Eighteen months in, the programme has a name, a logo, a target operating model and a slide showing four workstreams in green. Attrition is flat. The metrics that were meant to move have moved, some of them considerably. Nothing anybody does on a Tuesday is different from what they did on a Tuesday two years ago.

That gap is the subject of this piece. These articles have been working through leading as a practice rather than a personality: what a leader can and cannot cause, what a crisis does to the distinctions a group relies on, why the word urgency was the wrong one, why nobody turns up for your problem, and why a message sent is not a message received. The claim underneath them is that leading is taking a position on how change resolves. You do not start the event and you rarely choose its timing. What falls to you is the resolution: whether the disturbance ends in something released, or in somebody blamed, or in a programme that absorbs it and leaves the organisation exactly as it was.

The nine methods below consolidate that into things you can observe this month and act on next. They group under the three families of question that run through every phase of this series and change only which thinker holds them: Identity, Information and Interaction, which ask who the organisation takes itself to be, what it can see, and how its parts talk to each other. In transforming, Identity asks who is leading and what they can actually do. Information asks what can be said, heard and counted about the change. Interaction asks what makes people move. Michael Jackson holds the first, Drucker the second, Saul Alinsky the third. Girard runs underneath all nine, because most of what organisations call transformation is imitation of a competitor, and mimetic desire is the instrument that tells you which is which.

Two questions sit under the nine. What is this change a response to? And how is it going to resolve? An organisation that cannot answer the first is running a programme rather than a transformation. An organisation that has not answered the second has left the answer to whoever is most frightened.

Identity: Who Is Leading, and What They Can Do

The Identity questions here are about the leader before the intervention: the method they arrived holding, the difference between what they can cause and what they can only make survivable, and the discipline of not filling the silence. Jackson is the foundational thinker. The critical systems thinking he set out in Critical Systems Thinking and the Management of Complexity in 2019 begins from an uncomfortable premise: every systems method carries assumptions about the kind of situation it faces, works well inside that kind and badly outside it, and none of them is neutral. You do not observe your organisation and then select an approach. You arrive with an approach, and it has already decided what you are able to notice.

1. Name the method you brought with you. Before diagnosing anything, write down which method you are using and what it renders invisible. A leader trained in operating models sees structure and misses meaning. A leader trained in coaching sees relationships and misses the incentive that makes those relationships rational. A leader trained in delivery sees dependencies and cannot see why a technically complete programme changed nothing. Jackson’s answer is pluralism, holding several methods and choosing between them deliberately rather than defaulting to the one that made your career. The method: write two sentences before your next diagnosis, the first naming the lens you are using and the second naming what that lens cannot show you. Then ask a colleague who works differently what they see in the same situation. The gap between the two accounts is the part of the problem you were going to miss.

2. Separate what you can cause from what you can only make survivable. Kotter taught a generation to start by creating urgency, and the word has not aged well, because manufactured urgency is what organisations reach for when no event has occurred. A real event throws received wisdom into question and nobody has to be persuaded that something has happened. Urgency by memo asks people to feel something the situation has not given them, and they can tell. What a leader can do is different and smaller and harder: keep the disturbance at a heat the organisation can survive without resolving it early, which is Heifetz‘s work in one line. The method: for the change you are leading, list what you can genuinely cause and what you can only hold open. If the first list is long, you are probably describing a project. If the second list is empty, you have not started.

3. Leave the silence. A leader hears a problem and answers it, usually within a couple of seconds, usually competently, and every time this happens the problem travels back up. The coaching tradition Lyssa Adkins set out in Coaching Agile Teams in 2010 makes a discipline of the pause: the few seconds after someone stops talking, in which the person with the problem either finds their own next move or hands it over. Whoever speaks first owns it. We are all faster than we think we are, and most of us have never counted the seconds. The method: in the next three conversations where somebody brings you a problem, wait, visibly, before responding. Then ask what they think should happen. Count how often you get an answer that was already there.

Information: What Can Be Said, Heard and Counted

The Information questions are about whether the change exists in the organisation’s language and its numbers, or only in the deck. Drucker is the foundational thinker, and his sharpest claim about communication is the one most often ignored: communication is an act of the recipient, not of the sender. Nothing has been communicated until somebody has received it, in their own terms, in a form they can use. The utterance was never the thing.

4. Check who received it. Ask three people at different distances from you, in separate conversations, to say what the change is and what it means for their work. Do not correct them. Write down what they say. The variance between the three accounts is your actual communication, and the number of them who describe the change in terms of their own work rather than yours is your actual reach. Argyris explained why this is uncomfortable enough that most leaders skip it: the people best at defending an unexamined position are usually the ones who most sincerely believe they have communicated. The method: run the three conversations, then rewrite the message from the words that came back rather than the words you sent.

5. Listen for the vocabulary nobody issued. After a genuine event, a new language installs fast, and it installs from below. People start using a phrase for the new thing that was in no announcement. The presence of that vocabulary is the strongest evidence you will get that something has actually happened, and its absence is the strongest evidence that nothing has, whatever the engagement scores say. Weick understood organisations as continuous sensemaking, and sensemaking leaves a residue in the words. The method: for one month, collect the phrases people use for the change that came from nobody official. If the list is empty, the change is still yours alone. If the list is full of the previous programme’s vocabulary, you are being absorbed.

6. Ask which decision each rising number changed. Delivery metrics are almost all up. Deployment frequency, lead time, the volume of change, the counts on the dashboard you review monthly: up, and rising faster since the tooling arrived. Take each number that improved and name the decision that got made differently because of it. Where a number moved and no decision changed, it is not a measure, it is a mirror, and the organisation is admiring itself. Beer‘s instrument applies here with force: the purpose of a system is what it does, and a measurement system that reliably produces reassurance has reassurance as its purpose. The method: take last quarter’s improved metrics and, for each, write the decision it changed. Retire the ones with a blank beside them.

Interaction: What Makes People Move

The Interaction questions are about the conditions under which people actually act, which are rarely the conditions described in the change narrative. Alinsky is the foundational thinker, and the reason to read a community organiser writing in 1971 rather than another management author is that he had no authority whatsoever and could not confuse compliance with movement. His starting point is that people act on their own interests, in their own experience, and that an organiser who cannot state a change in the language of somebody else’s self-interest has not yet begun.

7. Write the sentence in which this is in their interest. For each group whose behaviour has to change, write one sentence in which the change makes their working life better, in words they would use. Not the organisation’s benefit; theirs. If you cannot write the sentence for a group, you do not have that group, and no amount of communication will get them, because there is nothing there to receive. Alinsky would add the harder half, which is that where the change genuinely is not in their interest you must say so plainly and negotiate rather than dress a loss as an opportunity. People forgive a cost named honestly. They do not forgive being sold one. The method: write the sentences, one per group, and show each one to somebody in that group.

8. Look for the rituals, not the training. Everyone has already done the training. The leadership programme, the values workshop, the two days on psychological safety: they exist, they were attended, and they changed the language for about a fortnight. Collins explains why. Groups are made in repeated encounters where people are together, attending to the same thing, with something at stake, and what those encounters produce is energy and standards the group will defend on its own. A training day has no stakes. It has a facilitator. The method: list what recurs in your organisation weekly or fortnightly with a real outcome attached, and ask whether the change you are leading appears in any of them. If it lives only in all-hands presentations and a quarterly offsite, it has no ritual and it will not survive contact with a busy month.

9. Name what the change ends. A transformation that only adds is an initiative. Ask what stops: which report, which committee, which product, which promise to a customer you have been keeping badly for years. Ackoff put the general case sharply, that the highest move is to dissolve a problem by redesigning the system that generates it, and dissolution always ends something. Most organisations I have worked with cannot answer this question, and the inability is the finding. An organisation that cannot stop anything has no capacity for change, only capacity for accumulation. The method: publish the list of what ends, with dates, before the list of what begins.

How It Resolves

Girard’s account of a group in crisis, worked out across Violence and the Sacred in 1972 and everything after it, is the least comfortable idea in these articles and the most useful. When the distinctions a group relies on collapse, and everyone wants what everyone else wants because they want it, the tension has to go somewhere. The oldest resolution, and the cheapest, is to find somebody to carry it: a function, a supplier, a leader, a team that was already unpopular. The organisation feels better immediately. Nothing has been learned and the same crisis returns on a schedule.

We can watch the mimetic half of this happening in public at the moment, and we in technology are the ones doing it. Every large organisation is building the same thing, on the same timeline, with the same three vendors, and very few of us can say what event in our own situation made it necessary. That is not strategy, it is imitation of a rival, and imitation is fragile in exactly the way Christensen‘s incumbents were fragile: every step is defensible and the direction was never chosen. Methods one, two and seven are the ones that catch it, because each asks whose change this is.

The nine do two jobs. Methods one, two, three and five test whether anything has actually happened: whether the leader has seen the situation or their method, whether the disturbance is real or manufactured, whether the organisation is generating its own language for it. Methods four, six, eight and nine test how it is resolving: whether the message has landed anywhere, whether the numbers are informing decisions or providing comfort, whether the change has a ritual to live in, and whether anything is being let go. Method seven tests both, since a change nobody else wants has neither happened nor resolved, whatever the programme board decides.

Transform comes from transformare, to change the shape, and the subject of the verb never changes: the thing that gets a new shape is the thing you already have. You cannot transform an organisation by creating another organisation, which is what a transformation office is, and it is why the successful ones dissolve themselves early and the unsuccessful ones are still there in year four with a headcount to protect. The work is done in the organisation you have, in its meetings, with its people, in its language. The position a leader takes on how the change resolves is, in my view, nearly the whole of the job.

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

Every serious engineering decision is a transfer of cost to people who are not in the meeting. The migration deferred, the interface left broader than it needed to be, the model trained on data whose provenance nobody wrote down, the platform whose only operator knowledge lives in one person’s head: each one buys something now and posts the invoice forward. Discounting the future is not an occasional lapse in this trade. It is the default arithmetic, and it is built into how we approve work.

Derek Parfit spent the last part of Reasons and Persons, in 1984, taking that arithmetic apart. His target was the social discount rate, the practice of weighing a future harm less because it is further away, and his argument was blunt: remoteness in time is not itself a moral reason for anything. If a policy will kill people in two hundred years, the deaths are not less bad for being late. What Parfit could not do, and knew he could not do, was tell a person on a Tuesday what to actually do about it. Alasdair MacIntyre, writing three years earlier, had the mechanism that Parfit’s argument needs.

1. The discount nobody will defend out loud

Ask a technology leader whether a defect that harms users in five years counts for less than one that harms them next quarter, and nobody says yes. Look at how work is prioritised and the answer is unmistakably yes, at a fairly steep rate.

Some of that is legitimate. A financial discount rate exists because money can be invested, because the future is uncertain, and because we may be richer later and better able to fix things. Parfit accepted all of that and separated it cleanly from the thing being smuggled in beside it: a moral discount, in which the interests of people further away in time simply count for less. The first is a claim about resources and probability. The second is a claim about the worth of persons, and once it is stated in those words nobody will sign it.

Most organisational discounting is the second, defended with the reasons that justify the first. The technical debt register is the artefact where you can watch it happen. Everything on it was recognised, written down, agreed to be real, and then weighted at nearly zero because its consequences fall outside the horizon anybody is measured on.

2. Where the arithmetic runs out

Parfit is worth reading precisely because he does not let his own side win easily. Two of his results block the obvious move of just doing the sums on future welfare.

The first is the non-identity problem. Decisions made now determine who exists later, not just how well they live, so the people who inherit a depleted world are people who would never have existed had we chosen differently. They cannot straightforwardly claim to have been harmed by the choice that made them. That is philosophically devastating. It is also immediately recognisable. The engineers who will maintain the system you are approving are people who will be hired because of it, into roles it creates. They cannot be consulted, and in an odd sense they cannot be wronged by the arrangement that summoned them.

The second is the repugnant conclusion. Aggregate welfare far enough and total utilitarianism recommends an enormous population of lives barely worth living over a smaller flourishing one, which nobody accepts and which no one has yet decisively escaped. Parfit worked at it for the rest of his life.

What this establishes is a real and slightly uncomfortable position: the obligation to the future is genuine, the temporal discount is indefensible, and impersonal aggregation will not tell you what to do. Obligation without guidance. That is where most corporate ethics statements sit. It is why they change nothing.

3. What a practice carries that a calculation cannot

MacIntyre begins After Virtue with the claim that modern moral language is fragments of an older scheme, rearranged and no longer working, so that public argument consists of incompatible assertions with no way to settle them. Anscombe had opened that line in 1958, arguing that we kept the moral ought after abandoning the framework that gave it force. MacIntyre’s constructive answer is the practice.

A practice, in his precise sense, is a coherent form of activity with goods internal to it, standards of excellence that partly define it, and a history. Medicine is one. Farming is one. Software engineering is one, on a good day. The internal goods are the ones you can only get by doing the thing well: the satisfaction of a design that holds up, a system a stranger can understand. The external goods are money, status and promotion. Those can be had by doing the thing badly. That is why practices are always under pressure from the institutions that house them.

The thing to notice for Parfit’s problem is what a practice does with time. Standards of excellence outlive every individual practitioner. They arrive from people you never met and they leave in the hands of people you will not meet, and the obligation each practitioner carries is stated in a form small enough to act on: leave the practice at least as capable as you found it. That is not a calculation about future welfare. It is a duty to a craft, and it discharges obligations to strangers that no one could ever compute.

4. Each corrects the other

Held on its own, either idea fails in a way you can watch happening in real organisations.

Virtue without the long view becomes a guild. A tradition confident in its own standards drifts into protecting the practitioners rather than the practice, and every professional body in history has done a stretch of this. MacIntyre knew it; he wrote that a living tradition is an argument extended through time, including an argument about what the tradition itself is for, and a tradition that has stopped arguing has become an interest group. Parfit’s impersonal test is the external check that keeps it honest: never mind what our standards say, who bears the cost of this and does their distance from us make it acceptable? That question comes from outside the guild, and a guild cannot generate it.

Aggregation without a practice produces the ethics committee. Real concern, sound reasoning, no transmission mechanism. No engineer whose Tuesday is changed by any of it. The gap between the two is precisely where every corporate responsibility programme I have watched has failed. The values were not wrong. There was no practice carrying them, so they lived in a document and the practice went on doing what it had always done.

5. Who is the entity that owes

Parfit has one more result that lands directly on this, and it is the one that unsettles people most. His account of personal identity is reductionist. Psychological continuity and connectedness are what carry a person forward, not some further deep fact of the self, and identity itself is not the thing to care about. Push that into an organisation and the awkwardness is obvious. The company in ten years shares almost no people, no code and possibly no market with the company now. In what sense is it the same thing, and how can it be bound by a promise this one makes?

MacIntyre’s answer is narrative. A life, and by extension an institution, is intelligible as a story with a subject who can be asked to account for it, and the continuity that matters is the continuity of the story and the practices inside it rather than of the substance. This is, I think, the more useful answer for a leader. It identifies what has to be maintained: not the entity, which will change entirely, but the practices, the standards and the people who carry them. Every departure of a senior engineer with no apprentice is a break in that continuity, and it is a bigger break than a reorganisation.

6. What it looks like on a Tuesday

The practical translation is unglamorous and it is already familiar from this series, which is the point. The obligations to people you will never meet are discharged through the ordinary standards of the craft.

Write down the intent behind a system, because the people who inherit it cannot ask you. Keep it small enough to be understood, since Ousterhout‘s complexity is a cost paid entirely by successors. Make it possible to remove things, because a system that can only accumulate, as Parnas saw, hands the next generation a problem it cannot solve. Put people’s hands on the difficult work, since Sennett‘s craftsman is made in the doing and a practice with no apprentices has one generation left. None of these requires a view about population ethics. All of them are how the far future gets served by somebody who has a sprint to finish.

The two thinkers meet on a single sentence, and it is the one worth taking away. You cannot calculate your obligations to the people who come after, and you can keep the practice they will inherit in a condition you would have been glad to receive.

Placement: Transforming, the fourth of the series’ four phases, after Learning, Deciding and Building. The capacity this phase produces is release, the willingness to let a change carry on without you and be revised by people who were not there. An obligation to inheritors is what release looks like when it is taken seriously rather than described.

(An Organisational Prompt is something you can do now....)

Organisational Prompt

Name the inheritors.

Take the largest system your teams are building this year. On one page, write who inherits it in five years: the engineers who will operate it, the team that will extend it, the customers who will depend on it, the person who will be woken at three in the morning by it.

Then write the one thing you are currently planning to leave them that you would be angry to receive from a predecessor. There is always one and everybody on the team can name it in about four seconds.

Fix that one thing this quarter, and say plainly why you are doing it. Not for the roadmap, not for the risk register, but because the people who will hold this next are owed the same standard you were owed. That sentence, said out loud by someone senior, does more for a practice than any values document ever written.

Further Reading

Derek Parfit, Reasons and Persons (1984). Part Four contains the argument on future generations, the non-identity problem and the social discount rate; Part Three is the account of personal identity that makes the ownership question so hard.

Alasdair MacIntyre, After Virtue (1981), for practices, internal and external goods, and the narrative unity of a human life.

Elizabeth Anscombe, Modern Moral Philosophy (1958), the paper that reopened virtue ethics, free at https://www.pitt.edu/~mthompso/readings/mmp.pdf.

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

Tasks completed per developer, up 33.7 per cent. Pull requests merged per developer, up 16.2. Median time a pull request sits waiting for review, up 441. Bugs per developer, up 54. Incidents per pull request, up 242.7. Those five come out of one dataset, two years of telemetry across 22,000 developers and more than 4,000 teams, published by Faros this year under a title that does the work of a paragraph: the Acceleration Whiplash.

A board on which every number is rising is not reporting performance. What it reports is that the numbers have stopped pulling against one another, and the pulling was the point. The four DORA keys were arranged as two pairs in opposition, so that no team could buy speed with stability. The counterweight has come off the other pan, and most of us are still reading the number as though it means what it used to.

1. The opposition was the invention

Nicole Forsgren spent four years measuring software organisations before Accelerate appeared in 2018, and the durable result was never the four metrics. It was the shape they were arranged in. Deployment frequency and lead time for changes report how fast work moves. Change failure rate and time to restore service report what happens when it goes wrong. Two of each, deliberately, because the industry believed speed and stability traded against each other and the data said the reverse: the fastest organisations were usually the most stable ones too.

That finding held because both pockets were filled from the same purse. Every change cost an engineer the hours to write it, so throughput limited itself, and any shortcut taken to raise throughput surfaced in the failure numbers within a quarter or two. You could not run one number up without the other coming down. Speed and stability were never opposites; they were two readings of the same discipline.

2. Writing got cheap and reading did not

Generation changed the price of exactly one step in that loop.

The Faros telemetry has pull request size up 51.3 per cent and median review time up 441 per cent, against 91 per cent in the previous year’s dataset. It also has 31 per cent more pull requests merging with no review at all. Organisations rarely decide to stop reading code. They discover that they have. The cheapest way to keep the visible number moving was to keep the review step on the board and stop reading inside it, and that is the route the work took. Writing is cheap now. Reading is not, and reading is where the cost of writing went.

What is being bought is smaller than the story around it. Brian Houck’s first quarter data for 2026 puts AI at 27.4 per cent of production code on average, with adopting teams gaining roughly 7.8 per cent in pull request throughput. Real, measurable, worth having. Nobody wrote that number into the business case.

3. What your measurement system is actually for

Stafford Beer had a rule for reading any organisation without arguing about intentions: the purpose of a system is what it does. Turn it on the board. A metric set’s purpose is whatever behaviour it reliably produces. So if yours makes volume visible, leaves review expensive, and rewards volume while measuring nothing about review, then unreviewed volume is what your metric set is for.

It is working.

The defence we usually reach for at this point is maturity: strong practice, good tests, a platform, a real deployment pipeline. Faros went looking for that protection in the 2026 data and did not find it. Organisations with high baseline maturity show the same deterioration downstream as everyone else. Whatever is protecting you, it is not your maturity model.

4. Goodhart, quoted properly

The line everybody quotes belongs to Marilyn Strathern, who wrote in 1997 that a measure which becomes a target stops being a good measure. Charles Goodhart’s own formulation from 1975 is less quotable and more use. His claim was that an observed statistical regularity collapses once pressure is placed on it for control purposes.

Read the last decade of engineering measurement against that sentence. A research programme observed a regularity across thousands of organisations. The industry then adopted the regularity as a control target, printed it on quarterly slides and attached it to promotion cases. That alone is enough to break the relationship. No new technology required. Generation did not snap the four keys; it arrived at a set of measures already carrying years of control pressure, and leaned.

That reading is mine rather than DORA’s. Their own recent work is more careful than most of what gets written around it. The 2024 report put a 25 per cent rise in AI adoption alongside a 1.5 per cent fall in delivery throughput and a 7.2 per cent fall in stability. The 2025 report named the mechanism. AI is an amplifier, and what it amplifies is whatever the organisation already was.

5. Rebuilding the counterweight by hand

Abi Noda and Laura Tacho published the DX Core 4 in December 2024, with Forsgren, Margaret-Anne Storey, Michaela Greiler and Thomas Zimmermann around the work. It folds DORA, SPACE and DevEx into four dimensions: Speed, Effectiveness, Quality and Impact. Speed is headed by diffs per engineer. Effectiveness is the Developer Experience Index, a fourteen-item survey. Quality is change failure rate. Impact is the share of engineering time going into new capabilities rather than maintenance.

Diffs per engineer is the provocation and they know it. Their conditions are explicit: never at the individual level, never a target and never tied to a reward, always counterbalanced by an oppositional measure. They state the design principle in the open, which is rarer in this trade than it should be: the dimensions exist to hold each other in tension, so that speed cannot be bought with developer experience. What Core 4 rebuilds by hand is the balance that came free until 2024, and it rebuilds it as a rule rather than a happy property of the work.

Houck’s 2026 revision adds a distinction worth stealing whatever framework you run. Keep diagnostic telemetry apart from outcome metrics. Adoption rates, token spend and tool coverage tell you what is being consumed, and they tell you nothing about whether anything got better. An input sitting on the same slide as an outcome tends to be read as one. Adoption is not a result. It is a bill.

6. The number nobody has

Neither framework measures complexity.

Ousterhout named the quantity from inside the code: what a developer must hold in their head to change something safely. Skelton and Pais named the same quantity at the scale of a team and made it the constraint on how you draw boundaries. Nobody has a number for it, and it is the quantity that has moved furthest since 2024. Not DORA, and not Core 4.

Everything worsening on your board is its shadow. Review takes longer because there is more to hold. Bugs rise because less of it is being held. Incidents rise because what was held was held by nobody in particular, which is what shallow ownership feels like at two in the morning. More to hold, less of it held, and no number anywhere on the board that says so.

Three proxies are available from tooling you already own: median pull request size, median hours from opening to first human comment, and the share of a team’s services any single engineer will change unsupervised. None of them is complexity. All of them move when complexity moves, and that is enough to run a serious conversation on, which is more than your board is giving you now.

7. Where to stand

Taiichi Ohno took reports seriously and facts more seriously, and the difference between the two was a walk to where the work was happening. Numbers tell you where to go and stand. They do not tell you what is happening once you are there, and no board has ever produced the sentence you need most, which is somebody saying out loud that they approved a change they did not read.

That board was designed for an organisation in which writing was the expensive step. Reading is the expensive step now. Not one of the four keys knows it.

Placement: Building, the third of the series’ four phases, which run Learning, Deciding, Building, Transforming. Measurement sits on the Interaction question, the one about how teams, components and agents relate to each other. Conway returned here to report what team structure had already done to the architecture. Measurement is where that report either turns into numbers or does not.

(An Organisational Prompt is something you can do now....)

Organisational Prompt

Put the review queue on the same slide as deployment frequency.

Take one team and its last fifty merged pull requests. For each one, record four things: lines changed, hours between opening and the first human comment, whether any human comment exists at all, and roughly how much of the diff was generated.

Three numbers come out. Median size. Median wait. The count that merged with nobody’s comment on them.

Put those three beside your deployment frequency and change failure rate, on one slide, for the same period. If deployment frequency is up and any of the three has worsened, you have bought throughput with review, and the purchase was made for you rather than by you.

Then say which of three things you are doing about it, out loud, in a meeting that ends in a decision: pay for review properly, cut the size of what has to be reviewed, or accept unreviewed merges in a named part of the estate for a named period. Doing the third by accident is the only option not available to you. Review is either on the board or it is not, and review that is not on the board is not happening.

Further Reading

Accelerate (2018), by Nicole Forsgren, Jez Humble and Gene Kim. The 2025 State of AI-assisted Software Development report and its companion AI Capabilities Model are free at https://dora.dev/publications/.

Abi Noda, Laura Tacho, Margaret-Anne Storey and Michaela Greiler: Measuring developer productivity with the DX Core 4, free at https://getdx.com/research/measuring-developer-productivity-with-the-dx-core-4/. The benchmark data behind it is published too.

Faros: AI Engineering Report 2026: The Acceleration Whiplash, summarised at https://www.faros.ai/blog/key-takeaways-from-the-dora-report-2025. Telemetry rather than survey, which is why it disagrees so sharply with how developers report they feel about the same work.

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

Copilot. Agent. Human in the loop. Adoption. Four words, none of them chosen by anyone in particular, and between them they have already settled how most organisations think about the largest change to arrive in the trade in thirty years. A copilot sits beside you while you fly the aircraft. An agent acts on your behalf, which implies it can be given a task and left. A human in the loop is a step inside a machine’s process rather than a person the machine works for. Adoption is something you measure the rate of, so it sounds like an outcome, when what it describes is a bill.

George Lakoff has spent since 1980 explaining why none of that is decoration. Frames decide which facts count as relevant, which inferences follow without anybody drawing them, and which sentences are even sayable. Get the frame in first and the argument is mostly over. So the naming of a change does the work the announcement takes credit for. And the naming is done carelessly, by whoever wrote the first email.

1. Frames are not spin

The standard managerial view of language is that it is packaging: you decide the substance, then you decide how to communicate it, and communication is what you do to the substance afterwards. Lakoff’s work, from Metaphors We Live By with Mark Johnson in 1980 onwards, says the order is wrong. We think in metaphor. Argument is war, so positions are attacked and defended and somebody wins. Time is money, so it is spent, saved and wasted. Nobody selected those and everybody reasons inside them.

The consequence for a change programme is unwelcome. A frame is built by long repetition and holds because the world keeps confirming it, so facts that do not fit it are not weighed and rejected. They bounce. That is why your carefully evidenced deck moved nobody. The evidence was fine. The frame it landed in had already assigned it a meaning.

2. Negation invokes

Lakoff’s most quoted demonstration is the elephant. Told not to think of an elephant, you think of an elephant, because a frame is activated by the words that name it whether they affirm or deny. Now count how many times your organisation has said, in writing, that this is not about replacing anyone. Each repetition installs replacement as the subject under discussion.

The same trick runs through the tax relief framing he analysed for years: relief presupposes an affliction and somebody who lifts it, so the argument is lost before it starts by anyone who accepts the word. If you find yourself denying a frame, you have adopted it. Say the thing you do mean, in its own words, and let the denial go unmade.

3. The window opens after the event

Frames are slow. That is the rule, and it is why most internal communication achieves so little. You are pushing against structures laid down over years by everything the organisation has lived through. The exception is short and it is the one a leader can use. When an event has knocked the supports out from under the old frame, and people can see for themselves that the old account no longer explains what just happened, new frames install with startling speed.

Girard describes the crisis in which the distinctions a group lives by collapse. What follows that collapse is a naming contest, and it is settled within a fortnight, by whoever gets to the all-hands first. Whatever the organisation starts calling the thing in those two weeks is what it will believe about the thing for years, and the people who win the contest are rarely the people who thought hardest about the wording.

4. What our words about AI already decided

Take the four from the opening and read what each licenses. Copilot frames the model as an assistant to a pilot who remains in command. It is comforting, it keeps accountability where it was, and it implies the human is still flying when the human is now mostly reading.

Agent frames it as a worker given tasks. That licenses delegation and imports the whole vocabulary of command. Command is the wrong frame for a thing that can only promise its own behaviour, as Burgess argues at length. Human in the loop puts the machine’s process in the main clause and the person in a subordinate one; it is a phrase from control engineering and it has been doing quiet work on how organisations reason about responsibility ever since it escaped. Adoption frames consumption as achievement, which is why adoption rates keep appearing on slides where an outcome ought to be.

None of these was selected. All four arrived from vendors, papers and posts. Each has been shaping design decisions, org charts and governance documents in organisations that believe they are reasoning from first principles. We in technology are particularly bad at this, because we think of ourselves as people who deal in mechanisms rather than words, which makes us the easiest audience in the world for a metaphor arriving inside a product name.

5. Choose the governing metaphor on purpose

The remedy is not to purge metaphor, which cannot be done, but to pick one and know what it costs. Every metaphor lights part of the thing and hides the rest, so the test for any candidate is two questions: what does this license, and what does it hide?

Try one. Treat the model as a very fast, very confident graduate who has read everything and worked nowhere. That frame licenses supervision, review, apprenticeship and a named human owner, all of which you want. It hides the novel parts, the scale and the speed, and no graduate can be run in parallel a thousand times, so it will need correcting on those.

Compare it against agent and see which one produces better decisions about who is answerable when something breaks at three in the morning. The winner is not the more accurate metaphor, because neither is accurate. It is the one whose hidden part you can afford.

6. Listen for the words nobody issued

The strongest evidence that a change has actually landed is linguistic and it comes from below. People start using a phrase for the new thing that was in no announcement, no deck and no training. The phrase is usually blunt, often faintly rude, and it is the sound of an organisation building its own account of what happened.

Collect those. If the list is empty after a quarter of effort, the change exists only in your language and nobody has received it. If the list is full of the previous programme’s vocabulary, the old frame has absorbed the new thing and is running it as a subsidiary. And when a phrase does emerge that captures what you meant better than your phrase did, take theirs. The frame that holds is the one people already use.

Placement: Transforming, the fourth of the series’ four phases, after Learning, Deciding and Building. This one sits at the Language position, in the short window after an event when what a thing gets called is still open. Naming it well is not communication. It is the structure everything afterwards gets built on.

(An Organisational Prompt is something you can do now....)

Organisational Prompt

Audit five words.

Write down the five phrases your organisation uses most often about the change you are leading. Take them from what people actually say in meetings, not from the strategy document.

Beside each one, write two lines: what this word licenses, meaning what it makes reasonable to do next, and what it hides, meaning the question it stops anybody asking. Copilot licenses handing over the typing and hides who is reading. Adoption licenses celebrating a rate and hides whether anything got better.

Then change exactly one word, and change it where repetition happens: the weekly note, the standing meeting, the template everyone copies. Not in a campaign about language, which will fail and deserve to. Say the new word four times in the places people hear things four times, and watch whether it comes back to you in a month from someone who thinks it was theirs.

Further Reading

George Lakoff and Mark Johnson, Metaphors We Live By (1980), the book that made conceptual metaphor a serious subject and still the fastest way into the argument.

George Lakoff, Don’t Think of an Elephant (2004), short, political, and the clearest statement of why negating a frame reinforces it.

Lakoff’s essays and lectures on framing are collected at

https://georgelakoff.com

including the material on how frames are entrenched by repetition rather than argument.

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

Two teams in the same company. Same stack, same hiring bar, same director two levels up. One of them ships steadily, argues well in the open, and turns a competent joiner into a good engineer inside a quarter. The other has the stronger people on paper and cannot get out of its own way. Every technology leader has met this pair, and the words we reach for are culture, leadership and luck. All three name the thing without accounting for it. What is the mechanism?

For fifty years Randall Collins worked on that gap, and his answer is uncomfortable in a useful way. The difference sits in the situations people pass through rather than in the people, and situations either charge people or flatten them. Move an engineer from the second team to the first and in six weeks you have a different engineer with the same skills. Nothing about her changed. The rooms did.

This phase of the series asks three questions about building: what a builder is and what a builder can promise, what a builder has to know, and what holds a team together while it works. Burgess answered the first. Ousterhout answered the second. Collins answers the third, with a theory that was never about software at all.

The last article left five practitioners agreeing on one finding: the input mode has changed, the discipline has not, and the load has moved to review. Review is not a stage in a pipeline. Two or more people hold one object in their attention at the same time and arrive at a shared judgement about it. It remains the oldest social technology we have, and the one currently under the most pressure.

The interaction lever has changed register twice already in this series. Illich asked whether an institution’s tools serve the activity or quietly take it over. Beer asked whether the channels could carry the variety the situation was throwing at them. Both are design questions, answerable at a whiteboard. Building asks something smaller and much harder to fake. When people are making a thing together, what is the actual mechanism by which they cohere, and what happens to that mechanism when the making stops passing through them?

Collins was born in 1941 and, as an undergraduate at Harvard, heard Talcott Parsons lecture on Durkheim’s account of religion, in which what holds a congregation together is the ritual rather than the belief. He describes that lecture as the revelation that set the direction of everything after it. A psychology master’s at Stanford followed, then a doctorate at Berkeley in 1969 under Herbert Blumer, Erving Goffman and Reinhard Bendix. He is Dorothy Swaine Thomas Professor of Sociology at Pennsylvania and was president of the American Sociological Association in 2010. Conflict Sociology came in 1975, The Credential Society in 1979, The Sociology of Philosophies in 1998, Interaction Ritual Chains in 2004, Violence in 2008, and Napoleon Never Slept in 2015.

Two of his teachers are already in this series. Goffman gave us the interaction order in the learning phase. Parsons gave us the systems view of institutions. Collins took Goffman’s close observation of face-to-face life and turned it into a mechanism with inputs and outputs, which is what makes him usable on a Monday morning.

1. The situation is the unit, not the person

Sociology usually starts with individuals and builds up to groups. Collins inverts that. His unit of analysis is the situation: a bounded encounter between people, with a beginning and an end. Individuals are what their chain of situations has made them, and they are being remade continuously, including this week.

That sounds abstract until you apply it to hiring. If the person is the unit, a strong team is a procurement problem: find the good ones, pay for them, assemble them. Every organisation I have watched has run that experiment at some point, and the results are famously unreliable. Collins tells you why. The engineer who was extraordinary at her last company was extraordinary inside a chain of situations, and you bought her without buying any of them. She arrives, the situations here are different, and within two months everybody privately concludes that the interview process failed.

It also explains the thing managers find hardest to say out loud. You do not have good people and bad people. You have situations that charge people and situations that flatten them, and the same person passes through both in a single week. When you describe someone as low energy you are usually describing where they have been sitting.

The practical consequence is that the lever is not the person, and it is not the culture either. It is the design of the recurring situations that person is in. Situations are a far smaller thing to change than a culture, and you can start on them immediately.

2. Four ingredients, and most of your meetings have two

Collins names four ingredients, and all four have to be present at once. Miss one and the gathering still happens. The group does not.

Bodily co-presence: people in each other’s physical or attentional field, aware of one another as bodies rather than as names on a list. A barrier to outsiders: everyone knows who is in this and who is not. A shared focus of attention: at any given moment everyone is attending to the same thing, and knows that the others are too. And a shared mood: an emotion that converges across the group rather than being felt privately by each person.

Test your standing meetings against those four. A daily standup where twelve people report serially to a facilitator has co-presence of a sort and nothing else. Each person attends to their own turn, and there is no moment at which the group looks at one thing together. A quarterly planning session has a barrier and no mood. A recorded all-hands has neither.

Then test an incident call, which passes on all four and passes hard. Everyone is present and knows exactly who else is. The boundary is sharp; you are on the call or you are not. There is one graph and everyone is looking at it. The mood is unmistakable and it is shared. This is why teams bond in incidents and do not bond in planning, and why people who have been through a bad night together will do things for each other for years afterwards. Nobody designed that ritual. In many organisations it is the only one running properly.

The uncomfortable ingredient is the barrier. Open invitations, optional attendance and a recording for anyone who missed it are all reasonable and considerate policies, and every one of them weakens the ritual. A gathering that anyone can drop into is a broadcast. There is nothing wrong with a broadcast, and you should stop expecting it to produce a group.

3. Emotional energy is a disposition to act

The currency Collins tracks is emotional energy, and the phrase invites a soft reading, so take his definition exactly. On page 49 of Interaction Ritual Chains he calls it a sustained feeling of confidence, elation, strength, enthusiasm and initiative in taking action. The last three words carry the concept. The concept names a disposition to act, and you can observe it without asking anybody how they feel.

High emotional energy looks like this. Someone picks up an ugly ticket nobody assigned. Someone argues with a senior person in an open channel. Someone volunteers to own the migration, or starts the work rather than scheduling a meeting about starting the work. Low emotional energy looks like waiting: perfectly polite, perfectly professional waiting, with every process followed.

Because it is produced in situations and carried out of them, emotional energy is not a personality trait. Leaders get this wrong, and it is expensive. The same person is a fountain on one team and flat on another. You did not hire a fountain and you did not hire a flat person. You put a person into a chain. The chain did what chains do.

So the management question at the end of any recurring meeting is not whether people found it useful. Did they leave more likely to act? Morale is what you measure in a survey. Emotional energy is what you can watch happening at four o’clock on a Thursday, and the two frequently disagree.

4. Why the codebase used to be sacred

When the four ingredients combine, Collins says you get four outputs: solidarity in the group, emotional energy in the individual, symbols charged with group significance, and standards of morality that make violating those symbols feel wrong.

The third and fourth outputs are the ones that matter for building. Every engineering team you have admired had sacred objects. The green build. The runbook. The naming convention that somebody defended in a review three years ago (nobody now remembers who) and everyone still follows. Nobody enforces these, and that is precisely the point. They are charged, in Durkheim’s sense, by the rituals that produced them, and violating them feels wrong well before anyone works out which argument to make.

This gives you an account of something engineers report constantly and struggle to justify. The flatness people describe about generated code is a symbol problem. A thing that nobody assembled together carries no charge. When the codebase stops being something the group made in each other’s presence, it stops being a sacred object, and the standards that were held in place by a feeling of wrongness lose the thing that was holding them up.

That is where standards actually go. They do not go in a decision, and they do not go in a memo. They dissolve quietly when the ritual that charged them stops running, and the first anyone notices is a review comment that nobody bothers to make. The measured collapse in refactoring that the last few articles pointed at has a cost explanation and a Collins explanation, and I think you need both of them. Tidying got cheaper, and at the same time it stopped feeling like something you owed anyone.

5. Thinking happens in rooms, and the chain is the asset

The Sociology of Philosophies is Collins’s largest book and the one most worth an executive’s afternoon. He traces two and a half thousand years of intellectual creativity across Greece, China, India and Europe, and finds the same structure everywhere. Ideas are produced in networks of face-to-face contact between rivals. The great names do not work alone; they cluster, they teach each other, they argue with people they know personally, and the chains of contact run for generations.

Out of that comes his law of small numbers. In any field, the count of actively contending positions stays small, roughly between three and six, because attention is a limited resource and a field can only sustain so many live arguments at once. Rivals need each other. Nobody can hold a position in a space where a hundred positions are being asserted, so the field self-organises into a handful.

Both findings land directly on engineering. Your architecture debate has room for about four live positions; past that, people stop attending and start waiting for the decision, which is why the review with nine competing proposals produces less thinking than the one with three. And the deeper claim is the one to take into any argument about remote work, asynchronous review or autonomous agents: on the historical evidence, co-presence produces thinking, and not merely comfort. The claim is a strong one. Collins spent six hundred pages on it, and it deserves better than being waved away as sentimentality about offices.

The word chains does the rest of the work. Each situation is stocked by the ones before it: the symbols, the shared references, the accumulated knowledge of who is reliable under pressure. A team’s capacity to hold its attention together is an asset that was built up over time, and it can be spent as well as built.

Which reframes the reorganisation. When you split a team and redistribute its people, the slide shows headcount moving and capability preserved. What moved is people. What did not move is the chain. The new group starts accumulating from close to zero: no shared symbols, no history of joint attention, no established sense of what the mood in the room means. That recovery takes a couple of quarters and is never planned for, because the asset you destroyed was never on any register.

It also gives a better account of onboarding. A new joiner is slow because they have no chain, and information alone will not supply one; you can hand over information in a week. They have not yet been in the room when the group looked at something hard together, so the symbols mean nothing to them and they cannot read the mood. Every senior engineer knows the moment when a new person becomes one of us, and it is always a situation rather than a milestone.

6. What an agent cannot do, and what that costs

Now the part this phase has been building towards. What happens to a ritual when one of the participants is not having an experience?

An agent supplies output. It cannot supply co-presence, it cannot share a mood, and it cannot receive emotional energy or generate any. This is not a complaint about the technology. It carries no prediction that the tools will stop improving. Three of Collins’s four ingredients require another party who is also having an experience, and capability does not supply that.

Three consequences follow. None of them appears on a delivery dashboard.

The number of situations falls while output rises. Work that used to pass through two people now passes through one person and a model. The pairing, the design conversation at the whiteboard, the fifteen minutes of arguing about an interface: those were rituals, and they have been quietly optimised away, because the value in them was never the artefact they produced.

Shared focus fragments. Each engineer now works inside a private context: their session, their prompt history, their accumulated instructions to the model. Attention has never been more concentrated and never less shared. Two people on the same team can spend a fortnight in the same service without a single moment in which they were looking at the same thing together.

And review changes character. Review used to be a ritual with mutual focus: two people holding one object and reaching a judgement, with the reciprocity that makes a ritual charge rather than drain. Reviewing a large volume of generated code is closer to inspection. It runs one way, nobody is having the experience with you, and the person doing it comes out flatter than they went in. Five practitioners reported last time that the load has moved to review. On this account it moved onto the one activity that was carrying the social work, and we degraded that activity at precisely the moment we made it more important.

Collins is alive to the question himself. In an interview published in Theory and Society in 2024, following the inaugural meeting of the Norwegian Microsociological Network, he addressed whether bodily co-presence still carries the weight it did when he wrote the book, how online rituals compare with offline ones, and what artificial intelligence might do to the ritual landscape. The honest summary for our purposes is that the mechanism is not in question. What is in question is whether anything can substitute for the ingredients, and so far nothing has.

7. Solidarity is not the same as being right

The theory has a limit, and you should carry it into any argument where you use him.

Collins explains why groups cohere. He does not explain whether the group is any good. A team can be tightly bonded around a bad framework, a defensive practice, or a sacred object that should have been deleted in 2019, and the solidarity will be entirely real. High emotional energy in a team that is confidently building the wrong thing is worse than low energy, because it is faster.

A second limit is argued well by Boyns and Luery in 2015: the framework is underdeveloped on its negative side. Failed and hostile rituals also generate energy of a kind. The blame-heavy incident review, the war room that runs on adrenaline, the channel where people bond over shared contempt for another department: all four ingredients present, all four outputs produced, and the energy is corrosive. It charges the people in the room and burns the chain they are standing on.

So this lever needs its two companions. Identity asks who is actually accountable for the thing, which is where Burgess put the weight. Information asks what has been made visible to whoever arrives next. That is Ousterhout’s question. Interaction asks whether the group can hold its attention together long enough to answer either one. Collins is the lever people skip, because it looks like the soft one. It quietly determines whether the other two are answerable at all.

8. What this lever asks of you now

You cannot mandate emotional energy, and the attempt is visible from a long way off. What you can do is design the situations, and there are fewer moves available than you might expect, which is good news for anyone with a full calendar.

Run fewer rituals and give the survivors all four ingredients. One artefact, in front of everyone, at the same moment, with a clear boundary around who is in the room and who is not. A weekly session in which a team looks together at one thing it built will do more than five status meetings, and it will cost less.

Put the pairing where the work now is. If the writing has moved to the model, then pair on the review. Two people, one screen, one diff, one judgement, out loud. That restores the mutual focus the old ritual supplied and attaches it to the activity that now carries the risk.

Keep at least one artefact that the group makes in each other’s presence. It matters less what it is than that it exists: the architecture diagram, the incident write-up, the interface everyone argued about for an hour. Without a jointly assembled object there is nothing for the standards to attach themselves to, and standards that float free do not survive their first busy quarter.

And audit the rituals that drain. Every organisation has a recurring meeting that reliably sends people out flatter than they came in, and it survives because it is on the calendar and because cancelling it would look like a governance failure. On Collins’s account that meeting is doing active damage. Every session of it spends the chain, and you pay for the privilege.

The next article takes all of this into structure. Conway told us that the systems we ship copy the communication structures that produced them. Collins tells you which of those communications are real, because a channel that carries no shared attention transmits information and produces no group. And the telos of this phase, simplicity, turns out to depend on him in a way that is easy to miss. Joint attention is finite and it does not scale. A system that no group can hold in shared focus cannot be kept simple by anybody, whatever the tooling promises you.

(An Organisational Prompt is something you can do now, in your own organisation, with the ideas in the article.)

Organisational Prompt

Count the room, then count what leaves it.

Take one team and one fortnight. List every recurring situation in which that team is together: the standup, the review session, the planning meeting, the on-call handover, the demo. Before each one starts, mark which of the four ingredients is actually present. Are people in each other’s attentional field, or on mute with a second screen. Is it clear who is in and who is watching. Is there one thing everyone is looking at at the same moment. Does the mood converge. Or is each person alone with it.

Then stop assessing the meeting and start watching the twenty minutes after it. Do not send a survey. Watch. Who starts something. Who picks up work nobody assigned. Who raises a hard question in a channel. Who goes quiet and does not reappear until lunch.

Two weeks of that gives you a short list you will find hard to unsee. Any recurring situation with two ingredients or fewer, producing no action afterwards, is a chain you are paying for and not banking. Fix one ingredient or delete it, and do not replace it with a new meeting.

Then the question that costs something. Name the last artefact your team assembled together, in each other’s presence, in the past month. If you can name it, protect that ritual above everything else on the calendar. If you cannot, you have found the reason the standards have been slipping, and it was never a discipline problem.

Further Reading

Randall Collins, Interaction Ritual Chains, Princeton University Press, 2004. The four ingredients and the four outcomes are laid out in the first two chapters, and the definition of emotional energy is on page 49.

The future of interaction rituals, Collins interviewed by Lars Johannessen in Theory and Society, 2024, on co-presence, online rituals and what artificial intelligence may do to the ritual landscape.

Sinisa Malesevic and Steven Loyal, an introduction to the special issue on Collins, Thesis Eleven, 2019. The best short route into the whole body of work.

David Boyns and Sarah Luery, Negative Emotional Energy: A Theory of the Dark-Side of Interaction Ritual Chains, Social Sciences, 2015. Open access, and the strongest criticism in print.

Sociologizing with Randall Collins, a 2018 interview covering emotions, violence and the attention space.

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

In 1969 Drucker gave a talk in Tokyo that became chapter thirty-eight of his big management textbook. The chapter is called More Talk and Less Communication, and it opens with a riddle he takes from the Zen Buddhists, the Sufis and the rabbis of the Talmud. A tree falls in a forest and nobody is there. Is there a sound? Drucker says no. Sound waves, yes. No sound, because sound is created by perception, and perception is the whole of what communication is.

He sets out four fundamentals in that chapter, and says they were learned mostly by doing the wrong things. Fifty-seven years on, the wrong things are still being done, at greater volume and lower cost than Drucker could have imagined. Uttering is now free. Communicating is exactly as expensive as it ever was, which means the gap between what an organisation utters and what it communicates has never been wider.

1. It is the recipient who communicates

The first of them reorganises everything. Communication is perception. The recipient communicates; the speaker merely utters. That is no modest point about listening skills. The whole act moves to the recipient. The question before speaking is therefore not whether the message is clear, which is a question about the sender, but whether the message sits within the recipient’s range of perception. Can the recipient receive it? If not, no amount of clarity will help, and the sender has produced sound waves.

Every organisation has a mechanism for confirming that the sound waves went out. The email was sent. The town hall was recorded. The strategy is on the intranet. None of these tells you anything about the recipient. A leader who takes them for evidence has stopped measuring communication and started measuring utterance.

2. We hear what we expect to hear

The second explains why the memo that was received was still not the memo that was sent. We perceive largely what we expect to perceive. The unexpected is either not perceived at all or quietly bent into the shape of the expected, so the organisation perceives a familiar thing where an unfamiliar one was said. Announce a new operating model to people who have seen four, and they will perceive the fifth, because the previous four have trained them to expect it. They are probably not being cynical. They are perceiving normally.

Drucker’s instruction follows from this and it is uncomfortable. Find out what the recipient expects before you speak, because the expectation decides what the recipient can perceive. Then decide whether to use those expectations or to break them, and understand that breaking them is a separate and much harder piece of work than saying the thing.

This is also why the weeks after something genuinely happens are worth more than any communications plan. An event removes the evidence that held the old expectations in place. For a short period the recipient can perceive the unexpected. Most organisations spend that window drafting.

3. Communication makes demands

The third disposes of the idea that you can communicate neutrally. Communication always demands something. The demand is that the recipient become somebody, do something, or believe something. That demand lands where it fits what the recipient already wants, and fails where it appeals to nothing the recipient wants. Simply informing people of a change that will alter their work is not available as an option, and the attempt is usually heard for what it is.

Set that against the standard transformation announcement, which demands a great deal and appeals to nothing. The demand is that people work differently, learn new tools, accept new measurement and give up practices they are good at. The reason offered is the organisation’s competitive position. That is a demand with nothing on the recipient’s side of it.

4. Information is the opposite of communication

The fourth has aged into relevance. Communication and information, Drucker writes, are different and largely opposite, though each depends on the other. Communication is perception. Information is logic, formal and impersonal, carrying no meaning of its own. From which he draws the conclusion that more information can make the problem worse rather than better. He was writing when the information explosion meant mainframes and photocopiers, and he noted that the revolution of his day had not produced information at all. It had produced data.

An organisation can now generate more information than at any point in history, at almost no cost, and every piece of it is impersonal by definition. The AI-drafted summary of the AI-drafted strategy is information about information. Perfectly logical, and it demands nothing, appeals to nothing, and lands nowhere. Drucker’s fourth fundamental predicts precisely this outcome, several decades before the tooling existed to produce it at scale.

A corollary in the same chapter tends to get skipped. Downward communication cannot work. What travels downward are commands and prearranged signals, and those are received perfectly well. Understanding travels the other way, upward, from those who perceive towards those who want to reach their perception. So the answer to a communication problem is almost never another message.

5. A liberal art, which is not a compliment

Twenty years after the Tokyo talk, Drucker published the sentence that got him accused of going soft. Management, he wrote in 1989, is what tradition used to call a liberal art. Liberal because it deals with the fundamentals of knowledge, self-knowledge, wisdom and leadership; art because it is practice and application. Managers should draw on psychology and philosophy, on economics and history, on ethics and the physical sciences. And then the clause that stops it being a plea for a broader reading list: they have to focus that knowledge on effectiveness and results.

This ran against the consensus of its moment, which held that management was a science and that quantitative analysis was its method. Drucker never described himself as a management theorist anyway. He called himself a social ecologist, concerned with people’s existence in the man-made environment as a natural scientist is with the physical one, and only fifteen of his thirty-nine books were about management.

The claim reads differently now. The technical half of management, the part that is analysis and reporting and modelling, is the half being automated fastest. What that leaves is knowledge, self-knowledge, wisdom and leadership, which is the part Drucker was pointing at when everyone thought he was being vague.

6. Say it back

All four converge on one test, and it costs nothing to run. Take the last significant thing you announced. Ask four people separately, without notes, to tell you in one sentence what it means they should do differently. Four different answers means you uttered. Four answers that match each other but not your intent means they perceived something, just not what you sent. One or two people saying they were not sure it applied to them is the most useful result of all, because it tells you exactly where the range of perception ends.

The failure this catches is not a failure of writing. It catches a belief. The sending was never the event.

Placement: Transforming, the fourth of the series' phases; Learning, Deciding, Building, Transforming, and this is its Language position. Drucker's third appearance turns on the same distinction each time: what was said is not what was received, and after an event the naming that sticks is whichever one gets there first. The Event Is Not Yours to Make opens this stretch.

(An Organisational Prompt is something you can do now....)

Organisational Prompt

Four people, no notes

Pick the biggest thing you have announced in the last quarter. Ask four people at different levels, separately and without warning, to say in one sentence what it means they should do differently. Write the four answers down next to each other. Do not correct anybody, and do not send a clarifying note afterwards; that would be another utterance. Take the gap between the four answers as your actual communication result for the quarter, and spend the next one closing it upward rather than downward.

Further Reading

Drucker, Management: Tasks, Responsibilities, Practices (Harper and Row, 1973). Chapter thirty-eight, More Talk and Less Communication, is the source of the four fundamentals and is worth reading on its own even if the rest of the volume stays on the shelf.

Drucker, The New Realities (Harper and Row, 1989). Where management is named a liberal art, and where the naming is tied to effectiveness rather than to breadth for its own sake.

Drucker, Managing Oneself (Harvard Business Review, 1999). The self-knowledge half of the liberal art, written as a practical exercise rather than a philosophy.

The Drucker Institute at Claremont, and the Drucker Forum in Vienna. Both carry the management-as-a-liberal-art work forward, and the Forum’s annual proceedings are the best guide to what the idea looks like in current practice.

Disclaimer

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

Unfreeze, change, refreeze. It is on the second slide of every change management deck in circulation, it is attributed to Kurt Lewin, and he never proposed it. Stephen Cummings, Todd Bridgman and Kenneth Brown went back through the papers in 2016 and found the model assembled after his death in 1947, by other people, from fragments of an argument that was doing something else.

What Lewin actually described is more useful and considerably less comfortable. Any current behaviour, in a team or a company or a family, sits at a level held by opposing forces pushing against each other. He called it a quasi-stationary equilibrium: not stillness, but a draw. Nothing is resting. Everything you can see is the visible result of a fight that is already going on, and the level moves when the balance of that fight changes, not when somebody announces a new level.

1. The equilibrium is the finding

Picture a stream running at a constant depth. It looks static. It is not: water is arriving and leaving continuously, and the depth holds because the two rates match. Add water and the flow out rises to match. You are back where you started, with more water moving and nothing to show for it.

Organisational behaviour works the same way. The current level of anything you dislike, code review that takes four days, a team that will not challenge its lead, a policy everybody routes around, is held in place by forces pushing towards it and forces pushing away from it. Both are live. What you are looking at is their sum. So the useful question is never why people behave like this. It is what holds this level here. And what the forces would do if you took one of them away.

2. Take the brakes off, do not press harder

Two ways exist to raise the level. Increase the driving forces or reduce the restraining ones. They arrive at the same number and they do not produce the same organisation.

Push harder and you raise the tension in the system. The level moves while the pressure is on, everybody feels the strain, and the moment attention goes elsewhere the restraints, still fully loaded, take it straight back. Every mandated behaviour change you have watched fail did exactly this. It looked like insufficient commitment. It was ordinary physics. Take a restraint out and the level moves and stays moved, without the strain, because nothing is holding it back any more.

This is why the AI mandates are producing the results they are. The driving force is enormous: executive attention, budget, a stated expectation that everyone will use the tools. The restraints are untouched. Review capacity has not grown. Nobody has said who is answerable for generated code. The appraisal criteria still reward the outputs they always rewarded. Add force to that and you get compliance, tension, and a snap back the week the programme ends. Remove one restraint instead. Name the owner of a generated change and the level moves on its own.

3. Behaviour sits in a field, not in a person

Lewin’s formula is that behaviour is a function of the person and their field, and that both terms belong in the sentence. He meant it literally. You cannot read a person’s conduct off their character, because the same person in a different field behaves differently, which is a thing every leader has watched happen to a hire who was excellent somewhere else.

Bourdieu arrived at a version of the same thing from another direction, with dispositions laid down by a position and then carried into it. The practical consequence is identical and it is unwelcome: most attempts to change behaviour address the person, through training, through communication, through incentives aimed at individuals, when the thing holding the behaviour is the field. Lewin’s wartime food studies made the point plainly. Trying to change what families cooked, a lecture shifted a few per cent of them; a group that discussed the problem and made its own decision shifted around ten times as many. Same information, same room, different result. The second changed the field. The first only addressed the people standing in it.

4. Freezing was always provisional

The three-step caricature does its worst damage at this step. Lewin did write about making a new level permanent, and the phrase he used was permanent for a desired period. The whole point of settling was that the change should hold long enough to be worked in, not that it should hold forever. A quasi-stationary equilibrium is by definition temporary, and he knew that better than anyone, because he had spent his career describing why levels move.

The version the industry adopted made refreezing the end of the story, and organisations have been dutifully building end states ever since. A target operating model is an end state. So is a platform strategy with a completion date, and so is every AI usage policy being written this quarter, most of which will be wrong within two releases of the tools they govern. A structure that presents itself as final has made a promise about the future, and the promise is that it will resist the next event. It will keep that promise.

The corrected reading gives you the design rule. Settle deliberately, and settle in a form that can be unsettled: name the conditions under which this arrangement should be revisited, put a date on the review before anybody asks for one, and say out loud that this is the arrangement for now. An organisation that can say that has built something it can change. An organisation that cannot has built something it will have to break.

5. Study it by changing it

Lewin coined action research in 1946, and the underlying claim is that you understand a social system by trying to move it and watching what pushes back. The resistance is the data. You do not get to the restraining forces by analysis at a distance, because most of them are invisible until something disturbs them, and the people holding them usually cannot name them either.

Schön built a whole account of professional practice on the same insight, and Argyris, who worked in Lewin’s tradition, added the sharpest part: the forces people describe in the meeting are their espoused theory, and the ones actually holding the level are in what they do when the meeting ends. Run the small change first, then read the objections it produces. Those objections are your force field, drawn accurately and for free.

6. Provisional settling as a leader’s job

At this point in a change, the leader’s work is neither dramatic nor especially satisfying. The event has happened, the language has moved, and what is left is arranging the place so that the new behaviour is the easy one. That is structure, and it is mostly the removal of things.

Lewin’s contribution to that job is the discipline of asking what is holding the current level before proposing anything, and then arranging the new one so it can be dismantled by whoever comes next. It is a smaller claim about leadership than the field usually makes. It also survives contact with real organisations, which the three-step model has never managed.

Placement: Transforming, the fourth of the series’ four phases, after Learning, Deciding and Building. This one sits at the Structure position, where a change either gets an arrangement it can live in or waits to be undone. The capacity the phase produces is release, which means letting the settlement be revised by people who were not there when it was made.

(An Organisational Prompt is something you can do now....)

Organisational Prompt

Draw the brakes, not the reasons.

Take one change that keeps not sticking. On the left of a page, list what is pushing towards the new behaviour: the mandate, the tooling, the executive attention, the training. On the right, list what is holding the old behaviour in place, in specific terms: the appraisal criterion, the review queue, the person who has to approve, the report that still asks for the old number, the time nobody has.

Then leave the left column alone. It is already crowded and adding to it is what got you here.

Pick one item from the right and remove it this month. Not reduce, not review, remove. Watch whether the level moves without anybody being asked to try harder. That is the experiment, and the argument it produces afterwards will tell you which restraint to take out next.

Further Reading

Kurt Lewin, Frontiers in Group Dynamics (1947), the paper where quasi-stationary equilibria and force field analysis are set out, is available through many university libraries and is worth reading in the original for how little it resembles what is attributed to it.

Stephen Cummings, Todd Bridgman and Kenneth Brown, Unfreezing change as three steps: Rethinking Kurt Lewin’s legacy for change management, Human Relations (2016), free at https://journals.sagepub.com/doi/10.1177/0018726715577707.

Kurt Lewin, Action Research and Minority Problems (1946), which names the method and states the case for changing a system to understand it.

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

In 1939 Saul Alinsky went to work in the Chicago stockyards district, the neighbourhood Upton Sinclair had written about in The Jungle. He arrived with no programme. He spent months in kitchens and parish halls and union offices asking people what they wanted, and what he found was not what the reformers of the day assumed. People wanted wages, drains, rubbish collection and somewhere for the children to go. The Back of the Yards Neighborhood Council came out of those conversations, and it is still running.

Two things about that council are worth more than the whole shelf of change-management literature. Alinsky did not bring the grievance. The grievance was already there, decades old, and had no forum that could act on it. And he built the council out of 185 organisations that already existed, the parishes and unions and social clubs, rather than founding a new body to compete with them. Most corporate change does the exact opposite on both counts. It arrives with a grievance nobody feels, and a structure nobody asked for.

1. He never brought the problem

Alinsky’s starting position is that people act on their own interests, and that organising which appeals to duty or mission instead will not hold. Readers usually take this as cynicism. The opposite is true, because it takes seriously what people actually want. Self-interest is not selfishness. It runs across the material, the professional, the social and the moral. It includes wanting a say. It does not include your strategy. When a transformation programme asks people to act on the organisation’s interest, it is asking them to run on a fuel they do not carry. They will comply for a while because complying is easier than arguing, and the programme will read the compliance as agreement.

His other founding rule is to start where the world is rather than where you would like it to be. Accepting the world as it is does not weaken the desire to change it. Nowhere else is available.

2. Apathy is a conclusion

The quiet in your organisation is data, and it is probably not data about the people. Alinsky spent his career among people who had every reason to act and did not, and his reading of that was never that they did not care. People who have learned that action produces nothing behave exactly as though they do not care. The behaviour is identical. The cause is opposite. Martin Seligman later gave the mechanism a name and a laboratory, but Alinsky was working with it thirty years earlier as a practical fact about neighbourhoods.

Read your own organisation that way. When you announce something and nothing happens, the silence is a summary of the last three programmes, and I have never known that summary to be wrong. Somebody is doing arithmetic on your track record and arriving at the correct answer.

3. What can be manufactured

One distinction rescues this whole line of thinking from the manufactured-urgency trap. You cannot manufacture a grievance. Alinsky never tried. What the organiser manufactures is the occasion. The occasion is the confrontation on which a long-standing grievance becomes public, in front of witnesses who cannot afterwards claim not to have seen it. The grievance is found. Only the occasion is made. That distinction is the whole of it.

His ninth rule says the threat is usually more terrifying than the thing itself, and the story attached to it is instructive. Word was allowed to circulate that large numbers of people would occupy the washrooms at O’Hare, at length, in a coordinated way. The city imagined it. Passengers, newspapers, the lot. It moved on a commitment it had been sitting on for years. The tactic was never carried out. Nothing was manufactured except the credible prospect of the occasion.

His fourth rule is the one to take into any institution: make the opponent live up to their own book of rules. Organisations publish standards they do not meet. That gap costs nothing to find, and it is the cheapest leverage available to anyone inside, because you are not asking for anything new. You are asking for the thing already written down.

4. The rule that should worry you

Rule thirteen: pick the target, freeze it, personalise it, polarise it. It works, and it works for exactly the reason René Girard says it works, which is that a group under pressure will converge on a single figure with real relief and will experience the convergence as having finally got to the bottom of things. Alinsky is recommending, deliberately and in writing, the mechanism Girard spent a career describing as the thing communities do instead of learning.

He knew what he was doing. Rules for Radicals devotes a full chapter to the ethics of means and ends, and its argument is that judgements about means depend heavily on whether the person judging is a participant or a spectator. That is a serious position and it is not obviously wrong when the participants are people with no other lever at all.

It is wrong inside your own organisation. A leader who applies rule thirteen converts a systemic question into a personnel answer, and the relief in the room is real, which is precisely the problem. Relief was what the room wanted. The information was what it needed. Mary Parker Follett is the standing objection here, and she is probably right at the point where Alinsky is most dangerous: integration rather than domination, on the assumption that shared interests can be found rather than forced. Take the analysis. Leave rule thirteen.

5. The programme with no members

Almost every enterprise has run two AI efforts at once and only counted one. The official one has a charter, a budget, a steering group and a set of adoption targets. The other one happened in a channel nobody chartered, among engineers swapping what worked, and it was driven entirely by self-interest of the sort Alinsky would recognise: less tedium, faster feedback, fewer evenings lost to a migration script. Nobody organised it and it organised anyway. That is what people do when something serves them.

The official programme has all the position power and no members. The unofficial one has all the members and no authority to change anything that matters, which is why the governance forum still takes three weeks and the procurement process still takes four months.

Alinsky’s question for the official programme is short. What does the person you want to move actually want? If the answer is that they want to be part of the transformation, you have no answer, and the programme is running on your fuel rather than theirs. Fuel of that kind tends to run out around the second quarter.

6. The organiser leaves

Alinsky called it the iron rule rather than a preference: never do for people what they can do for themselves. The reasoning is not modesty. Doing the work for people leaves them without the capacity to do it next time, and without ownership of what was won. A victory the group did not win itself is hollow, and it can weaken the organisation rather than build it, because everyone now knows who to wait for. So the organiser finds the people others already listen to, works through them, and goes. Alinsky joined none of the organisations he founded.

The measure of an organiser is what still runs after they have gone. Apply that to your own arrangements and the reading is uncomfortable. The transformation office that has to keep existing is the evidence that the transformation did not take. A capability that needs its programme is not a capability yet; it is a dependency with a budget line. Back of the Yards has outlived its organiser by more than fifty years, which is the number worth being judged against.

Where this falls: Transforming, the last of four; Learning, Deciding, Building, Transforming, at Event. Alinsky's subject is the manufacture of an event: organising is what you do when the disturbance you need has not arrived by itself. The Event Is Not Yours to Make opened this stretch.

(An Organisational Prompt is something you can do now....)

Organisational Prompt

Ask what they want, not what they think

Pick three people at least two levels below you who do the work your programme is about. Ask each one what would make next Tuesday less annoying. Do not ask what they think of the strategy, the tooling or the transformation, because they will tell you what they believe you want to hear and that answer is worth nothing. Write down the three things. Fund one of them this month, visibly, and say whose idea it was.

Further Reading

Alinsky, Rules for Radicals: A Pragmatic Primer for Realistic Radicals (Random House, 1971). The thirteen rules get all the attention; the chapters on the organiser and on the ethics of means and ends are the ones that repay a slow read. A summary of the rules with the O’Hare episode is available.

Alinsky, Reveille for Radicals (University of Chicago Press, 1946). The earlier and warmer book, written out of Back of the Yards rather than about it.

The Industrial Areas Foundation, founded 1940 and still training organisers. Read it for what the method looks like after eighty-five years of practice rather than as Alinsky left it.

Bill Moyers, Who is Saul Alinsky?. A short corrective for anyone who has met the name only as a term of political abuse, with his own account of why he never joined anything.

Disclaimer

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

Chapter three of Leading Change explains how to raise urgency in an organisation that has none. Allow a financial loss to happen. Let errors escalate rather than be quietly corrected. Expose managers to the places where competitors are beating them. Get rid of the executive dining room. Set targets that cannot be hit by working the way people work now. Send managers to sit with unhappy customers and suppliers. Bring in consultants to force honest data into management meetings.

John Kotter published that list in 1996 and it has been the operating instruction for enterprise change ever since. Three of those tactics construct a crisis. The rest increase the flow of real information about a crisis that already exists. Nobody drew the distinction at the time, and most of what has gone wrong with transformation programmes since sits in the gap between the two halves of one list. Kotter noticed. He spent the following sixteen years correcting step one in public, across two more books, and I would take the correction over the original.

1. The step he found everybody skipping

The 1995 Harvard Business Review article came out of ten years watching more than a hundred companies try to remake themselves. He listed eight errors. He put not establishing enough urgency first, because that is where most of them died. More than half never got past it. The finding is empirical and it has held: an organisation with no felt need to move will produce plans, workshops, programme boards and no movement.

Complacency is the hard part, and Kotter is good on why. It is invisible to the complacent. It tends to grow out of past success rather than failure, so the organisations most exposed to it are usually the ones with a good decade behind them. He set a threshold as well. Roughly three quarters of management have to believe, genuinely, that staying as they are is more dangerous than changing. Three quarters is a high bar. Most programmes launch with the executive committee convinced and call that alignment.

2. The correction he published in 2008

Twelve years after Leading Change, Kotter published an entire book about its first step. A Sense of Urgency is a correction, though it never quite says so.

The correction is a third category. Everybody already knew about complacency and everybody knew about urgency. He put a third state between them and called it false urgency. From any distance it looks real. Meetings, task forces, long days, visible strain, people moving quickly between rooms. What drives it is anxiety and anger rather than a determination to win, and the activity it produces goes nowhere. False urgency, Kotter writes, is almost always produced by failure or by intense pressure applied to a group.

Intense pressure applied to a group is what step one instructs.

He left the reader to notice. An organisation in true urgency removes work. It stops doing several things so that it can do one. An organisation in false urgency adds a workstream and keeps everything else. Count what your transformation has stopped. If the answer is nothing, you have the counterfeit.

3. The second correction, in 2012

In 2012 he moved urgency off the crisis altogether. The Accelerate article, which won that year’s McKinsey Award, relocates the whole first step onto what he called a single Big Opportunity. He compared the two periods himself: the early work was about ridding an organisation of complacency, the later work treats sustained urgency as a competitive advantage in its own right.

A threat has to be believed. An opportunity can be checked. A threat is a claim about the leader. An opportunity is a claim about the world. When a leader announces that the company is in danger, the only responses available are belief and disbelief, and disbelief is the safer bet for anybody who has been through this before. When a leader points at something the organisation could now do and could not do last year, people can go and look. Some will look and disagree. That is fine, because the disagreement is then about the world rather than about the leader’s motives.

The rest of Accelerate follows from the relocation. The eight steps stop being a sequence and become accelerators that run at the same time and never stop. A volunteer army replaces the cascade. A network sits alongside the hierarchy, drawing people from every level and function, while the hierarchy carries on doing what hierarchies are good at. Holding the network together is a guiding coalition.

4. Why the manufactured version costs more than it earns

An instruction can be complied with. An event cannot.

Manufactured urgency therefore fails in a way that looks like success for about two quarters. Announce urgency and the organisation produces the observable signs of urgency, because producing observable signs is the thing organisations are best at. Attendance rises. A workstream appears. Somebody stands up at the town hall and says the right thing. None of this is cynical. People are responding to an instruction in the way instructions are meant to be responded to. Being able to respond at all is the proof that no event occurred. Nobody complies with an outage.

Reviewers of Leading Change reported managers telling them that routinely setting unreachable targets dampened urgency rather than raising it. Anybody who has worked in a large company for five years already knows this, and I include myself. The first announced emergency moves people. The second moves fewer. The third moves nobody. By then the organisation has learned that an announcement carries no information about the world, and that waiting is the correct response. Waiting is cheap. Waiting is usually right.

The bill arrives later and it is usually large. When something real finally happens, the organisation has no way of telling it apart from the last three things that did not. Every manufactured crisis spends the same credibility, and nothing replaces it. Ronald Heifetz gives the leader in the room the version to work with: adaptive work only happens inside a band of discomfort, too little and nothing changes, too much and people panic and look for somebody to blame. Kotter gave urgency a throttle. He never gave it a thermostat.

5. AI did not need a memo

Nobody needed a memo about AI.

The event happened somewhere around the point at which generating a working thing stopped costing anything, and engineers found out the way engineers find things out, in the evenings, on their own machines, before anybody had approved anything. Whatever else is true about the last three years, complacency was not the constraint. The felt sense that the status quo was fine did not survive first contact.

Organisations ran step one anyway. Adoption targets. Mandates. The AI-first memo. Licence utilisation on a dashboard beside the delivery metrics. An adoption target measures adoption. A mandate measures compliance with the mandate. Neither measures whether the work changed. This is manufactured urgency about a fire everybody was already standing in, and it produces what manufactured urgency usually produces, which is a compliance number. Seat utilisation measures whether people opened the tool. It has never measured whether anything changed about how the work gets done, and in most enterprises the two have come apart badly.

Run Kotter’s own 2008 test. Since AI arrived, what has the organisation stopped doing? Not what it has added: the platform team, the guild, the governance forum, the training programme, the centre of excellence. What has it stopped? Most AI programmes have added eleven things and removed none, which is Kotter’s own definition of the counterfeit, published eighteen years ago, applied to the thing his readers are doing now.

6. The part that survives

What survives the correction is step two.

The guiding coalition has four criteria and they have not moved since 1996. Position power: enough of the people who could block the work if they were left out of it. Expertise: the actual range of skills and perspectives the work requires. Credibility: reputations strong enough that when the group says something, people take it seriously. Leadership: proven leaders, who are found throughout an organisation and are not the same population as its managers.

He is emphatic that this is not a steering committee, and the distinction is sharper than most organisations allow. A committee has a mandate and a budget. A coalition has shared conviction. A committee is appointed. A coalition assembles. When the programme ends and the consultants leave, the committee dissolves on schedule and the coalition is still there, because it was never constituted in the first place.

Three of the four criteria can be arranged. Position power is a calendar invitation. Expertise can be hired by the day. Leadership can be developed, slowly, if somebody is paying attention. Credibility is the one that cannot be assigned, because it is held by whoever the people doing the work actually believe, and that person sits several levels below the steering group and has never been asked to join anything. I cannot remember a programme board that failed for want of position power. I have watched several fail for want of credibility. Look at who is in the room for your AI programme. If everybody is there because of their title, you have assembled the one kind of group Kotter spent a career warning about.

The coalition is not what causes the event. It is what carries one after it arrives.

Placement: Transforming, the fourth of the series' phases; Learning, Deciding, Building, Transforming, at the Event position. Kotter appeared first in Learning, where the eight steps were read as a theory of how change gets narrated. Here urgency is read as a constructed event rather than a discovered one, which is the harder claim and the more useful one. The Event Is Not Yours to Make opens this stretch.

(An Organisational Prompt is something you can do now....)

Organisational Prompt

Cross out the titles

Take the attendee list for your AI programme board and cross out everybody who is on it because of their job title. Count who is left. Those are the people whose names carry weight with the engineers, and if there are fewer than three of them you have a reporting line rather than a coalition. Find the two people your teams actually believe, add them, and give them the same decision rights as everybody else in the room.

Further Reading

Kotter, Leading Change: Why Transformation Efforts Fail (Harvard Business Review, 1995). Ten years of watching more than a hundred companies, reduced to eight errors. The full text is hosted by UCSF. Read it for the finding rather than the model: most transformations died at the first step.

Kotter, Leading Change (Harvard Business School Press, 1996). The book the eight steps come from. Chapter three is the one to read, because it sets out the tactics for raising urgency plainly enough to argue with.

Kotter, A Sense of Urgency (Harvard Business Press, 2008). A whole book on step one, twelve years later. Read it for the three-state taxonomy and for the diagnostic about what an organisation stops doing.

Kotter, Accelerate! (Harvard Business Review, November 2012), later expanded into a book. Kotter’s own firm hosts the article. The dual operating system, the volunteer army, and urgency relocated from a crisis onto an opportunity.

Disclaimer

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

Four companies in the same sector announced platform strategies within eleven weeks of each other last year. The architectures were not identical, but they ‘rhymed’. Each deck contained a slide about becoming the intelligence layer for its industry. None of the four had been asked for this by a customer.

Ask any of the executives involved and you will get a coherent account of market forces, and the account will be sincere. What nobody will say is that they are copying each other, because copying is not how it feels from inside. It feels like independent analysis arriving at the same conclusion. Independent analysis does sometimes arrive at the same conclusion. It also looks exactly like copying, from outside, to everyone except the people doing it.

Girard was born in Avignon in 1923, took a doctorate in history at Indiana in 1950, and spent the rest of his life in American universities: Johns Hopkins, Buffalo, then Stanford from 1981. He was elected to the Académie française in 2005 and died in 2015. Across all of it he pursued one idea, from literature into anthropology and then into theology, and the idea is this: we do not know what we want. We take our wants from other people. Everything else in his work follows from that, up to and beyond the part organisations most need to understand, which is what a group does when the resulting rivalry becomes unbearable.

These articles hold that the event which changes an organisation is not the leader’s to make. Girard holds the Event position, and he is there because he gives the best available account of where the crisis actually comes from. Not a competitor, a regulator or a technology. The crisis is generated inside the group, by the group, out of the ordinary mechanics of wanting things, and it has a default resolution that works, costs one person, and teaches the organisation nothing.

1. Desire is borrowed

Girard’s first book, Deceit, Desire and the Novel, came out in 1961 and looked like literary criticism. He had been teaching French literature in America and noticed something the novelists he most admired all seemed to know.

We tell ourselves that desire runs in a straight line. A subject wants an object because of what the object is. Girard called this the romantic lie. What the great novelists show instead is a triangle. Between the subject and the object stands a model, and the subject wants the object because the model wants it, or has it, or seems to.

Don Quixote does not spontaneously want chivalry. He wants what Amadis of Gaul wanted. The wanting is borrowed, all of it, and Cervantes says so on the first page. Emma Bovary wants the life she read about in books. Proust’s social climbers want the drawing room they cannot enter, and lose interest the moment they are inside. Dostoevsky’s characters torment each other over objects that neither party cared about before the other one did.

Girard’s move was to say that this is not a literary device. He read it as a description of how human wanting works, found by novelists because sustained attention to actual behaviour turns it up.

Sit with that for a moment, because it contradicts something most of us believe about ourselves. You did not independently arrive at your preferences in cars, schools, holidays or technology strategies. You acquired them from people whose position you were paying attention to. I find this uncomfortable, and after some years of looking I have not found a way around it.

2. The nearer the model, the worse it gets

The distinction that does the practical work is between two kinds of model, and Girard is unusually clear about it.

External mediation is when the model is remote. Another world, another era, another league. Don Quixote can imitate Amadis all he likes without ever meeting him, and the imitation stays admiring. There is no possibility of collision because the two do not occupy the same space, so the model remains a model.

Internal mediation is when the model is right there. Same market, same floor, same promotion round. Now the model wants the same object you do, and the model is in your way. The imitation continues, but its emotional register flips. The person you are copying becomes the person you resent, and the resentment is proportional to the closeness rather than to any wrong they have done you.

Notice what this does to the usual story about competition. We assume rivalry comes from difference: different interests, different incentives, different worldviews. Girard says the opposite. Rivalry comes from similarity, and it intensifies as the parties become more alike. The teams that fight hardest in your organisation are not the teams with nothing in common. They are the two platform teams with overlapping remits, or the two directors at the same grade with adjacent portfolios, who agree with each other about almost everything.

An uncomfortable corollary follows for anyone who has flattened a hierarchy. Removing distinctions removes distance, and distance is what kept the imitation admiring rather than rivalrous. Some of the most poisonous organisations I have been near were also the flattest, and everyone in them could tell you sincerely that they had removed hierarchy to reduce politics.

3. Everyone is building the same thing

Now take the last three years of technology strategy and read them through this.

Boards want AI capability. Ask a board why, and you get an answer about competitive necessity, and that answer resolves on inspection into a statement about what other boards want. The desire is mediated, and the model is the peer group, and nobody in the chain is lying. Each firm can point to real evidence that the other firms are moving. That is the mechanism exactly: the model’s wanting is the evidence, and the evidence is the model’s wanting.

This would matter less if the mediation were external. Nothing about it is. These are firms of similar size, in similar markets, hiring from the same pool, advised by the same three consultancies, reading the same analysts. Similar firms, similar models, similar desires. Internal mediation at industry scale, and internal mediation is the kind that turns into rivalry.

And something new sits on top of it, which Girard did not live to see. Everyone now consults the same models, trained on the same corpus, prompted in roughly the same way, and the strategic advice converges. Ask a frontier model to design your AI transformation and it will produce something coherent and defensible and very close to what your competitor received when they asked the same question the same week. The tool that was supposed to confer advantage is a powerful engine of undifferentiation.

I do not think this is a reason to avoid the tools. But the scarce thing has moved. Once the analysis is commodity, differentiation has to come from somewhere the model cannot reach: your actual constraints, your actual customers, the thing your organisation is oddly good at that nobody else can copy because it took eleven years to build.

The four platform strategies I opened with are what happens when nobody checks. They spend heavily. They arrive at the same place. They then discover that a thing everyone has is not an advantage, and that the money went on becoming less different from each other.

4. When the differences dissolve

Girard’s second book, Violence and the Sacred, published in 1972, follows the logic into what happens when mimetic rivalry runs unchecked in a group.

Rivalry spreads by the same imitation that started it. A conflict between two parties recruits others, who take sides by copying, and the copying accelerates. The original object of contention becomes irrelevant. Only the rivalry is left, and the rivals become more and more alike as it goes on, until they are behaving identically while insisting on their differences.

He calls the resulting state undifferentiation, and it is the centre of his account of crisis. The distinctions that ordered the group have failed. Nobody can say who is who. Myth represents this as plague, flood, monstrous twins, all of them images of collapsed distinction.

The organisational version needs no translation. Two teams with overlapping charters, both briefing upwards, both producing roadmaps that cover the same ground. A matrix in which nobody can say who decides. A programme where platform, enablement and the centre of excellence are three names for teams doing substantially the same work while each insists its remit is distinct.

Look at what the words are doing there. Two teams, one remit. Two remits, no difference. When a team cannot say what it decides that the other team cannot overturn, the team has no boundary, and a team without a boundary is not really a team. It is a claim on the same work. The more the two teams talk, the more alike the two teams become, and the more alike they become the harder they fight.

The danger is that it feels like a coordination problem, and coordination problems have familiar remedies. More alignment meetings. A shared roadmap. An offsite. Every one of those brings the parties closer together, which is to say it increases internal mediation, which is to say it feeds the thing it was meant to fix. The temperature goes up after the alignment session and everyone finds this mysterious.

5. The single victim

Here is where Girard’s argument becomes genuinely disturbing, and where he is most often accused of going too far.

His claim is that groups in this state do not resolve their crisis by negotiation or insight. They resolve it by converging, spontaneously and with near unanimity, on one member. The choice looks motivated to the participants and is substantially arbitrary. The requirement is that the victim be inside the group and marginal to it, so that expelling them costs nothing and provokes no retaliation.

The expulsion works. People resist that, and it is the part that explains why the mechanism lasts. The rivalry that was running everyone against everyone gets discharged in a single direction, and afterwards the group experiences real peace. Girard’s account of the sacred begins here: a community that has just discharged its violence onto one person and found calm reads the calm backwards, first as proof that the victim was guilty, then as evidence that the victim had some power over the group’s fortunes.

And the whole thing requires misrecognition. The mechanism only works while the participants believe they have correctly identified who was responsible. A group that knows it is scapegoating cannot get the benefit. That is why the phenomenon is so hard to see from inside and so obvious from outside, and why the people who did it will describe the meeting afterwards as the one where things finally became clear.

Modern organisations do not sacrifice anyone. They do something structurally identical and much politer. The executive whose departure explains two years of underperformance. The supplier who becomes the reason. The team that gets restructured out, whose removal is described as simplification. The team lead named in the incident review as the one who missed the alert.

Each converts a systemic question into a personnel answer. Each is followed by real relief. And each leaves the conditions that produced the crisis entirely intact, which is why the same organisation runs the same play against a different person about two years later. Same conditions, same crisis, different person.

6. Four marks on the page

In The Scapegoat, published in 1982, Girard does something more useful for a working leader than any of the theory. He gives you a way to read a document.

His argument is that texts written by persecutors, who do not know that they are persecutors, carry four recurring features. First, a description of general crisis, of a situation in which the normal order has broken down. Second, accusations against the victim of crimes that themselves destroy distinctions, so that the alleged offence conveniently matches the crisis it is meant to explain. Third, marks of selection on the victim, signs of being both inside the group and set apart from it. Fourth, the violence itself, narrated as necessary and as restoring order.

Where all four appear together, Girard says, you are reading an account of a real persecution written by people who believed they were describing justice.

Now go and read your last serious post-incident review with those four in hand. I have done this and it is not a comfortable exercise.

A crisis will be described in the opening paragraphs, usually with the word “systemic” somewhere. You will find the failure attributed to a specific act that, conveniently, explains the whole thing. The person will be identified in terms that mark them: relatively new, or brought in from outside, or holding an unusual role that sits between two teams. And you will find the remedy, framed as strengthening controls, which in operational terms means removing that person’s discretion or that person.

The same reading works on a reorganisation memo. A crisis, an account of what went wrong that centres on one team, marks of selection on that team, and the restructure of that team presented as clarification.

I want to be careful here. Sometimes a person did the thing. Sometimes a team really should not exist. Girard’s four marks are not proof of injustice and using them as a trump card would be its own kind of foolishness. What they give you is a prompt to check, at exactly the moment when checking is least welcome, whether the account being agreed explains the crisis or merely ends it.

7. Why it works less well than it used to

Girard’s later work is theological, and readers split sharply over it. Some treat it as the point of the whole enterprise and others as a departure from the anthropology. The two do not separate cleanly, so the theology is better stated plainly than skipped.

His claim: the Hebrew and Christian scriptures differ from myth in one decisive respect. They tell the story from the victim’s side. Where myth reports the crisis and the expulsion as the persecutors understood them, and so hides the mechanism inside its own account, these texts insist repeatedly that the expelled one was innocent. Once that perspective is available, the mechanism is exposed, and an exposed mechanism stops working.

You can accept or reject the theology and the sociological observation still stands. Modern societies are unusually preoccupied with victims. Historically that is strange, and Girard’s explanation for it is at least interesting: we inherited a text that took the victim’s side, and the concern outlasted the belief.

The consequence is not comfortable, though. A society that cannot scapegoat innocently is a society whose crises no longer discharge. We are all now sophisticated enough to know that blaming one person is unfair, and this has not stopped anyone blaming one person; it has made the blaming self-conscious, defensive and prone to escalation. Girard also noted the newest move, which is for rival parties to compete over who is the real victim, so that the language developed to protect victims becomes the weapon of the persecution.

Read the last eighteen months of public argument about AI and labour with that in mind. Both sides are speaking the language of the wronged.

8. What a leader actually does with this

Two moves follow, and the second is the one nobody teaches.

The first is refusal, which the opening article set out. As the room converges on a name, the person with standing declines the resolution, and does it before the relief arrives rather than after. That costs something and it is a single sentence. MacIntyre’s test is the useful check: does the proposed account explain why what we believed turned out to be inadequate? A name explains nothing. Heifetz then supplies what to do with the undischarged energy, which is to hold it at a temperature people can survive.

The second move is structural and it is where Girard becomes practically useful rather than merely illuminating. If rivalry is generated by undifferentiation, then the treatment is to restore real differences. Not cultural differences, and not personality types. Real differences: distinct remits, with boundaries somebody can point at.

Building spent months on the same instrument, arriving from an entirely different direction. Parnas said modules should hide their decisions from each other. Conway said the interfaces will mirror the organisation whether you like it or not. Beer said each viable unit needs genuine autonomy inside a defined scope. All three are describing differentiation. Girard tells you what differentiation prevents.

So when two teams are at each other’s throats, an offsite is not the intervention. A boundary is the intervention. Decide what each team owns. Write it down. Make the ownership real by giving each team control of something the other team cannot override, and accept that the result will look less integrated than the diagram you had before.

The criticisms of Girard should be stated, because he attracts uncritical followers and that does his argument no favours. He explains one thing and tends to explain everything with it. He is better at accounting for how crises resolve than at predicting who gets chosen. His use of anthropological evidence has been called selective by people qualified to judge, and I am not qualified to adjudicate that. Take him as a lens rather than a law, and he earns his place.

But hold on to the central claim, because I think it is right and I think it is nearly always missed. The crisis in your organisation is not arriving from outside. It is being generated inside, continuously, by ordinary people wanting what the people next to them want. You cannot stop it, because wanting works that way in everyone, including you. What you can decide is whether it ends with a name.

Where this sits: Transforming, the last of the series' four phases; Learning, Deciding, Building, Transforming. These articles start at Agency and move here to Event, and Girard supplies the sharpest move available: a crisis resolved by finding somebody to blame leaves an organisation feeling restored and entirely unchanged. The Event Is Not Yours to Make opened this stretch.

(An Organisational Prompt is something you can do now....)

Organisational Prompt

Read your last incident review as a persecution text

Find the most serious post-incident review or reorganisation memo your organisation produced in the last year. Print it. Set aside forty minutes and find a pen.

Mark it against Girard’s four. One: where is the crisis described, and does the description say that normal order had broken down? Two: what is the victim accused of, and does the accusation happen to explain the whole crisis rather than part of it? Three: what marks the person or team out as both inside and marginal? New, external, interstitial, unpopular, unusual role. Four: how is the remedy described, and does it restore order by removing someone’s discretion?

If all four are present, you are not looking at proof of injustice. You are looking at a document that resolved a crisis rather than explaining it, and the conditions are still in place.

Then do the structural half. Name the two teams in your organisation with the most overlapping remits, the pair that generate the most heat per unit of actual conflict. Write one sentence for each, saying what that team decides which the other cannot overturn. If you cannot write those two sentences, you have found what your next reorganisation is actually for.

Further Reading

The Scapegoat (Johns Hopkins University Press, 1986). The most direct statement of the mechanism and the four stereotypes, and the one to read first.

Deceit, Desire and the Novel (Johns Hopkins University Press, 1965) is where it all starts, and it works as literary criticism even if you take nothing else from it. The chapter on external and internal mediation is the one that pays for the book.

Violence and the Sacred (Johns Hopkins University Press, 1977) is the hardest of the three and the most ambitious, taking the argument into ritual, prohibition and the origins of the sacred.

The Internet Encyclopedia of Philosophy entry is free, careful, and fair about the objections. The St Andrews Encyclopaedia of Theology has a longer treatment as a full PDF if you want the theological argument set out properly.

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

The offsite is in its second hour and the pyramid is on the flipchart. Absence of trust at the base, then fear of conflict, then lack of commitment, then avoidance of accountability, and results at the top. Everyone in the room has seen it before. Everyone agrees the problem is the base layer, which is a comfortable thing to agree because it makes the problem nobody’s in particular. Six weeks later the leadership meeting behaves in exactly the way it did in June.

Something similar happens with the other one. Four boxes, four styles, directing through to delegating, and a diagnosis of where each of your people sits. Between them, Patrick Lencioni’s pyramid and Ken Blanchard’s quadrant have been taught to more managers than any other pair of ideas in the field. They are also, by a distance, the two with the thinnest evidence behind them. That combination is worth taking seriously rather than sneering at, because both contain something real, and what gets taught is usually not it.

1. The pyramid and the quadrant

Lencioni published The Five Dysfunctions of a Team in 2002 as a business fable, which is to say a short novel with a model at the back. The five stack: without trust people will not risk conflict; without conflict they will not genuinely commit; without commitment there is nothing to hold each other to; and without accountability the team drifts from results to individual status. Trust in his sense is specific. It means being willing to be seen not knowing something, which he calls vulnerability-based trust to distinguish it from predictability.

Blanchard’s model came out of work with Paul Hersey in the late 1960s and was reworked as Situational Leadership II with Drea Zigarmi and Patricia Zigarmi. The claim is that there is no best leadership style, only a style appropriate to the person’s development level on a particular task. Four levels, four matching styles: directing, coaching, supporting, delegating. Give a beginner delegation and they flounder; give an expert direction and they resent it.

Both are elegant, both fit on a slide, and both have sold in the millions. Blanchard has written more than seventy books. The Table Group has built a substantial practice on the pyramid. Neither man is a fraud and both are describing something they have watched happen many times.

2. What the evidence actually says

Situational Leadership is among the best-known and least well-substantiated models in the whole field, and the gap is not a secret. The serious leadership textbooks note it. General management textbooks reproduce the quadrant without mentioning that it has been tested and largely failed to confirm.

Claude Graeff set out the theoretical objections in 1983 and again in 1997: weak or absent argument for critical parts of the model, internal inconsistencies, and conceptual ambiguity that got worse rather than better in the second version. On the empirical side, only four studies have ever fully tested the three-way interaction the theory predicts, between how directive the leader is, how supportive, and how ready the follower. Taken together they offer minimal support, often only directional, and mostly at the low-readiness end where the advice is least surprising. Thompson and Vecchio tested three versions across ten Norwegian financial institutions and found the 2007 revision a worse predictor of performance and attitudes than the original.

There is a deeper problem, and in practice it does more damage than the statistics do. The leader is the sole judge of the follower’s readiness. That exposes the whole diagnosis to attribution error and halo effect, and the predictions hold up only where the leader’s rating and the person’s own rating agree. When they disagree, which is often, the model tells you to treat someone according to a picture of them they do not share.

Lencioni is a different case and a fairer one, because he never claimed to be doing research. The pyramid comes from consulting practice and is offered as such. The causal ordering has not been tested. If you want the same territory with evidence underneath it, Amy Edmondson’s work on psychological safety is where to go, and it is a group property you can measure rather than a virtue the leader is supposed to embody.

3. What survives in Blanchard

Strip away the quadrant and one instruction remains, and it is the one almost nobody follows: the diagnosis is per task, not per person.

The model says a person can be a beginner on one thing and an expert on another, this week, in the same job. What organisations do instead is type people. She is a D4. He needs a lot of direction. The label attaches to the human being, goes into the talent review, and survives contact with everything.

Now look at what has happened to your engineering organisation in the last two years. Your most senior developer, twenty years of judgement, is a rank beginner at orchestrating agents and evaluating generated code at volume. On that task she is D1: keen, and without the pattern library that makes her formidable elsewhere. Treating her as the expert she is on everything else means she gets no direction on the one thing that is new, and is too senior to ask for it. Meanwhile a two-year engineer with no habits to unlearn may be D3 on exactly that task and is being managed as a novice.

I have watched this exact mismatch produce a rollout that went badly in both directions at once. The per-task instruction is nearly forty years old and it has never been more useful than it is right now.

4. What survives in Lencioni

The pyramid’s value is not the diagnosis. It is the ordering, and the ordering is a claim about sequence that most leaders violate every time they try to fix a team.

You cannot get commitment from people who have not argued. What you get instead is the thing Lencioni names precisely, artificial harmony, in which the meeting ends with agreement and the actual disagreement moves to the corridor. And you will not get argument from people who cannot afford to be seen not knowing. So the sequence runs backwards from what most interventions attempt: teams get told to commit, when the missing thing is two levels down.

The other durable piece is that somebody has to go first, and it has to be the most senior person in the room. Not a facilitated exercise. An actual admission, about something that costs, in front of people who report to you. Adkins would recognise this as the same move at a different scale: the instrument is your own tolerance for exposure.

The AI version of artificial harmony is now easy to spot. Somebody demonstrates a workflow, the room nods, and three people privately think the output is worse than what they had. Nobody says so, because saying so reads as being against the direction of travel, and the direction of travel has executive sponsorship. That is a team that will commit to nothing and deliver a pilot.

5. Why the training does not stick

Both models are taught in two-day workshops, and the workshop is where the failure lives rather than in the content.

A workshop is a language event. People leave with vocabulary, which is real, and Lakoff would say we underrate what vocabulary does. But the field the vocabulary lands in has not changed. Lewin’s point holds: any current behaviour is held in place by restraining forces, and adding a new model on top of untouched restraints produces a snap back to the original state, usually within a quarter.

The restraints are not mysterious. Promotion still rewards the person who had the answer. The review meeting still runs to a slide deck that leaves no space for someone to say they were wrong. The senior person still speaks first. None of that is addressed by a workshop, and all of it teaches the room what is actually safe, more convincingly than a flipchart ever could.

So if you want either model to hold, change one restraint rather than running the training again. Remove the thing that punishes the behaviour you asked for. That is a smaller and more awkward job than booking a facilitator, which is why it is rarely the one that gets done.

Placement: Transforming, the fourth of the series' phases; Learning, Deciding, Building, Transforming. Both belong at Agency, where the question is how much support a person needs before they can act, and the honest answer changes with their readiness rather than with your preference. The Event Is Not Yours to Make opens this stretch.

(An Organisational Prompt is something you can do now....)

Organisational Prompt

Rate the task, not the person

Take the six people who report to you. For each one, write down two things they are currently doing: something they have done for years, and something that arrived in the last eighteen months. Rate their development level separately for each, from beginner to expert.

Almost everyone will land in two different places. Now check your last three interactions with each person against those two ratings. Were you directive on the old thing and hands-off on the new one? Most of us are, because we manage the person we know rather than the task in front of them.

Then one thing for the team. Name the disagreement your team is currently not having, the one that lives in the corridor. Take it into the next meeting yourself, and open by saying what you have got wrong about it. If you cannot think of anything you have got wrong, you have found the reason nobody else is arguing either.

Further Reading

Patrick Lencioni, The Five Dysfunctions of a Team (Jossey-Bass, 2002). Two hours to read, a fable, and honest about being one. The model is in the last thirty pages if you want to skip the story.

Claude Graeff’s Evolution of Situational Leadership Theory: A Critical Review (Leadership Quarterly, 1997) is free in full and is the careful version of everything above. Read it before you commission the training.

Geir Thompson and Robert Vecchio, Situational leadership theory: A test of three versions (Leadership Quarterly, 2009). The empirical assessment. It includes the awkward finding that the 2007 revision predicted performance and attitudes less well than the version it replaced.

Amy Edmondson, The Fearless Organization (Wiley, 2018), for the same ground with two decades of field research behind it.

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

She starts the sentence and you finish it. “I think the reason the pipeline keeps stalling is that we’re not really...” and there is a pause of about a second, and you say “yes, the review queue, we should probably batch them.” She agrees. You both move on. It felt collaborative and it took eleven seconds.

What just happened is that you took the thought off her before she had finished having it. She now knows what you think, which she can work with, and you will never know what she was about to say, which you cannot. Adkins built a career on that particular loss. Her argument is that the most valuable thing a leader can do in a conversation is very often nothing at all. That is a trained skill rather than a personality trait, and almost nobody senior has trained it.

1. Four stances, and the one everyone defaults to

Coaching Agile Teams came out in 2010 and became the standard text for a role that barely existed. Adkins had come from fifteen years of running project teams, and the book is unusually candid about what that did to her. The skills that made her good at the old job, holding the whole picture in her head and orchestrating who spoke to whom, turned out to be liabilities in the new one.

Her framework separates four things that usually get lumped together as “helping”. Teaching transmits information the person does not have. Mentoring offers your own experience as a model, which is a different act and lands differently. Facilitating structures a conversation so that a group can get somewhere it could not have got to alone. Coaching asks a question and then leaves the person to find their own answer, which is the hardest of the four by some distance. All four are legitimate and each suits a different moment.

The discipline is to use the lightest one that will work: coach before you mentor, mentor before you teach. What actually happens in most organisations is that everyone defaults to teaching, because teaching feels useful to the person doing it and comfortable for the person receiving it. It also builds a dependency, quietly, over about eighteen months, until the team cannot make a decision without you in the room and everybody agrees you are indispensable.

2. What one second costs

There is a piece of research from outside this field that gives the whole thing a floor to stand on, and I think it deserves to be better known in business than it is. In the late 1960s Mary Budd Rowe recorded hundreds of classroom lessons and measured the pauses. She found that teachers waited less than a second after asking a question before speaking again, and less than a second after a pupil stopped talking. She named the two gaps wait time one and wait time two, and found the second one mattered more.

When teachers were trained to extend both to three seconds or more, the changes were not marginal. Pupils’ answers got several times longer. They speculated more, offered evidence unprompted, asked more of their own questions, and said “I don’t know” far less often. There is a threshold at roughly 2.7 seconds, below which the benefit largely disappears. The finding has since been replicated from primary schools to universities and into business settings.

Read that back as a description of your last leadership meeting. Roughly a second of silence is what most rooms can bear, and the person with the most authority is usually the one who breaks it. Every time you break it you have taken the answer.

3. Sixty per cent who you are

Adkins’ most demanding claim is that the coaching stance is about sixty per cent who the coach is and forty per cent what the coach does. That sounds like the sort of thing said to fill a chapter. In practice it names something specific, which is that the four seconds are unbearable for reasons that have nothing to do with technique.

Silence in a meeting you are running feels like failure. It reads as your failure to have prepared, or the team’s failure to engage, and the discomfort is yours to hold. So you fill it, and the filling is experienced by everyone as helpfulness. Adkins’ instruction is that you coach yourself first. The instrument is your own tolerance for other people’s struggle, it can be trained, and until it is trained you will keep rescuing teams that were about to work something out.

Heifetz said much the same about a leader’s job being to hold the disequilibrium rather than relieve it. Adkins scales it down to the meeting. Same discipline, three seconds instead of three months, and considerably more chances to practise.

4. Answers are now free, which is the problem

Something has changed since 2010, and it has changed more for this stance than for any other idea here. An engineer with a question no longer has to ask a colleague. The model answers at once, fluently, in whatever format was asked for, without judgement, and often correctly. That is useful. I would not want to take it away from anyone, and I use it myself all day. It also means the market value of being the person with the answer has collapsed, and a leader whose contribution is answers is now competing with something faster, cheaper and more patient.

Two consequences follow, and the second is the more interesting. The first is about you: whatever you were adding by knowing things, add something else. The second is about the room. When answers are free, the expensive thing becomes the collision, the moment two people who see a problem differently have to work it out between them. Teams that route every question to a model stop having those. Nobody decides to stop, and there is simply no longer a reason to turn to the person at the next desk, so the disagreements that used to surface in that moment go unheld instead. Somebody has to protect the collision, and that is a coaching job rather than a technical one. It means noticing when a decision was reached by consensus with a chatbot, and asking the two people who quietly disagree to say so out loud.

5. Coaching yourself out of a job

The agile coaching profession has a failure mode it is honest about, at least in its better literature. The coach becomes permanent. The team gets good, and the coach stays, and stays, and eventually there is a coaching function, and then a centre of excellence, and then a transformation office with a budget line.

Adkins’ test is blunt and I have never seen an organisation apply it. Does the team perform as well without the coach as with? If the answer is no after two years, the coaching failed, however good the retrospectives were.

That test belongs here rather than to agile. Release is what this part of the argument is for: the change no longer needing the person who started it. Applied to the coaching stance it is quite exact. Every time you supply an answer you make yourself slightly more necessary, and necessary is the opposite of the thing you were hired to produce.

You cannot make the event that changes your organisation. What you can do is spend the years beforehand building a group of people who can think without you when it arrives. That work is done four seconds at a time, in ordinary meetings, by not saying the thing you were about to say.

Where this falls: Transforming, the last of four; Learning, Deciding, Building, Transforming. Coaching sits at Agency, the position these articles start from, because agency is grown rather than granted and the growing is mostly a matter of not filling the silence. The Event Is Not Yours to Make sets out the argument.

(An Organisational Prompt is something you can do now....)

Organisational Prompt

Measure your own wait time, then buy four seconds

Pick a meeting this week that you run. Ask one person you trust to sit with a stopwatch and time two things. How long do you wait after asking a question before anyone speaks, and how long after someone finishes before you speak? They should write down the two numbers and nothing else.

Most people are somewhere near one second and are surprised by it. Do not defend the number. It is the same number Rowe found in every classroom she recorded.

Then run one meeting the following week under a single rule, which is that you count. After you ask a question, count silently to four before anything else happens. After someone stops talking, count to four again. Four seconds is much longer than it sounds and the first two attempts will feel like a mistake in front of witnesses.

Write down what came out of the room in the second half of those pauses, because that is the material you have been buying yourself out of for years. If nothing came, you have learned something worth knowing about the meeting.

Further Reading

Coaching Agile Teams (Addison-Wesley, 2010) is the source for all of this, and still the best treatment of the four stances anywhere. Pearson has the front matter and a sample chapter free as a PDF if you want to see the shape of it first.

Mary Budd Rowe, Wait Time: Slowing Down May Be A Way of Speeding Up (Journal of Teacher Education, 1986). Free in full, six pages, and it will change how you run a meeting more reliably than most management books.

Her current work has moved from team coaching to leadership teams and systems coaching, which is a useful corrective if you think of any of this as a delivery-team concern.

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

The workshop was booked six weeks ago and the template arrived before the facilitator did. By eleven the wall is covered in coloured squares. The squares have been grouped, the groups have owners, the owners have dates. Everyone leaves lighter than they came in. Nothing in those four hours asked whether the people in the room actually want the same thing.

That is the commonest failure in transformation work and it is almost never noticed, because the method did its job. It produced squares, groups, owners and dates, which is what it produces. The question it could not ask was whether this situation was the kind of situation it suits. Michael Jackson spent forty years on that question. His answer is that every way of intervening in an organisation carries assumptions about the world, about knowledge and about power. No method is neutral. Choosing between them, in the open and with reasons, is the first serious act a leader performs.

This part of the argument began with a hard claim: the event that changes an organisation is not yours to make. You do not author the crisis. What you do is take a position on how it resolves. Jackson supplies the content of that position. If the leader cannot generate the event, the leadership sits entirely in what happens next. How is the situation looked at, which approach is brought to it, whose interests does that approach serve, and can anyone say why this approach and not another. Agency here is custodial rather than heroic. I have watched that choice get made a hundred times by people who did not know they were making it.

1. Every method is partial, and partial methods travel

Jackson trained as an operational researcher and spent most of his career at Hull, where he founded the Centre for Systems Studies and edited Systems Research and Behavioral Science for a quarter of a century. The field he inherited had a problem he could not stop seeing. It had produced a shelf of powerful approaches, each built by someone serious, each demonstrably successful somewhere, and each sold as though it worked everywhere.

Systems dynamics modelled stocks and flows. Soft systems methodology surfaced competing worldviews. The viable system model diagnosed whether an organisation could survive in its environment. Interactive planning had the organisation design its own idealised future. Every one of them worked. Each was also built on assumptions about what kind of mess it was facing, and those assumptions were rarely stated, because the person applying the method had usually stopped noticing them.

A method applied outside the conditions it suits does not announce the mismatch. It usually produces confident, well-structured, entirely wrong work. The wall fills with squares. The model runs. The report is delivered on time and the situation is unchanged, or worse, because the intervention has now consumed the credibility that a better one would have needed.

So the skill is not solving. The skill is choosing.

2. The grid that classifies the situation, not the tool

In 1984, with Paul Keys, Jackson published “Towards a System of Systems Methodologies” in the Journal of the Operational Research Society. It proposed a grid with two axes. The first asks about the system: is it simple or complex? Few elements and predictable behaviour, or many elements, many interactions, and behaviour nobody can fully model. The second asks about the people: are their relationships unitary, pluralist or coercive?

Unitary means shared objectives. Everyone genuinely wants the same outcome and disagreement is about how to get there. Pluralist means different interests and compatible ones. People want different things, they know it, and a negotiated accommodation is available if anyone convenes it. Coercive means the interests cannot be reconciled and the matter will be settled by power, whether or not anyone says so out loud.

Cross the axes and you get six kinds of situation, each with its own family of approaches. Simple and unitary suits hard systems engineering and classical operational research. Complex and unitary suits system dynamics, socio-technical design and Beer’s viable system model. Simple and pluralist suits assumption-surfacing methods. Complex and pluralist suits soft systems methodology and Ackoff’s interactive planning. Coercive contexts, simple or complex, want critical systems heuristics and the emancipatory approaches, and Jackson was honest that the complex-coercive cell was the least developed of the six.

Read the grid the wrong way, as I did for years, and it becomes a lookup table for consultants. Read it the right way and it is a diagnostic on the situation. Beer supplies a superb method for a complex-unitary context. Jackson’s question comes first and it is more awkward: is this context unitary at all?

Most enterprise transformation is run as though the answer is yes. The strategy deck says one company, one direction. Ask the middle of that organisation what they want and you will hear at least four answers, several of which are incompatible with the deck and all of which are rational from where they sit. Run a unitary method over a pluralist situation and you get compliance without commitment, which is the state most large programmes live in and cannot explain.

3. Where AI adoption actually sits on the grid

Take an AI adoption programme, the kind now running in every large enterprise. Complex, plainly: the technology changes underneath the plan, the effects are second and third order, and nobody can model what six thousand people will do with a tool that answers back.

The people axis is the one that gets skipped. The board wants margin. Engineering leadership wants throughput and, quietly, wants to keep its best people. The engineers want to do interesting work and would rather not train their replacement. Legal wants defensibility. The operating units want their own thing shipped, not the platform. No communications plan I have ever seen has fixed any of it. They are different positions held by people with different stakes, which is the definition of a pluralist situation.

And in places it is worse than pluralist. When the same programme is presented as a productivity initiative to the board and as a capability uplift to the workforce, and only one of those framings appears in the business case, the situation has a coercive element. Nobody needs to be a villain for this to be true. A situation is coercive when the interests are irreconcilable and the resolution will probably be settled by whoever holds the budget.

You can run a coercive situation as though it were unitary. Organisations do it constantly. Everything you learn afterwards is then filtered through what people think it is safe to say, and your instrumentation tends to fail at exactly the moment you need it. An organisation that cannot tell you the truth cannot be steered.

4. Five ways to look at the same mess

Jackson’s more recent work, consolidated in the 2024 practitioner’s guide, replaced the grid’s front end with something more usable. Before choosing anything, look at the situation through five systemic perspectives. Each has a long history, each has produced real insight, and each is blind in a different direction.

The mechanical perspective sees a system built to do a job, with parts that can fail and be fixed. It asks about efficiency, throughput, defects, waste. The interrelationships perspective sees feedback, delay and unintended consequence; it asks what loops are running and where the leverage sits. The organismic perspective sees a living thing that must adapt to survive; it asks about environment, viability and learning. The purposeful perspective sees people who hold different worldviews and pursue their own ends; it asks what the various parties think is going on and who has the power to make their version stick. The societal and environmental perspective asks who else is affected, including people with no seat at the table, and what the intervention does to them.

Run those five over the same AI programme and you get five different accounts of the same trouble.

Mechanically, the pipeline has a bottleneck at review; generated code arrives faster than anyone can verify it. In terms of interrelationships, the faster generation raises the review load, which lengthens the queue, which pushes people to review less carefully, which raises the defect rate, which raises the review load again. Organismically, the organisation is not learning: the metric it watches is output, so it cannot see the compounding cost. From the purposeful perspective, the reason it watches output is that output is what the business case promised, and the people who wrote the case need it to be true. From the societal perspective, the junior engineers who would once have learned by doing the work the machine now does are being quietly deskilled, and they were never asked.

All five are correct. Only the first one is on the dashboard. Jackson’s discipline is to insist that the exploration produces a primary issue and secondary ones, and that you are able to say why you picked the primary. Most programmes never get there, because they only ever had one lens and they mistook it for the situation.

5. EPIC, and the loop that evaluates the instrument

Critical systems practice is Jackson’s answer to the obvious objection, which is that a philosophy of pluralism does not tell anyone what to do on Monday. He set it out across four papers in Systems Research and Behavioral Science between 2019 and 2023 and consolidated it in the 2024 book. Four stages: Explore the problem situation, Produce an intervention strategy, Intervene flexibly, Check on progress. He calls it EPIC, and the acronym is doing more work than it looks.

Explore is the five perspectives, applied before any method is selected. Produce designs the combination: one dominant approach carried by the primary issue, others in support, with objectives and a schedule attached so the thing is real. Intervene runs it, and runs it flexibly, which means the dominant method can be swapped when the situation shifts and the swap is a planned move rather than an admission of failure. Check evaluates what happened.

Check is where the discipline lives, because it evaluates two things. It asks whether the situation improved, which every programme claims to ask. It also asks whether the methods were the right ones, which almost nobody asks, because by the time you get there the method has become the identity of the team running it. I have sat through a great many programme reviews and I cannot remember one that asked whether the method had been the right method. Say that a soft systems approach surfaced the real disagreement but had no way to fix the queue, and you have learned something you can use twice: once about this situation, and once about your own instrument.

The double loop has come up in this series before. Argyris found that organisations correct actions and protect the assumptions underneath them. Schön found that the practitioner’s real skill lies in reframing the problem while inside it. Ohno’s insistence that you go and see stands behind the whole of Check. Jackson applies the same move to the toolkit itself. The four stages are iterative in practice, not sequential in the way a bar chart is sequential, and the output of each pass is a sharper account of what kind of situation you are in.

6. Pluralism is not helping yourself to whatever you like

There is an easy misreading of all this, and it is worth killing early. Combining methods is not the same as being relaxed about methods. Jackson’s pluralism is disciplined, and it rests on three commitments that only do their work when they are held together.

Critical awareness means knowing what each approach assumes and what it therefore cannot see. Every method carries a theory of the organisation. Pick one without saying so and the theory arrives anyway, unexamined, and starts making decisions on your behalf.

Pluralism means using approaches from genuinely different traditions together, on purpose, each retaining its own assumptions rather than being flattened into a house methodology. The hard part is that the traditions are not compatible. A mechanical view and a purposeful view of the same organisation contradict one another, and the practitioner has to hold both without dissolving the tension into mush. Total Systems Intervention, which Jackson developed with Robert Flood in 1991, was the first serious attempt at this; critical systems practice is what it became after thirty years of finding out where it broke.

Emancipation means asking whose interests the intervention serves. That commitment is what makes the tradition critical rather than merely clever, and it does most of its work through boundary critique. Werner Ulrich’s argument, which Jackson carries, is that every improvement has a boundary. The boundary decides who counts as a stakeholder and whose knowledge counts as expertise. That is a value judgement, and it is usually presented instead as a technical decision that somebody made for sound operational reasons. An AI productivity programme whose boundary includes the engineering leadership and the vendor, and excludes the people whose work is being measured, has already answered the question of whose improvement this is. It just has not said so.

You can hear the objection forming, because I have made it myself. This sounds like a lot of philosophy for a Tuesday. Jackson would half agree; the tradition has been criticised for years, sometimes by its own, for being more comfortable in the seminar than in the situation. His response over the last decade has been to reach back to the pragmatists, to Dewey and James, and argue that the test of a perspective is what it lets you do. Use the ones that work here. Say why. Be ready to be wrong about it.

7. What a machine cannot do for you

An AI will produce a change methodology in about nine seconds. It will be coherent, it will be referenced, it will have phases and workstreams and a RACI, and it will look exactly like the output of a competent consultancy, because that is what it was trained on. The cost of a plausible method has collapsed. A wrong method costs what it always did.

That gap is the whole of the current risk. When method was expensive, the expense itself imposed a rationing discipline; somebody senior had to justify the engagement, and the justification was at least a moment where the question got asked. Now the deck arrives free and arrives first, and it arrives with the situation already classified, silently, by whatever the training data thought normal. Which is unitary. Almost always unitary, because that is how change programmes are written up.

Ask the machine to design your transformation and you will most likely get a unitary method for a pluralist situation, delivered with total confidence, in a document nobody will interrogate because it reads as though someone already did.

The same trap sits inside agentic systems themselves. An agent estate is mechanically legible in a way most organisational subjects are not: you can see the calls, the tokens, the latency, the failure rate. All of that is real and worth watching. It also makes the mechanical perspective irresistible, because it is the only one with numbers, and a perspective with numbers will beat four perspectives without them in any meeting you have ever attended. The purposeful view of an agent programme, which is that different parts of the organisation want incompatible things from it, generates no dashboard at all.

So the human contribution is not the plan. The machine will write a better-formatted plan than you will. The human contribution is the classification of the situation and the choice of approach that follows from it. Then the willingness to defend both, in a room where a free and beautifully formatted alternative is already open on somebody’s laptop.

8. Method as the exercise of agency

Return to where the argument started. Somebody in the review says a name, the room relaxes, and a crisis becomes a personnel matter. That room has one method for crises and it has never chosen it. Girard’s account of why is uncomfortable and hard to unsee: the group discharges its tension onto a single figure because the discharge works. Heifetz says the leader’s job is to hold the disequilibrium rather than relieve it. Jackson says something narrower and more usable. The leader’s job is to make the choice of approach explicit, so the default one can be argued with.

Kotter’s eight steps are not wrong. They are a complex-unitary method, they work well when the situation is complex and unitary, and they have been applied for thirty years to situations that were neither. Ackoff’s dissolving is a real move and it belongs to a particular kind of mess. Beer’s model diagnoses viability and assumes the parts want the whole to survive. Each of these thinkers gave the field a real instrument. Jackson is the one who insists that having several instruments obliges you to say which you are picking up and why.

This is what custodial agency looks like in practice. You did not cause the event. You cannot make people want the same thing. What you can do is refuse to run the default method on a situation it does not fit, and that refusal is visible, unpopular and cheap to make. It costs one meeting. In most organisations it may also be the only moment in the whole programme where a real choice is available, because once the method is set everything downstream follows from it and looks like execution.

Choose the approach in the open. Say what it assumes. Say who it serves. Then let people argue with it, which they will, and the argument is the first honest information you will have had about what kind of situation you are actually in.

Placement: Transforming, the fourth and last of the series' phases; Learning, Deciding, Building, Transforming, which starts at Agency. Jackson's subject is the first act of agency there: choosing which approach the mess actually wants, then holding it through disciplined practice rather than sponsoring it from a distance. The Event Is Not Yours to Make opens this stretch.

(An Organisational Prompt is something you can do now....)

Organisational Prompt

Classify the situation before you defend the plan

Take the largest change programme you currently own. Give yourself one page and forty minutes, alone first.

Write five short paragraphs, one per perspective. Mechanically, what is the bottleneck? In terms of feedback, what loop is running and what does it amplify? As a living thing, what is this organisation failing to learn? Purposefully, what do the four or five main parties each want, stated as they would state it rather than as the deck states it? Societally, who is affected who has no seat, and what is happening to them?

Then answer two questions in writing. Is this situation unitary, pluralist or coercive? And which method are you actually running, whether or not anyone named it?

If the honest answers are “pluralist” and “a unitary method”, you have found the thing that will defeat the programme, and you have found it early enough to matter. Take both answers to your next steering meeting and put them on the first slide. Whatever comes back at you is the most useful data the programme has produced so far.

Further Reading

Pragmatism and critical systems thinking: Back to the future of systems thinking is Jackson’s own short statement of where the tradition should go next. Free in full, and the clearest short thing he has written.

Martin Reeves interviews him for the BCG Henderson Institute in Critical Systems Thinking with Michael C. Jackson, covering EPIC and the five lenses, aimed at leaders rather than academics.

Critical systems practice 1: Explore, in Systems Research and Behavioral Science, 2020, is the first of the four EPIC papers and arguably the one that matters most, because getting Explore wrong tends to make the situation worse.

An extended summary of the 2024 practitioner’s guide, with concept maps, sits at CSL4D. Worth reading if you want the architecture of the book before you buy it.

The full statement is Critical Systems Thinking and the Management of Complexity, Wiley, 2019, with the history of every methodology in the grid and a fair account of the objections to his own position.

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

Forty minutes into the review, somebody says a name. It comes out as a clarification rather than an accusation, and the room relaxes immediately, because the question of what happened has just been converted into the question of who, and the second question has an answer. Twenty minutes later there is an action list. Nobody in that room is a bad person and nobody set out to do this.

What you have just watched is an organisation declining to learn, efficiently and with everyone’s cooperation. This is the commonest resolution of a crisis anywhere, it works, and it leaves the organisation exactly as it was, minus one person. These articles are about the alternative, and about how somebody has to choose it deliberately, in front of everyone, while the pressure to do the other thing is at its highest.

Three phases have run. Learning asked what an organisation has to be like before anything can get through to it, and found that learning is a condition rather than a process. Deciding asked how choices get made under constraint, and found that a decision is a design problem and design is a sequence of decisions. Building asked what happens to the craft when making becomes cheap, and found that judgement, verification and coordination are what remain. Each part handed something to the next. Building hands this one an organisation that has learned to describe what it made, and now has to decide what that description means for how it is led.

The claim here is narrow and it costs something to accept. Leading is taking a position on how change resolves. Not starting the change, not driving it, not owning it. Taking a position on how it ends, at the moment when it could end several ways and the cheap way is the one everybody is already moving towards.

1. The event you did not make

Start with what a leader cannot do, because most leadership writing starts with what a leader can.

The argument running under this whole series is that organisations are not changed by people. They are changed by events. An event is something that happens which makes the existing account of what is possible untenable, in public, in front of witnesses who cannot unsee it. The demo that works. The outage that reveals what everyone privately suspected. The competitor who ships the thing your architecture said would take two years. After an event, the sentence “that is not how we do things here” stops being an argument and becomes a description of a place that no longer exists.

Leaders do not manufacture these. That claim gets resisted, because a great deal of consulting revenue depends on the opposite one. Kotter’s first step is to create a sense of urgency, and thirty years of practice have shown what that produces when there is no actual event underneath it: anxiety without information. People can tell the difference between a crisis and a slide about a crisis. Manufactured urgency spends the organisation’s attention and buys nothing, and it makes the next real event harder to hear, because the alarm has already been sounded once for nothing.

What a leader can do is create the conditions in which a latent crisis surfaces as an event rather than being discharged quietly. Michael Jackson’s word for the leader’s job in a systems intervention is close to what is needed here: the agency is custodial. You hold the space. You choose the approach and defend it in the open, and you sustain what previous change achieved. The rupture is not yours to generate. Believing you did is the most common and most expensive self-deception in this trade.

2. Why the room reaches for a name

Back to the review meeting. Something has gone wrong, the account that explained the world has failed, and the discomfort in the room is real and rising. René Girard spent a career on what happens next, and his answer is unpleasant enough that people usually hear it once and then forget it deliberately.

Girard’s starting point is that desire is imitative. We want what we see others wanting, which means that the closer two parties are, the more likely they are to want the same thing and to become rivals over it. Rivalry spreads by the same imitation that started it. What began as a difference of position becomes a general condition in which everybody is behaving identically and nobody can tell the factions apart. Girard calls this the crisis of undifferentiation, and it is recognisable in any organisation where a real problem has turned into eight teams briefing against each other about the same problem.

The crisis resolves, historically and reliably, through a single victim. The community converges on one person or one group, the choice is more arbitrary than anyone involved believes, and the expulsion produces genuine relief. Peace returns. And the mechanism conceals itself: the participants do not experience it as scapegoating, they experience it as having correctly identified who was responsible. A room that has just scapegoated somebody does not feel like a mob. It feels like a group that has finally, after some difficulty, got to the bottom of things.

You do not need a corporate execution for this to be running. The modern forms are gentler and just as effective. The reorganisation that removes one function. The supplier who becomes the reason. The departing executive whose leaving explains everything that happened for two years. Each converts a systemic question into a personnel answer, and the relief is real, which is precisely the problem, because relief is what the organisation was seeking rather than understanding.

3. Refusing the cheap resolution

If the leader cannot make the event, the leadership sits in what happens to it. And the first thing that happens to any event is that the organisation tries to close it, fast, with the least expensive story available.

Refusing that is the central act. Nothing about it is grand. Usually it is one sentence in one meeting, said by the person with enough standing to be listened to, and it costs them something with everybody present.

Ronald Heifetz supplies the operational half of this. His term for the state the organisation is in after an event is disequilibrium, and his finding is that adaptive work only happens inside it. Too little and nobody changes anything; too much and people panic, regress, and reach for a scapegoat. The leader’s job is thermostatic: keep the heat inside the productive band, and resist the pressure to relieve it. Every instinct in the room, including yours, wants the temperature down. The reward for lowering it is immediate. The cost is that the organisation returns to exactly the state that produced the event, having learned that the event was one person’s fault.

Alasdair MacIntyre gives this its philosophical shape, and his term is more precise than “crisis”. A tradition, meaning any community with shared standards for what counts as good work, enters an epistemological crisis when its own resources stop being able to resolve its own problems. Argument stops converging. The tests that used to settle things no longer settle them. The way out, when there is one, is not a return to first principles: it is the arrival of a new account that solves the problem and also explains why the old account failed and had to fail. That second condition is the demanding one, and it is what separates a real resolution from a scapegoat. Naming a person explains nothing about why the standards stopped working.

So the test in the room is quick. Does the proposed resolution explain why the previous account was inadequate? If it does not, it is a story about a person, and the organisation is about to buy peace with information it needed.

4. The window that opens straight afterwards

Something happens to language in the weeks after a genuine event, and organisations consistently fail to use it.

The received wisdom is that changing how an organisation talks takes years. Mostly true. The exception is short and it is the one you can use. George Lakoff’s work on framing explains why. Frames are not decoration on top of an argument; they are the structures that decide which facts count as relevant and which conclusions feel obvious. They are built up by long repetition, and they hold because the world keeps confirming them. Lakoff’s famous demonstration is that a frame cannot be fought by negating it, because the negation invokes it. Tell people not to think of an elephant. The elephant is already there. Argue against tax relief and you have conceded that tax is an affliction.

Now put that beside an event. The event’s whole function is to break the confirming evidence. For a short period the old frames are unsupported, and vocabulary that would have been rejected as naive six weeks earlier installs quickly and sticks. This is why the language an organisation uses about AI was mostly set in a two-month period nobody planned, by whoever happened to be speaking with confidence at the time.

The stakes are practical, because the metaphor decides the policy. Call the system a copilot and you have implied a human in command, a defined cockpit, and a division of labour that mostly does not exist. Call it an agent and you have implied something that acts on its own account, which raises the question of what it may promise on your behalf and to whom. Call the programme adoption and it is a rollout with a completion date. Call it capability and it is a standing investment with no end. None of these are the same organisation eighteen months later, and the choice between them is usually made by accident, in the first fortnight, by whoever wrote the deck.

The leader’s job at this position is to have the vocabulary ready before the window opens, and to use it immediately when it does. That sounds cynical. It is the opposite: the alternative is not neutrality, it is somebody else’s frame arriving by default.

5. Settling, which is not the same as finishing

New language that does not become structure evaporates. Everybody in this trade has watched a vocabulary flourish for one quarter and disappear, because it never got into the standards, the funding lines, the review gates or the job descriptions.

Kurt Lewin is the thinker at this position, and there is a correction to make before he can be used, because most of what is taught under his name he did not write. The three-step model of unfreeze, change, refreeze is the most cited framework in change management, and the archival work of Stephen Cummings, Todd Bridgman and Kenneth Brown shows it was assembled by later writers and attributed backwards after his death in 1947. Lewin used the ice-block image in passing. He did not propose a three-stage method, and the model that carries his name arrived with an assumption he would probably have rejected: that change is an interval between two periods of stability.

What Lewin actually gives you is better. Field theory holds that behaviour is a function of the person and the environment together, so if you want different behaviour you change the field rather than exhorting the people. Force field analysis follows directly: any current state is held in place by driving and restraining forces in rough balance, and the productive move is almost always to remove restraint rather than to add drive. Most transformation programmes do the opposite. They add incentives, campaigns, dashboards and mandates on top of restraints nobody has touched, and then wonder why the field snaps back.

And action research, which he also founded, treats the intervention itself as the way you learn about the system. Studying an organisation and then changing it is the wrong order. You change something and find out what it was.

Read that way, the settling at the end is provisional by design. Structure has to be built so that the next event can unfreeze it, which means writing down what would have to be true for this arrangement to be wrong. An end state that has quietly announced it will resist the next event will keep that promise.

6. Why the rotation starts where it does

Each part of this argument moves through the same four positions, and the order changes according to what it is for. This one runs Agency, then Event, then Language, then Structure.

It starts at Agency because that is the only position a leader occupies before anything has happened. No event has happened yet. What exists is a person with a role, some standing, a set of prior commitments, and a choice about what kind of leader to be when something does. What they do with that determines whether the latent crisis in the organisation surfaces as an event at all, or gets discharged in the quiet ways organisations have for keeping the peace.

Then the event arrives, from the system rather than from the leader, and the leader’s position on its resolution decides everything downstream. Language follows, in the window the event opens. Structure comes last, holding the language long enough for it to become how things are done, and built to be opened again.

Where the previous phases ended by handing something forward, this one ends by handing something back. The settled structure becomes the ground on which the next disruption lands, which is where Learning began. The whole shape of the argument sits in that handover: not a programme with a completion date, but a cycle that leaves the organisation better at recognising and using the next event than it was before.

7. Release

Each part has a single word for what it is trying to produce. Learning was fidelity, the discipline of staying faithful to what an event revealed rather than to what was comfortable. Deciding was actuality, the insistence on the situation as it is rather than as reported. Building was simplicity. This one is release.

Release is the least fashionable of the four and the hardest to fake. It means the change no longer requires you. The vocabulary is in the standards, the structure holds without your attention, the people who must do the work are doing it, and your continued presence adds nothing to it. A leader who cannot release has not finished a transformation; they have built a dependency and given it a name.

The personal cost of this is real and rarely stated. The reward structure of most large organisations is set up to punish exactly this. Visible ownership of a live programme is career capital. A programme that no longer needs you is a programme somebody else can be given. I have watched capable people hold on to a transformation months past the point where holding on was the only thing slowing it down, and in every case they could have told you the theory perfectly.

Release is also what makes the flywheel turn. An organisation that has been through one full cycle and been released is measurably better at the next one. It carries a memory of an event that produced something rather than a casualty. Its language is language it trusts. Its structure is structure it knows to be provisional. The next event arrives into an organisation with more capacity than the last one had, which is the only definition of transformation this series is willing to defend.

8. What this is not

Two things, so that expectations are set correctly.

Nothing here is about leadership styles. There will be nothing here about whether you should be more visionary or more servant, and no quadrants. The question is not what kind of person you are; it is what you do in the twenty minutes when a crisis is resolving and the cheap option is on the table.

Nor is it heroic. The custodial reading of agency cuts both ways: it relieves you of the fantasy that you cause change, and it hands you a duty that is harder than causation. You have to be present, in the room, holding a position that costs you something, at the moment when everyone in it would rather the discomfort stopped. That is a smaller job than the one on the conference stage, and the one that decides whether anything survives.

The rest of these articles is about how to do it, thinker by thinker, and about what to do when it goes wrong, which it will.

Where this sits: Transforming, the last of the series' four phases; Learning, Deciding, Building, Transforming. It starts at Agency, because building hands over knowledge and capability rather than a crisis, and nothing about that compels anybody to act. What an organisation gets out of the work that follows is release: the change carries on without the internal agent who initiated it. From Building to Transforming is the bridge into it.

(An Organisational Prompt is something you can do now....)

Organisational Prompt

Write the two sentences before you need them

Think of the last significant failure in your organisation, the one that got a review and an action list. Find the meeting notes if they exist.

Answer one question honestly: did the resolution explain why the previous account of how things worked was inadequate, or did it identify a person, a team or a supplier? If it identified someone, the organisation bought relief and paid for it with the information it needed.

Now write two sentences and keep them where you can reach them in a meeting. The first refuses the cheap resolution without accusing anyone of seeking it: something close to “before we agree what happened, I want to be clear about what we thought was true that turned out not to be.” The second is the vocabulary you want installed if the window opens: one phrase, chosen deliberately, that you will use the same way every time for the next six weeks.

Both sentences take ninety seconds to prepare and are almost impossible to compose under pressure. That is why you write them now.

Further Reading

René Girard, The Scapegoat (Johns Hopkins University Press, 1986). The mechanism set out at book length and without the literary apparatus of the earlier work. Unsettling in a way that is difficult to shake off afterwards.

Ronald Heifetz, Leadership Without Easy Answers (Harvard University Press, 1994). Still the most useful thing written about what a leader does when there is no answer to supply.

George Lakoff, Don’t Think of an Elephant! (Chelsea Green, 2004, updated 2014) is the popular version. His Framing 101 essay is free in full and covers the essentials in a few pages.

Stephen Cummings, Todd Bridgman and Kenneth Brown, Unfreezing change as three steps: Rethinking Kurt Lewin’s legacy for change management (Human Relations, 2016). The archival work showing that the field’s founding model was never proposed by the person it is credited to. Read it before you draw the three boxes again.

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

A working system is not a transformed organisation. The system is a fact about what the organisation can now do, and the distance between that fact and a transformation is what this article is about. Building ended with something real: a system in production, a group of people who know things they did not know a year ago, and a vocabulary for the work that nobody had before they did it. All of that is hard-won. None of it, on its own, changes anything. There are enterprises running excellent systems, built by teams who learned a great deal, inside structures that are precisely what they were.

Each part of this series ends with something the next part cannot use directly. The gap is the point. Learning ended with a capacity, and a capacity is not a description, so Deciding had to begin by producing one. Deciding ended with a specification; a specification is not a system, so Building had to begin by constructing one. Building now ends with a system and everything the building of it revealed, and none of that is a change in the organisation. Which means the last part has to begin somewhere else entirely: with a person taking a position.

1. What Building Produced

Building ran its own path. It began at Structure, because a specification has to be constructed before it can do anything, and construction is first a structural act. The work gets divided into teams and the system gets divided into components. Those are the same division, whether or not anyone admits it. It moved to Agency, in two senses at once, the autonomy of the team that owns a component and the promises the component itself makes about what it will and will not do. It moved to Event, where the built thing met reality and returned a verdict, and where the useful result was always the disconfirming one. And it ended at Language. The operational knowledge of a system in use is a description nobody could have written in advance.

That final position feeds straight back to the first, which is why Building is iterative in a way the earlier phases are not. Continuous integration and continuous delivery are not delivery fashions; they are this cycle running fast enough for the learning to compound.

Building hands over three things rather than one. First comes the artefact, the system itself, which is the least interesting of the three. Second comes a capacity, which Building called simplicity and which is more precisely the ability to subtract. Remove a component. Stop a line, kill a feature, say no to generation that nobody asked for. Third comes a vocabulary, the operational language of people who have watched their own assumptions meet production and lose.

There is also a fourth thing, unlisted and rarely welcome. Building redistributes competence. The people who did the work now know something the people who commissioned it do not, and some of what senior people used to know has quietly stopped being the valuable kind of knowing. A build that changed nothing about who is expert in what was probably not much of a build.

2. Why a Working System Is Not a Transformation

The intuitive model of what happens next is diffusion: the thing worked, so it spreads, and leadership’s job is to sponsor the spreading. That model has now failed publicly three times in the working memory of most readers, and each failure looked the same from the outside.

Agile spread as a set of ceremonies while the funding model, the reporting lines and the annual planning cycle stayed where they were. DevOps produced DevOps teams, which is the one structure the idea rules out. Cloud arrived and was billed as a data centre with a different invoice. In every case the tools were absorbed, the vocabulary was captured, and the shape of the organisation came through unchanged. I have sat through all three of those waves, and each one felt at the time like the exception. The absorption was not stupidity. It was the organisation doing what an organisation does with new capability in the absence of anybody taking a position on it.

The mechanism runs like this. A build changes what is possible. It does not decide what the new possibility means, and meaning does not assign itself. Where nobody takes a position, the organisation resolves the question by default, and the default is restoration. The new capability gets fitted into the old shape, and the shape wins, because the shape is what everybody is paid by. A capability absorbed without a position taken on it is a capability neutralised.

Which is why Transforming is not a delivery function or a communications exercise. Its subject is the resolution: how the disturbance a build creates gets settled, who settles it, and whether the settlement leaves the organisation able to do this again. The hypothesis here is that transforming is taking a position on how change resolves. Everything that follows rests on that one sentence.

3. Why Transforming Starts at Agency

Each part enters the cycle at the point the previous one forces, and the last begins at Agency for a reason that is easy to miss.

There is no external shock arriving. Building did not hand over a crisis; it handed over knowledge, capability and a slightly disturbed distribution of expertise. Nothing about that compels anybody to act. So Transforming cannot open at Event, because there is no event yet, and the whole error of the conventional change literature is to assume one will be provided by circumstance. It opens instead with somebody deciding what to do in a situation that does not demand anything in particular.

Michael Jackson holds this position, set out across two books in 2003 and 2019, and his contribution is more practical than his field’s reputation suggests. Faced with a genuine mess, one with no agreed problem statement and no agreed owner, the leader’s real skill is not choosing a solution. The skill is choosing which approach to use. Then it is orchestrating that approach through disciplined practice rather than sponsoring it from a distance. Different messes want different methods, methods carry assumptions about power and purpose, and picking the wrong one produces confident work on the wrong question. That act of choosing, made explicitly and defended, is what agency looks like when it is executed rather than announced.

There is one further claim about that agency, and it is the claim most likely to annoy people who have read a lot of leadership books. The leader does not generate the event. The agency here is custodial: it creates and holds the space in which an event can emerge, protects the change already made from the organisation’s appetite to restore itself, and stays out of the way of what follows. That is a smaller account of leadership than the field usually offers and a considerably harder one to live up to, because it removes the part everyone enjoys.

4. What Transforming Will Argue

The rotation runs Agency, then Event, then Language, then Structure, and the shape of its argument can be set out here.

After Agency comes Event, and this is the sharpest move here. René Girard, in Violence and the Sacred in 1972, read collective crisis as mimetic. People want what the people around them want, the distinctions that used to order a group collapse, and rivalry becomes general. The group then resolves the whole thing by converging on somebody to blame and expelling them. Afterwards it feels restored. It is unchanged. The scapegoat mechanism is the most efficient machine ever devised for converting a real crisis into no learning at all, and every one of us has watched it run, usually shortly after an incident review. The leader’s first act at this position is refusal: declining the sacrifice at the moment the organisation most wants it. Refusal alone will not do, because the energy is real and has to go somewhere, and that is where Ronald Heifetz, writing in 1994, works alongside Girard. The craft is holding the disequilibrium at a temperature people can survive, long enough for the adaptive question to be asked, without turning the heat down so far that the question goes away. Underneath both sits Alasdair MacIntyre’s epistemological crisis: the point at which a tradition can no longer explain its own failures in its own terms. A tradition in that position becomes able to change, and is most tempted to blame an individual instead.

Then comes Language, and George Lakoff supplies the mechanism, as he has since 1980. Frames are ordinarily slow to move, which is why most internal communication achieves nothing. The exception is the period immediately after an event has dissolved the supports the old frame rested on, when new frames install with startling speed. What the organisation starts calling things in the weeks after a crisis becomes what it believes about them for years, and the naming is usually done carelessly, by whoever gets to the all-hands first. I have watched a company’s whole understanding of an outage get settled by one phrase in a Friday email. The window is short, it closes without warning, and it does not reopen because somebody senior has finally got round to thinking about the wording.

Then Structure, where Kurt Lewin returns with his most abused idea corrected. Refreezing, described in 1947, was never meant as the point at which change stops. Read properly it is provisional settling: the new arrangement holds firmly enough to be worked in and loosely enough to be revised. A structure that presents itself as the end state has quietly announced that it will resist the next event, and it will keep that promise.

The capacity what comes out of it is is release. Not delegation, which is a management technique, and not detachment, which is a mood. Release is the ability to let the change carry on without you. It means having made an organisation that no longer requires the person who held it open, and being willing to see the thing you started transformed by an event you did not choose. Release is the hardest of the four capacities, which is why it comes last.

5. The Handover

So the bridge can be stated plainly. Building produced a system, a capacity to subtract, an operational vocabulary, and a disturbance in who knows what. Transforming takes that disturbance and refuses to let it settle by default. It begins with somebody choosing an approach and standing behind it, holds the crisis open rather than resolving it onto a person, names what happened while the naming still moves, and settles into a structure that admits it is provisional.

Look at the four phases together and a pattern shows up that was not designed and is probably not a coincidence. Learning ran Event, Language, Structure, Agency. Deciding began one position later, at Language. Building started one later again, at Structure. Transform begins at Agency, which is the fourth, and the four phases together make one complete iteration of the same four moves, at the scale of the whole change rather than any part of it.

Which means the last part does not close anything. The settlement it produces becomes the arrangement the next event will disturb. An organisation that has been through this once is better at recognising an event and better at making one, because it has a vocabulary for what happened and people who survived it. This is the argument the series has been making since the first article. People do not change organisations; events do. And an organisation that resolves an event well ends up with more events available to it, not fewer, which is why positive change compounds and why the alternative is not stability but slow arrest.

Building asked how an organisation makes the thing real and stays honest while reality tells it what it got wrong. Transforming asks the question the whole series has been walking towards, and it is a personal one: what kind of leader can you be in the organisation you are actually in? That is where the series goes next.

One orientation before the change of question. The series runs four phases; Learning, Deciding, Building and Transforming. Each starts one position later than the last in the same four moves, so the four together complete a single turn: Learning at Event, Deciding at Language, Building at Structure, Transforming at Agency. From Deciding to Building was the previous bridge and The Build Is the Test opened the stretch this one closes.

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

Ask an engineering organisation how much faster it has become and you will be handed two numbers that cannot both be right. Developers report they are somewhere between thirty and sixty per cent faster, and they are not lying; the experience is real, and they can show you the diff. The delivery numbers, measured where work becomes something a customer can use, have barely shifted. The 2024 DORA research found that as AI adoption rose across teams, throughput dipped slightly and stability dipped further.

Both numbers are accurate, which is what makes the gap worth staring at. Something between the keystroke and the delivery numbers is eating the gain, and these articles have spent fifteen articles naming it. The answer has stayed consistent enough to state in one sentence: making went free, the constraint moved somewhere else, and the organisation did not follow it. Money went to the part of the work that had just become cheap. Nothing went to the parts that were always expensive and had merely been hidden behind the labour of typing.

There are five failures underneath the gap. Not one of them is a technology failure, which is why no tool has fixed any of them. They are failures of building: of who builds, of what the building has to carry, and of how the builders relate to each other and to the machines now working alongside them. Building is the subject, and building is what the organisation has quietly stopped being able to do.

1. Brooks sorted this in 1986 and nobody re-read him

Fred Brooks published No Silver Bullet in 1986, and the distinction it draws still sorts every claim anyone makes about an engineering tool. Difficulty in software comes in two kinds. Accidental complexity is the labour surrounding the work: typing, boilerplate, wiring, ceremony, the tax you pay to express an idea in a machine. Essential complexity is the irreducible part: deciding what to build, working out where the seams belong, judging whether the result is any good. Brooks’s claim was that no tool ever touches the second kind. Forty years have not dented it.

AI is the most complete accidental-complexity machine the field has produced. Its effect on that layer is not reduction; it is collapse. The cost of incidental work has fallen far enough to change what the job feels like from the inside, which is why the thirty to sixty per cent is honest. What survives is everything Brooks called essential, and it is now exposed, unsupported, and carrying the whole weight of the outcome.

So the classics did not survive the wave by luck. Modular decomposition, information hiding, small batches, refactoring, review: these are the essential core, left standing once the scaffolding around them came away. The discontinuity and the endurance of the old disciplines are one claim seen from two sides.

Which produces the diagnosis this article is going to argue. Your organisation cannot build because the capabilities that now determine whether a build succeeds were thinly staffed for twenty years and nobody noticed, because typing concealed them. Verification was a background cost. Design was something a few people did on Fridays. Answerability was assumed and never assigned. Remove the concealment and all three are suddenly the job, performed by an organisation that never resourced any of them.

2. Nobody is answerable for the change that just merged

Mark Burgess holds the question of who, or what, builds, and Promise Theory gives the discipline in one line: an agent can promise only what lies inside its own locus of control. Promises are voluntary and best-effort. Impose an obligation from outside and you create what Burgess calls a fictitious coupling, a dependency the system believes in but cannot honour. Fictitious couplings are what make systems brittle, and they are what most AI governance consists of.

Look at what your organisation issued in the past year. An acceptable-use policy, a set of coding standards, prompt guidance, a slide in the all-hands about human oversight. Every one of those is advice offered to somebody who may choose to follow it on a Thursday afternoon when the release is late. Birgitta Boeckeler’s distinction is the operational form of Burgess: guides tell an agent what to do and can be ignored; sensors observe what it actually did and can stop the work. Static analysis is a sensor. A blocked merge is a sensor. A markdown file of standards is a guide, and so is a governance framework, however many pages it runs to. Most organisations, sorting their own controls honestly, find that everything they have is a guide.

Underneath that sits a slower failure. Addy Osmani named the seventy per cent problem: the model gets you most of the way and leaves the last portion, the edge cases, the integration, the production behaviour, exactly as hard as it ever was. Charity Majors has been blunt about what that does to a career ladder, because the first seventy per cent is the work that used to make juniors into seniors. Richard Sennett explains why the loss is not recoverable by training. Craft knowledge is built through resistance: the hand meets a material that pushes back, and the head learns something it could not have been told. Remove the resistance and you remove the mechanism.

Put the three together and the shape is unpleasant. Verification is becoming the whole job at the exact moment the pipeline that produced people capable of verifying is being closed. This is not a prediction. It is arithmetic on a hiring plan.

3. The organisation cannot stop

Taiichi Ohno held that overproduction is the worst of the seven wastes, and his reason was not sentimental. Overproduction manufactures the appearance of progress and hides the other six behind it. A team producing four times the output feels four times as productive and is accruing four times the liability.

Cost was only ever half a tax on production. The other half was the friction restraining overproduction, and that half is gone. Every line you once had to write by hand was a line you thought twice about. Nobody thinks twice now, and the danger has moved from the price of generating to the price of keeping. Generated code cannot be thrown away once it exists; it must be read, maintained, secured and supervised for the life of the system. The bill arrives late and it arrives forever.

Generated code is code. Somebody reads it like code, reviews it like code, patches it like code and audits it like code, and none of those costs fell when the cost of writing the code fell. What changed is how much code there is.

Ohno’s two counter-disciplines are pull and jidoka. Pull means real demand triggers the build, and almost nothing in a modern backlog process can tell real demand from available capacity. Jidoka means stopping the line the instant a defect appears rather than producing confident defective output at volume. Pull and jidoka are both refusals, and refusal is the capacity most organisations have bred out of themselves.

I have sat in a great many steering meetings and I have never once watched an executive receive good news for stopping something. Stopping reads as a throughput failure, so it is renamed a pause, a rebaseline, a phase two. The andon cord exists on every Toyota line because pulling it is safe. In most enterprises it is not safe, and the cord is therefore decorative, which is a fact about the reward system rather than about anyone’s courage.

Lorin Hochstein found the sharpest version of this in the incident world. Using a model to assemble the ingredients of an incident report saves genuine toil. Using one to write the report removes the act in which the organisation learns anything, because the understanding was never in the document; it was in the writing. Any artefact whose value lies in the thinking it forces loses that value the moment it can be produced without the thinking. The document survives. The learning does not.

4. Somebody has to hold the graph

John Ousterhout holds the question of what the build must carry, and his definition of the enemy is precise. Complexity is anything that makes a system hard to understand and change. It shows up as change amplification, where a small modification touches many places; as cognitive load, where somebody must hold too much in their head to proceed; and as unknown unknowns, where it is not even obvious what you would need to know to change something safely. Those three name precisely what generated abundance costs.

His prescription is deep modules behind simple interfaces. The interface is what a person must understand; the depth is what they are spared. Get this right and machine leverage is enormous, because the agent can churn inside the module while a human reasons about the surface. Get it wrong, expose the complexity through a wide shallow interface, and every generated line becomes a line somebody has to comprehend.

Ousterhout is building directly on David Parnas, and Parnas supplies the half the industry dropped. His 1979 paper was called Designing Software for Ease of Extension and Contraction; the field took the extension half and built fifty years of practice on it. Contraction is the property you only discover you lack at the moment you need it. Every system has a uses relation, the graph of which component may invoke which. Where that graph is a hierarchy, every level downward is a shippable subset: you can cut anywhere and what remains still runs. Where it contains loops, the system has exactly one deliverable configuration, which is all of it.

No agent has ever seen your uses relation. Asked to add a capability, it attaches the capability wherever attachment is easiest, which is correct behaviour for the task it was given and corrosive behaviour for the graph. Each edge is defensible. The graph degrades anyway, one reasonable decision at a time, and the loss surfaces two years later when a regulator kills a product line or an acquisition needs one service carved out and the answer comes back that it cannot be separated.

5. The batch got bigger while everyone was talking about speed

Kent Beck and Martin Fowler built a practice on one insight: the way to survive change is to shrink the interval between decision and feedback. Small batches, continuous integration, refactoring as a daily habit. The economics were never in dispute. A small change that breaks tells you what broke it; a large change that breaks tells you almost nothing.

Generation quietly reverses this. A model produces a large, plausible, internally consistent change in one pass, and large plausible changes are precisely what the practice was designed to prevent. Review capacity did not scale with generation capacity, because review is a human act performed at human speed. The measurable signature is already in the data: refactoring is falling as a share of changed lines while short-interval churn climbs, which is what code being generated faster than it is understood looks like from the outside.

Fowler’s design stamina argument is the one to put in front of a board. Neglecting internal quality buys you speed for a few weeks and costs you speed for years, and the crossover arrives sooner than anyone estimates. That argument was made against schedule pressure. It now has to be made against a tool that removes the only remaining reason to keep changes small, which is that they were tiring to write.

The practical correction is deterministic enforcement rather than probabilistic instruction. You cannot prompt your way to a small batch. You can configure a pipeline that refuses one.

6. The topology did not change, and the ritual quietly stopped

Melvin Conway observed in 1968 that a system’s structure copies the communication structure of the organisation that produced it. That has not stopped being true; it has gained a new term. Your organisation chart now holds agents that nobody drew, reporting to nobody, communicating through whichever developer happens to be supervising them. If the architecture you are getting is not the architecture you asked for, the chart is the first place to look, and this year the chart is incomplete.

Matthew Skelton and Manuel Pais gave the operational version in 2019 with cognitive load as the sizing constraint on a team. A stream-aligned team should own no more domain than it can hold in its collective head. The load has just risen sharply, because supervising output you did not write is more expensive than writing it, and almost nobody has reduced a team’s domain to pay for it. Instead the same team owns the same surface and now spends its day reviewing machine drafts of it.

Randall Collins explains what that does to the team itself, and his argument is the least comfortable here. Interaction ritual chains generate emotional energy through bodily co-presence, mutual focus of attention, and shared mood. Energy is a product of particular interactions, and it decays without them. Collins rejects the managerial reading of cadence as a schedule. A stand-up is not a ritual because it recurs; it is a ritual when people are jointly attending to the same thing and leave with more energy than they arrived with.

Four people each running four agents have fewer of those moments than the same four people had in 2019, and the agents supply none of the deficit. An agent can make a promise; it has nothing at stake in keeping it, no relatedness, and no capacity to care whether it holds. That is the structural line between a human and a machine builder, and it falls exactly where the energy is generated.

Stafford Beer closes the section with the shortest test in the series. The purpose of a system is what it does. Whatever your build system claims about quality, its purpose is what it produces in production under load. Judged by that standard, a great many delivery organisations are running a machine for converting review capacity into unreviewed change, and doing it well.

7. Your metrics are working perfectly

Measurement is where the four failures above go invisible, and the dashboard is usually the last instrument anybody thinks to check.

DORA gave the field four measures that earned their authority honestly. Deployment frequency, lead time, change failure rate, time to restore. DX Core 4 extended the frame to speed, effectiveness, quality and impact, and made developer experience something you could measure. They are good instruments, and both are being read badly, because organisations under pressure to show an AI return reach for the two numbers that move fastest and mean least: acceptance rate and lines generated.

Apply Beer’s test to your own dashboard. If the reported measure is volume of generated code, the purpose of that dashboard is to increase volume of generated code, whatever the accompanying narrative says about quality. The metric is not broken. It is working perfectly, and it is producing exactly what it was designed to produce.

The measures worth watching now sit on the other side of the constraint: the ratio of building new to maintaining old, the share of changed lines that are refactoring, the interval between a change being generated and a human accepting it, and the time it takes to answer the question of what would break if you removed something. Nothing on that list flatters anybody in a quarterly review. All four tell you whether you can still build.

8. The five failures are one failure

Taken separately, each of these looks like a discipline problem with a discipline answer. Taken together they compound, and the order of compounding is the same order this series has used since Learning.

Identity fails first. No one is clearly answerable for a merged change, and the people who would have grown into answerability are no longer doing the work that grows it. Information fails next, because an organisation without answerable builders has nobody holding the design, nobody willing to stop the line, and nobody drawing the graph. Interaction fails last and hardest, because the topology and the cadence were the only remaining mechanisms for catching what the first two missed, and both have been degraded without a decision being taken about either.

Say it plainly. Nobody is answerable, nothing gets stopped, nobody holds the graph, the batches keep growing, and the topology stays where it was. Five failures, and each one feeds the next.

They also fail in the same way, which is why listing them separately is slightly misleading. Each one is a failure to stop. A change merges because nobody could stop it without a name against it. A line producing waste keeps running. A dependency joins a graph nobody holds. A batch grows. A topology absorbs agents nobody drew. Stopping is the single capacity underneath all five, and stopping is what free making removed the last natural restraint on. Cost used to stop things. Nothing stops them now except a decision, and a decision of that kind has to be taken by somebody answerable for it.

The failure stays invisible for a long time, and the invisibility is structural. Every individual step is defensible. Every agent addition is reasonable. Every deferred refactor has a good reason on the day. Every unpulled cord protects a date somebody committed to in good faith. The system degrades through a sequence of locally correct choices, which is precisely the mechanism Parnas described for software aging and Ohno described for waste, and it is roughly why exhorting people to be more careful tends not to work.

There is one further reason this wave punishes harder than the last three. When an organisation failed to adopt agile properly, it delivered slower than it might have; the cost was opportunity. When an organisation fails to build properly under free making, it accumulates a system it cannot understand, cannot subtract from, and cannot staff, at a rate limited only by its licence count. The cost is carried, it compounds, and it arrives on a balance sheet rather than in a retrospective.

9. Simplicity is a capacity, not an aesthetic

Each part of this series leaves an organisation with a capacity, and Building leaves simplicity. Not tidiness. The capacity to subtract. Subtract, and you can build. Fail to subtract and you can only add, and you will keep adding until the system you are adding to is a system nobody can hold.

An organisation that can build has three properties a visitor could verify in a morning. Somebody read every merged change and accepted it by name. Controls sit in the pipeline as sensors rather than in a deck as advice. And it can stop: it can kill a feature, halt a release, and remove a component without a fortnight of archaeology.

Everything else here is instrumentation for those three. Deep modules and clean interfaces exist so a person can be answerable for something they can actually understand. Small batches exist so stopping is cheap enough to do. Team topology and cadence exist so the people doing the verifying have the energy and the context to do it. Promise Theory explains why an imposed obligation will not substitute for any of it, and Brooks explains why no tool will.

The uncomfortable part is that none of this is new. I have not described a novel discipline; I have described the practice of good engineering organisations in 2015, now made compulsory. Free making did not invent a new set of virtues. It removed the slack that let organisations get away without them.

The next article closes Building and hands to Transforming, where the question stops being what you can build and becomes what kind of leader you can be while the building changes underneath you.

Where this sits: this is the synthesis that closes Building, the third of the series' four phases; Learning, Deciding, Building, Transforming. The articles on building start at Structure and ask whether an organisation can build an effective response. What an organisation gets out of that work is a system, of software, process, organisation and so on, and a descriptive operational vocabulary nobody could have written in advance. The Build Is the Test opened it, and Why Your Organisation Can't Decide is the matching synthesis one step back.

Further Reading

Fred Brooks: No Silver Bullet: Essence and Accidents of Software Engineering (1986), freely available as Brooks’s own University of North Carolina technical report at https://www.cs.unc.edu/techreports/86-020.pdf. The essay this whole argument rests on, and the fastest way to sort any vendor claim you will hear this year.

John Ousterhout: A Philosophy of Software Design (2nd edition, 2021). Short, opinionated, and more useful the more code your organisation is generating that somebody will later have to understand.

David Parnas with Paul Clements: A Rational Design Process: How and Why to Fake It (1986). Free and brief. The gap between how the work actually happened and what should be written down is probably wider now than at any point since they wrote it.

Mark Burgess: Thinking in Promises: Designing Systems for Cooperation (O’Reilly, 2015). Read it for why coordination by voluntary promise outlasts coordination by command, whether the agents are people or machines.

Taiichi Ohno: Toyota Production System: Beyond Large-Scale Production (Productivity Press, 1988; first published in Japanese, 1978). The origin of overproduction as the worst waste, and the conscience of this whole argument on restraint under abundance.

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

A change lands in the repository on a Thursday afternoon. Nobody typed it. Two people skim it, the tests are green, and it is approved inside eleven minutes because nine more are queued behind it. Three years ago almost every step in that sequence was expensive, and the expense was doing quiet work nobody had to name.

Making is now the cheap part of building, and these articles have circled one consequence of that: an organisation can produce far more than it can understand, and it pays for what it cannot understand every day the system stays alive. The hypothesis running through this group is that building is what remains when making becomes free. Take out the typing and what is left is judgement, verification, coordination and restraint, which is to say everything that was ever essential about the job. This piece consolidates the group into nine methods. Each one is a diagnostic you can observe this week and a method you can act on.

They fall into three families of question that run through every phase of the series and change only which thinker holds them: Identity, Information and Interaction, which ask who the organisation takes itself to be, what it can see, and how its parts talk to each other. In building, Identity asks who or what is doing the making and what it can be held to. Information asks how precisely the thing being made is described and what the organisation will have to carry once it exists. Interaction asks how teams, components and agents relate while the work is happening. Burgess is the foundational thinker for the first, Ousterhout for the second, Collins for the third. Brooks sits underneath all three, and his distinction between essential and accidental complexity is the instrument that sorts what has changed from what has not.

Two questions sit under all nine. Can anyone still hold this, meaning understand it well enough to answer for it? And can anyone still change it? A system nobody holds is a liability with good test coverage. A system nobody can change is finished, whatever the roadmap says. Most organisations I talk to are losing both at once and measuring neither.

Identity: Who, or What, Builds

The Identity questions are about the builder before anything is built: what it can promise, who answers when the promise fails, and whether the capacity to do this work at all is being renewed or quietly spent. Burgess supplies the frame. An agent, human or otherwise, can only make promises about its own behaviour; an obligation imposed from outside is a statement about the wishes of the imposer, not a fact about what will happen. Autonomy is not something an organisation grants. It is a description of where control already sits, and the only useful question is what each part can promise from where it stands.

1. Ask what the builder can promise. An agent can promise to produce code that compiles, satisfies the tests it was given and matches the interface it was shown. It cannot promise that those were the right tests, that the interface was the right interface, or that the feature was worth having. Those promises were never available to it, because the information they depend on sits outside its locus of control. Most of the trouble organisations are having with agents comes from writing instructions as though they were obligations. Reality then declines to comply. The method: for each class of work you hand to an agent, write two lists, what it can promise and what it cannot, and make sure every item on the second list is held by a named person. The unheld items are your exposure. They are usually the interesting half of the work.

2. Ask who can answer for code nobody wrote. Authorship used to carry accountability without anyone arranging it; the person who wrote the thing understood it, and understanding is what answering for something requires. Generated code cuts that link cleanly, and the review queue is where the consequence shows up. Beck and Fowler made the small single-purpose change the cheapest discipline in software, because a reviewer can hold one. Given a nine-hundred-line diff with four purposes in it, the same reviewer approves the parts that look familiar. A signature on a change you could not explain is not a review. It is an admission. Willison calls the disciplined version of working this way vibe engineering, and what separates it from vibe coding is not the tooling; it is whether a named human is answerable for what shipped. The method: take the last ten merges and ask, for each, whether a named person could describe what it does and what it would break without reopening it. Count the ones where the answer is no. That count is the size of the unowned surface you are now running in production.

3. Ask where the next senior engineer comes from. Sennett spent a career on a single unfashionable claim: skill is made in the doing, and the hand teaches the head. Majors and Osmani describe where the doing now goes. An agent takes a feature from nothing to nearly finished, and nearly is the whole problem, because the last stretch is where the edge cases, the integration and the judgement live. You used to reach that last stretch by walking through the first seventy per cent by hand, badly, several times. We are automating the exact work our seniors were made of. The bill arrives in about five years, which is why nobody is looking at it. The method: name three tasks your juniors learned on three years ago, find out who or what does them now, and put a person deliberately back on at least one of them.

Information: What the Build Can See, and What It Will Carry

The Information questions are about description and cost. Ousterhout is the foundational thinker, and his definition is the one to hold: complexity is whatever makes a system hard to understand or modify. It is not a property of size and it does not arrive in a single bad decision. It accumulates in increments, each of which looked reasonable at the time. Nobody decided to make it this complicated. Free generation multiplies the increments and changes nothing else.

4. Count what you must know to change one thing. Take a small change made in the last fortnight, one that touched a few lines, and ask the engineer what they had to hold in their head to make it safely: which callers depended on the ordering, which flag was still read somewhere, which of the three similar-looking modules was the live one. The number of things that had to be known is the real cost of the change, and no dashboard reports it. Parnas gave the constructive form in 1972. A module earns its boundary by hiding a decision likely to change, so the change stops at the boundary instead of travelling through everything that touched it. The method: run the count on three recent small changes. Where one line required a dozen things to be known, you have found an interface that is not doing its job. Better now than next quarter.

5. Find the intent somewhere other than the code. When generation was expensive, the code was the artefact worth keeping and the specification was a promissory note that went to a wiki to die. That economics has inverted. The code is now the cheap, regenerable part, and the statement of what was wanted, and why, and what must never happen, is the thing that survives and the thing an agent can be held against. The method: pick a module written mostly by a model in the last three months and ask what it was for. If the only answer available is the code itself, you have a system that can be regenerated and not one that can be checked. There is nothing left to check it against.

6. Count what you removed. Ohno put overproduction first among the seven wastes, because it hides the other six and feels like productivity while it happens. Free making removes the only friction that ever restrained it. Generated abundance is not free to discard either; once a thing exists it must be understood, maintained, secured and supervised for the life of the system, so the invoice arrives late and then keeps arriving. Ask a team to add something and you get an estimate. Ask them to remove something and you get a silence. The method: count deletions and retired interfaces alongside the additions, quarter by quarter. A quarter with no removals is a quarter in which the carrying capacity of the team fell, whatever else the numbers said.

Interaction: How the Parts Relate While the Work Happens

The Interaction questions are about relations: between teams, between teams and the software they own, and between people and the things now generating on their behalf. Collins is the foundational thinker, and his account of ritual is the least tool-like idea in this whole group. An interaction ritual, in the account he set out in Interaction Ritual Chains in 2004, needs people assembled, attention on the same thing, a boundary saying who is in it, and something at stake. What comes out is emotional energy, solidarity, and standards the group defends when nobody is watching. A model can supply output all day. It cannot supply one unit of that.

7. Ask what your topology is able to produce. Conway observed that an organisation ships its communication structure, and Skelton and Pais turned the observation into a sizing rule: size the software to the team, one complicated domain per team, and when the software outgrows what the team can hold, split the software. Free generation broke that rule quietly. Software now outgrows its team in weeks, and the team finds out from the length of the review queue. The method: for each team, ask whether it could still explain everything it owns. Where the answer is no, split the software rather than adding another reviewer to the queue; adding reviewers treats the symptom and confirms the disease.

8. Ask what is at stake when the team is together. A stand-up where people report status has nothing at stake and generates nothing; the outcome was fixed before anyone spoke. A design session where two credible options are live, or a review where the change can genuinely be sent back, has stakes, and the group leaves it with more energy than it brought. Collins would say the second is a ritual and the first is an assembly, and only one of them makes standards that hold. The method: find the moment in your week where the team is in the same place, on the same thing, and the outcome is uncertain. Protect it in the calendar. If you cannot find one, that is your finding, and it explains more about your delivery than your tooling does.

9. Ask what your build system actually produces. Beer offered the most unforgiving instrument in the series: the purpose of a system is what it does. Point it at the pipeline. A pipeline that reliably produces green tests over code nobody has read has that as its purpose, whatever the policy says. A review process that approves any diff under a thousand lines the same day has throughput as its purpose. It calls that quality. The same instrument was aimed at the decision process earlier in this series; aimed at the build, it is sharper, because the build leaves evidence. The method: take the last ten changes, write down what the process claims about them (reviewed, verified, wanted) and what it demonstrably did, and treat the gap as the specification of your real build system.

Permanent Principles, Transformed Practice

The nine do two jobs, and separating them is worth a paragraph. Methods one, two, three and five test whether anyone still holds the thing: whether the promises are placed where they can be kept, whether a person can answer for what shipped, whether the skill to answer is still being made, and whether intent survives anywhere outside the generated artefact. Methods four, six and seven test whether anyone can still change it: what a change costs to understand, whether the system can be reduced as well as extended, and whether the shape of the organisation can produce the shape of the software. Methods eight and nine test both, over time, by looking at what the group’s habits and the pipeline actually produce rather than what they were designed to produce.

None of this is new engineering wisdom. That is the point, and it is Brooks’s point. Strip out the accidental complexity, the boilerplate and the glue and the first draft of the test scaffolding, and what remains exposed is the essential kind: what to build, where the boundaries fall, whether the result is any good. Modular decomposition, small reviewable changes, standard work, restraint under abundance, apprenticeship: none of these survived by luck. They are what the job turns out to have been all along, now that the part we mistook for the job has become cheap. The principles are permanent and their practice is transformed, and an organisation that reads the first half without the second will simply do 1998 faster.

Build comes from the Old English byldan, to construct a dwelling, from bold, a house. You build what you then have to live in. Every method here is a way of asking whether the thing being made this quarter is somewhere the organisation can still live in three years, and the reason the question has become urgent is that we have never before been able to build this much this fast without noticing what we were committing to. Carrying capacity, not generation rate, is in my view the binding constraint on every engineering organisation now. You cannot buy more of it with tools. You get it by keeping systems small enough to hold, intent explicit enough to check, and teams whose hands are still on the work.

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

An engineering leader I spoke to last year had all four DORA metrics green and three senior engineers telling her, separately and unprompted, that they felt slower than they had a year ago. She assumed the measurement was broken. It was not. Her instruments were reporting, accurately, that a large individual gain was being absorbed somewhere between the keystroke and production, and she had no framework that could show her where.

That gap is the subject of this article. Measurement will not rescue you: the disciplines here, deep modules and contraction and separated commits and sized domains, are judgement calls no number will make for you. What measurement can do is tell you whether you are lying to yourself. This year it is doing that with unusual clarity.

1. Four frameworks, each built to see what the last one missed

The lineage is short and worth knowing, because each step was a response to a real blind spot.

DORA came first, in the research Nicole Forsgren, Jez Humble and Gene Kim published as Accelerate in 2018. Four metrics: deployment frequency, lead time for changes, change failure rate, and time to restore service. Prescriptive, quantitative, evidence-backed, and narrow. They describe the pipeline from commit to production and almost nothing else.

SPACE followed in 2021, from Forsgren, Margaret-Anne Storey, Thomas Zimmermann and others, adding satisfaction, performance, activity, communication and efficiency. Broader, and descriptive rather than prescriptive, which meant teams had to invent their own measures and mostly gave up.

The DevEx framework arrived in 2023, from Abi Noda with Forsgren, Storey and Michaela Greiler, and narrowed onto three things developers actually experience: feedback loops, cognitive load, and flow state. Note the middle one. Ousterhout’s quantity and Skelton’s constraint finally appear in a measurement framework, and they are measured by asking people.

Then in December 2024, Noda and Laura Tacho at DX published the Core 4, built with the authors of the other three as advisors. Four dimensions, Speed, Effectiveness, Quality and Impact, each with one key metric and three secondary ones. It exists because executives kept asking what exactly to measure and getting “it depends”.

2. The most argued-about number in the set

The Core 4’s key Speed metric is diffs per engineer per week. Not lead time, not deployment frequency.

The reasoning is about translation rather than about truth. Take lead time to a finance director and you get asked why it matters; say each engineer ships four and a bit changes a week and the conversation continues. A real problem, solved.

It is also, under free generation, the single metric most exposed to inflation, and DX know it. Diffs per engineer goes up when an agent writes the diffs. It goes up when pull requests get split. It goes up when nobody is reading them. If you take one number from this framework into a board pack, take a different one.

What redeems the design is the part that looks like a weakness. The Core 4 deliberately mixes measured metrics with self-reported ones, including perceived rate of delivery and perceived software quality. Noda’s analogy is an endurance athlete’s rate of perceived exertion, and it is the right analogy. A runner whose heart rate says one thing and whose legs say another has learned something, and so has an organisation whose diffs are up thirty per cent and whose engineers report feeling slower.

The gap between measured and perceived is not an embarrassment to be reconciled. It is the finding.

3. What DORA found when it looked at AI

DORA’s 2025 report on AI-assisted development surveyed around five thousand people, and found adoption at roughly ninety per cent, which means the question has stopped being whether to adopt.

Two results matter. First, throughput reversed. In 2024 AI adoption was associated with reduced delivery throughput; in 2025 the association is positive. Teams have learned to ship faster with these tools. Second, and simultaneously, delivery instability continues to rise: more change failures, more rework, longer to recover. The systems downstream of generation have not adapted to the pace of generation, which is the argument these articles have been making since the Ohno article.

The third result is the one I keep coming back to. AI showed no measurable effect on friction or on burnout. Not an improvement, not a deterioration. DORA’s reading is that these are properties of the organisational system rather than of the individual workstation. Put less politely: your people are not exhausted because of their tools, and a better tool will not fix it.

DORA also did something braver than most research programmes manage. It threw away its own famous ranking. The Elite, High, Medium and Low performer ladder is gone, replaced by seven team profiles that combine delivery performance with friction, burnout and perceived value, running from foundational challenges through legacy bottleneck up to harmonious high-achievers. Roughly forty per cent of teams land in the top two, which is a more encouraging finding than the old ladder ever produced. A fifth metric appeared as well: rework rate.

And the headline conclusion is one sentence long. AI is an amplifier. It magnifies the strengths of high-performing organisations and the dysfunctions of struggling ones. Which makes AI adoption an organisational transformation rather than a tooling decision, and puts DORA and this series in the same position by different routes.

4. The capability that decides the sign

Buried in the AI Capabilities Model, which names seven practices that amplify AI’s benefit, is a result that should change how you sequence your investment.

Without a user-centric focus, AI adoption is associated with a negative impact on team performance. With it, the association turns positive. The same tools, the same money, opposite signs, depending on whether the organisation is oriented towards the people using its software.

That is Ohno’s pull signal arriving in survey data. An organisation that generates because generation is available produces waste at speed. An organisation that generates in response to a real demand produces value at speed. The tool does not know which one it is in.

5. What none of it measures

Almost nothing these articles say matters is measured at all, which is why I opened by saying measurement will not rescue you.

Faros telemetry from 2026, across roughly twenty-two thousand developers, gets closest. Developers using AI handle 67.4 per cent more pull request contexts a day and 17.7 per cent more task contexts, both sharply up on the prior year. Work restarts are up 13.8 per cent. Twenty-six per cent more in-progress tasks sit for a week or more with no activity at all. That is the nearest thing anyone has published to measuring cognitive load directly, and it is measuring context-switching as a proxy for it.

Everything else these articles say matters remains unmeasured. Coupling, which Beck via Constantine identifies as the actual cost driver of your software. Contractability, which Parnas makes the test of whether the system can still be changed. The accumulated ritual chain of a team, which Collins says determines whether it can take on something hard. Each of these is a stock rather than a flow, and every dashboard ever built measures flows.

So use the frameworks for what they are good at, which is detecting the gap between what you believe and what is happening. Then accept that the accumulations require somebody to go and look.

The next two articles close the Building group: the nine methods for building, and an honest account of why organisations cannot do it.

Placement: Building, the third of the series' four phases, after Learning and Deciding and before Transforming. Measurement cuts across all three levers rather than sitting on one, because a metric is at once a description (Information), a claim about who is accountable (Identity) and a signal that changes how the parts relate (Interaction). The Build Is the Test opens this stretch of the argument.

(An Organisational Prompt is something you can do now....)

Organisational Prompt

Put the measured number and the felt number on the same page.

Take one team and one quarter. On the left, the measured column: diffs per engineer, lead time, change failure rate, rework rate. On the right, a perceived column, gathered by asking every member of the team two questions and writing down what they say. Has your rate of delivery gone up or down this quarter? Has the quality of what you ship gone up or down?

Do not average the two columns and do not attempt to reconcile them, because reconciling is how organisations destroy the only useful thing on the page. Put them side by side and look at the sign of each.

Where measured and perceived agree, you can trust the instrument. Where they diverge, you have found the thing your dashboard cannot see, and the divergence tells you which direction to look. Diffs up and perceived delivery down means the gain is being absorbed downstream. Quality measures flat and perceived quality down usually means people are shipping code they could not walk you through line by line.

Then add three numbers nobody asks for. The proportion of your codebase changed in the last year that no current employee could explain. The age of your oldest open branch. The time it would take to remove one dead feature. None of them is on a dashboard. All three are knowable in a fortnight, and together they will tell you more about next year than the four keys will.

Further Reading

DORA’s publications index carries the 2025 State of AI-assisted Software Development report and the AI Capabilities Model companion, both free. Read the seven team profiles first and find yours before you read anything else in it.

Abi Noda: Introducing the DX Core 4 (December 2024), with the fuller research paper behind it. Short, prescriptive, and unusually candid for a vendor about what the earlier frameworks could not do and why.

Faros AI’s analysis of the DORA 2025 report sets the survey findings against its own telemetry across tens of thousands of developers. The context-switching numbers are the most useful data in the field right now.

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

An icosahedron has twelve vertices, thirty edges and twenty faces. In 1994, in his late sixties and at the end of a career spent telling large organisations that their information architecture was killing them, Stafford Beer published a book proposing that you should organise a meeting in that shape.

Put a topic at each vertex. Put a person on each edge. Because every edge touches exactly two vertices, each of the thirty people is now a member of two topic teams, and each team has five people. Give everybody a critic’s role on two further topics. There is no chair, no top and no bottom, and no agenda handed down, because the twelve topics are generated by the thirty people in the first session. Beer called the group an infoset, called the method Team Syntegrity, and thought it was the most important thing he ever built.

Deciding used Beer for the Viable System Model and for the observation that the purpose of a system is what it does. This article takes the part that got left out. Building’s Interaction lever asks how the parts of a system relate, and syntegrity is the only serious attempt anyone has made to design that relating from first principles.

1. The problem it solves

Beer’s whole career rests on one law, borrowed from Ross Ashby and published in 1956. Only variety can absorb variety. A regulator has to be able to produce at least as many distinct responses as the situation can produce distinct states, and where it cannot, the difference does not vanish. It shows up as the situation doing something you had no way of handling.

Hierarchies deal with this by attenuating. Information is summarised at every level on its way up, because the layer above cannot hold what the layer below knows. Attenuation is not an abuse of the structure; it is the structure. And each summary is a decision about what to leave out, made by somebody with an interest in how the result reads.

The Viable System Model answers this with channels and with the audit path, which lets the centre look directly at operations rather than through the reporting line. Syntegrity answers a different question. Suppose you have thirty people who between them know what needs to be known, and no time to pass it through a hierarchy. What shape should the conversation be?

2. The shape, and why it is that shape

The word fuses synergy with Buckminster Fuller’s tensegrity, a structure that holds itself together through distributed tension rather than through a strong centre. That is the design intent stated in the name.

Two properties of the icosahedron are doing the work. The first is that every person sits at most two nodes from every other, so anything said anywhere can reach anybody in two hops without climbing anything. The second is that no position is distinguishable from any other. Rotate the solid and it looks the same, which means the protocol cannot be gamed by seniority, because seniority has nowhere to sit.

The meeting runs three iterations over about three days. In each, a person is a member of one team, a critic of another, and an observer somewhere else, and the roles rotate. Beer’s term for what happens next is reverberation: an idea raised in the green team on the first morning comes back to you on the second afternoon, altered by two teams you were not in. He analysed the propagation mathematically, with eigenvalues, which tells you something about how seriously he meant it. Practitioners who run these report something in the region of ninety per cent of the information ending up genuinely shared across all thirty.

Since 1994 it has been used on the future of the City of London, at Union Bank of Switzerland, by the Colombian Ministry of the Environment, and by an Israeli and Palestinian group discussing West Bank settlements. This is not a thought experiment.

3. What the last two articles would say about it

Collins would notice immediately that syntegrity satisfies all four of his ritual ingredients by construction. Bodily co-presence over three days. A hard boundary, since the infoset is exactly thirty named people. A mutual focus, because each session addresses one statement. And a shared mood, which three days of that intensity guarantees whether or not anyone intended it.

That is probably the honest explanation of why participants describe syntegrations the way they do. Beer built a machine for manufacturing emotional energy and described it in the language of cybernetics, which is the language he had.

Skelton and Pais would notice something else. Syntegrity is the interaction mode chosen deliberately rather than left to drift, at the largest scale anyone has attempted. Their three modes are a vocabulary for pairs of teams. Syntegrity is a protocol for a whole population, making the same argument at a different scale: the shape of the interaction is a design decision rather than a matter of taste.

4. Attenuation just became free

The reason to care about a thirty-year-old protocol is that its central warning just got much cheaper to ignore.

Beer’s warning was always about attenuation. Every point at which information is summarised on its way to a decision is a point at which variety is destroyed, and the destruction is invisible from above, because what arrives looks complete. The report reads well. It reads well because somebody made it read well.

Summarisation is now free, instantaneous and fluent. Every layer of your organisation can attenuate at machine speed, and the output is more polished than the human version was, which makes it harder rather than easier to notice what has gone. An organisation can now attenuate everything and amplify nothing, and the dashboards will look better than they have ever looked.

Lorin Hochstein made the specific version of this argument about incident reports, and his distinction generalises. Using a model to gather the material is fine. Using it to produce the account removes the act in which somebody had to decide what mattered. Beer would put it more bluntly: you have automated the step where variety is lost and left the step where variety is generated to look after itself.

The syntegrity answer is not to ban summarisation, which is impossible and would be stupid. It is to notice what compensates for the losses. A channel with real capacity, in which thirty people hear each other directly, is the only thing that ever has, and such channels are getting rarer while cheap imitations of them get easier to build.

5. Where the argument runs out

Three honest objections, and the third is the serious one.

It is expensive. Thirty people for three days is a real cost, most organisations will do it once, and once is not an operating model. Beer designed an event, not a way of running a company.

It gets strange. The later formalism drifts into chakras, enneagrams and elaborate colour schemes, and this loses him readers who would otherwise take the geometry seriously. Read the structure and skip the mysticism; the structure stands without it.

The evidence is thin, in the way most organisational evidence is thin, and it is worth saying so plainly rather than leaning on the numbers. The ninety per cent figure comes from practitioners rather than from controlled study. And the people reporting how much the experience changed them had just spent three days in a highly charged group, which Collins would tell you affects the reporting.

None of that touches the underlying claim, which is the one to carry away. The shape of your conversations determines what your organisation can know, and almost nobody designs that shape on purpose.

The next article takes this to measurement, and to what the numbers can and cannot see.

Where this falls: Building, the third of four phases; Learning, Deciding, Building, Transforming. Beer is carried forward from Deciding, where he held Interaction, and POSIWID becomes the unforgiving test: a system in production is judged by what it does, not by what its specification claimed for it. The Build Is the Test opens this stretch of the argument.

(An Organisational Prompt is something you can do now....)

Organisational Prompt

Count the attenuations between the work and the decision.

Take one significant decision you made in the last quarter. Trace backwards to where the information came from, and count the number of times it was summarised on the way. Not the number of layers on the org chart. The number of separate acts of summarising.

For each one, write down two things: who did the summarising, and what they had at stake in how it read. Then mark which of them were done by a person and which by a model, and note that the second kind leaves no trace at all.

Most organisations find four to six, and I have seen nine. Beer’s guarantee inside a syntegration is that nobody is more than two hops from anybody, and the gap between two and six is roughly the amount of variety that never reached you.

Then pick the single attenuation you would most like to remove, and remove it. Sit in the room where the work happens, once, for an afternoon, with nobody presenting. What you learn will be uncomfortable and specific, and it is the thing your reporting line was built to prevent you from learning.

Further Reading

Stafford Beer wrote Beyond Dispute: The Invention of Team Syntegrity in 1994, for Wiley, and it is odd, brilliant and uneven in roughly equal parts. Read the protocol chapters and treat the metaphysics as optional.

Joe Truss and Allenna Leonard: The Coherent Architecture of Team Syntegrity. Free, and the clearest technical account of how the thing is actually run, including the smaller variants for groups that cannot field thirty people.

The International Society for the Systems Sciences primer on Beer is a short free summary of the protocol and of where his work sits, which is a reasonable place to start if the book looks daunting.

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.

There is a sentence in Team Topologies that most readers skate past on the way to the diagrams, and it is the one that matters most this year. Size the software to the team, never the reverse. One complicated domain per team. When the software outgrows what the team can hold, split the software.

Read that against what has happened to your codebase since 2024. The software grew. The team did not. Nobody made a decision and nobody proposed a change to the operating model. The rule broke from a direction Matthew Skelton and Manuel Pais had no reason to anticipate when they published the first edition in 2019.

Conway appeared in this series during Deciding, for the claim that an organisation ships its communication structure. This article picks up what two practitioners built on top of that observation, and what happens to their construction when a large share of the code is written by something that has no place on the org chart.

1. From an observation to a manoeuvre

Conway’s 1968 paper made a descriptive claim. A system’s design copies the communication structure of the organisation that built it, because the interfaces people can negotiate are the interfaces they have conversations across.

Skelton and Pais turned the description into an instruction, and the move is the reverse Conway manoeuvre. If your architecture will end up mirroring your team structure whatever you do, then decide the architecture you want and build the team structure that produces it. Stop drawing target architectures and start drawing target org charts, because only one of those two things is under your control.

Their vocabulary is four team types and three interaction modes, and its value is that it is small enough to hold in a meeting. Stream-aligned teams own a flow of change to some part of the business. Platform teams reduce what stream-aligned teams have to know. Enabling teams raise capability and then leave. Complicated-subsystem teams exist where genuine specialism is unavoidable. The modes are collaboration, self-service, and facilitating, and the discipline is that a pair of teams should be in exactly one mode, deliberately chosen, and should change it deliberately rather than by drift.

The second edition arrived in September 2025 with new case studies. The framework has become the default vocabulary for engineering org design in large organisations, which is worth noting mainly because default vocabularies stop being examined.

2. The constraint underneath all of it

The four types are not the theory. Cognitive load is the theory, and everything else is a consequence.

The idea comes from John Sweller, who distinguished the intrinsic difficulty of the material, the extraneous difficulty added by how it is presented, and the germane effort that actually builds understanding. Their application is a constraint on org design: a team can own only as much domain as it can hold in its collective head. Draw the boundary past that line and you have not created ownership. You have created a name on a wiki page.

This is where the sizing rule comes from, and it is worth stating in the direction they state it. The team is the fixed thing. The software is the variable. When a domain grows past what a team can carry, the answer is to split the software, not to add people to the team, because adding people raises the coordination cost faster than it raises capacity.

Notice that Ousterhout named the same quantity two articles ago from inside the code. For him it is what a developer must hold to make a change safely. Skelton and Pais measure the same quantity at the scale of a team, and use it to decide where boundaries go. Two disciplines, one constraint, arrived at independently.

3. Which direction the rule broke

Now put generation into the model, and notice that it attacks the sizing rule from the side nobody defends.

The intended failure mode is a domain that grows because the business asked for more. That is visible, budgeted and argued about in a room. What is happening instead is that the volume of code inside an unchanged domain boundary has risen sharply while the team roster has not moved. The domain looks the same on the diagram. The amount that must be held to change it safely has gone up, and quite a lot of it was written by an agent, at speed, by somebody who read the diff quickly.

Three responses are available, and only three. Split the software, which is the answer the book gives. Narrow the domain, which means giving something away and is therefore politically expensive. Or accept shallow ownership, in which the team’s name stays on the service and its actual grip on the service quietly weakens.

Almost every organisation is taking the third option, and almost none has decided to.

There is also an uncomfortable dependency between this remedy and the last article but one. Splitting the software is only available if there is something to split along, which means the uses relation has to be a hierarchy rather than a tangle. Parnas’s loops and Skelton’s sizing rule meet here in an unpleasant way. Where generated code has accumulated dependencies nobody drew, the Team Topologies remedy is unavailable in exactly the situations that call for it.

4. The org chart has non-human nodes on it now

Conway’s law does not care whether the communicating parties are people.

Suppose a stream-aligned team’s real interface to a domain runs through an agent, and the agent’s context is a directory of markdown files somebody wrote in March. That arrangement is now part of the communication structure, and the architecture will come to reflect it. Whatever the agent finds easy to attach to is where things will attach. Whatever the agent has no visibility of will drift.

Skelton has been direct about this in his 2026 talks, and his framing is bounded agency. Organisations that had already learned to give humans clear, limited, well-supported autonomy turn out to be well placed to give agents the same thing. The problem has the same shape either way: what may this party do, what may it reach, and through which interface. He is explicit that agents should not have unbounded access to data and resources, which is the platform argument restated for a new kind of consumer.

Which makes the platform question sharper than the framework originally posed it. A platform justifies itself by the cognitive load it removes from stream-aligned teams and by nothing else. If your agents reach around the platform and touch whatever they like, you do not have a platform. You have a set of conventions and a strong hope, and Skelton’s vending-machine test applies unchanged: capabilities should be available through a clear self-service interface, not through a ticket and a favour.

5. What is being claimed, and by whom

A good deal is currently being asserted about how the four types change under AI, and it is worth separating the sources.

The community claims run roughly like this. Stream-aligned teams shrink towards three to five people. Complicated-subsystem teams dissolve because agents absorb specialist implementation. Enabling teams pivot towards context engineering and tool evaluation, and platform becomes the highest-return investment in the organisation. Every one of these is plausible. None of them is yet supported by evidence I would put in front of a board, and they come from commentators rather than from Skelton and Pais, who have been noticeably more careful.

The authors’ own claim is narrower and stronger. The principles were never about humans in particular. They were about any system trying to build and maintain software under a limit on what can be held at once, and that limit does not disappear because some of the holding is being done by a model. Cognitive load is indifferent to what kind of mind is being overwhelmed.

So the useful question for a leader is not which team type is obsolete. It is whether your teams still own what they are named after, and the sizing rule gives you a way to find out that does not depend on anybody’s opinion about the future.

The next articles move to measurement, which is where all of this either shows up in numbers or does not.

Placement: Building, the third of the series' four phases; Learning, Deciding, Building, Transforming. Conway returns here on the Interaction lever, which asks how teams, components and agents relate. In Deciding his law was a warning about what a structure would produce; in Building it is a report on what it did. The Build Is the Test opens this stretch of the argument.

(An Organisational Prompt is something you can do now....)

Organisational Prompt

Ask a team to draw its own domain, and see whether it can.

Take one stream-aligned team. Give them a whiteboard and twenty minutes with no manager present, and ask them to draw everything they own: services, jobs, data stores, integrations, the lot. Then ask each person privately to mark the parts they could change safely on their own next week.

Two numbers come out. The size of the domain, and the fraction of it any individual can actually act on. The gap between them is your cognitive load overrun, and it is the number Team Topologies says should govern your org design.

Then ask when each part was last substantially changed, and by whom. Suppose a meaningful share of the domain was generated in the last year and nobody who works on it marked it as safe to change. That team no longer owns the service in any sense that will hold up at two in the morning. It owns the name.

Three responses exist and you must pick one out loud: split the software, narrow the domain, or say plainly that you are accepting shallow ownership and why. The third is a legitimate choice for a short period. It is not a legitimate accident.

Further Reading

Matthew Skelton and Manuel Pais: Team Topologies (second edition, September 2025). Short, practical, and unusually honest about its own limits, which is rarer in this genre than it should be. The site carries a readiness survey and the new case studies.

Matthew Skelton: Team Topologies as the infrastructure for agency with humans and AI (2026). The authors’ current position on agents, and considerably more restrained than most of what is being written around them. The slides are worth a scroll on their own.

Melvin Conway: How Do Committees Invent? (1968). Free on Conway’s own site, six pages, and it says more than the one line everybody quotes from it.

I write about the industry and its approach in general. None of the opinions or examples in my articles necessarily relate to present or past employers. I draw on conversations with many practitioners and all views are my own.