Problems with the Traditional Change Theory We All Inherited
Why Kotter’s eight-step model describes transformation as leaders imagine it, not as organisations experience it.
We have been talking around him for weeks. Every time this series has critiqued linear change management, top-down planning, or the assumption that transformation can be designed and controlled from above, it has been describing the world that John Kotter codified. It is time to address him directly.
Kotter’s 8-Step Model for Leading Change, published in 1996 and refined in Accelerate in 2014, remains the default operating system for enterprise transformation. If your AI transformation programme has a phased roadmap with “create urgency” in Phase 1 and “anchor in the culture” in Phase 8, it is running Kotter. The model is not wrong. It is incomplete. And its incompleteness has a specific shape: Kotter’s framework describes the world as it looks from the executive floor. It is the view from the balcony with the stairs removed. CEOs do not drive change in organisations. They frame it. Middle management enables it. The people on the front line actually enact it. Every thinker in this series, from radically different starting points, converges on this insight.
1. What Kotter Actually Says
Kotter’s framework deserves a fair hearing. It organises change into three phases: creating a climate for change (establish urgency, form a guiding coalition, create a vision), engaging the organisation (communicate the vision, empower action, generate short-term wins), and implementing change (consolidate gains, anchor new approaches in culture). He identifies genuine failure modes: complacency, weak coalitions, under-communication, declaring victory too soon. Anyone who has watched a transformation stall because leadership lost interest will recognise the patterns.
The question is not whether these patterns are real. The question is whether the framework that explains them is adequate. Kotter’s contributions should not be discarded: his observation that organisations are over-managed and under-led is validated by every thinker in this series. His insistence on political work, building coalitions and removing obstacles, is pragmatically sound. His insight that culture follows practice, not the reverse, is confirmed by Giddens, Bourdieu, and Weick. What must be examined is the set of assumptions the model rests on, because those assumptions determine what it can and cannot see.
2. The View from Above: Why the Model Cannot See the Organisation
The first limitation is perspectival. Kotter’s model assumes change is initiated at the top, designed at the top, and cascaded through the organisation. Everyone else is an object of change: recipients of the vision, targets of the communication plan. Stacey rejects this entirely. In his theory of Complex Responsive Processes, the CEO publishing an AI strategy is not programming a machine. She is making a gesture from within a web of relationships, and the responses will determine what actually happens, not the plan. The teams quietly using AI to solve real problems outside the governance framework are not insubordinate. They are the emergent strategy trying to tell you where the value is.
The model is also sequential. Steps cannot be skipped; each phase builds on the previous. This assumes a level of predictability that Snowden would place in the complicated domain, suitable for problems with known solutions, and inadequate for adaptive challenges where the destination is discovered through the journey. Weick inverts the sequence entirely. Kotter says: create the vision, then act. Weick says: act first, then discover the vision retrospectively. You cannot know what your AI transformation vision should be until you have experimented enough to discover what AI makes possible in your specific context.
Bateson’s levels framework reveals the depth of the problem. Kotter’s model operates at Learning I: it assumes the frame is correct (the organisation knows what it needs to become) and the task is execution within that frame. But AI transformation is a Learning II challenge: the frame itself must change. The organisation does not yet know what it needs to become, and the eight steps, however well executed, cannot produce the frame-questioning that Learning II requires. They can produce excellent execution of the wrong destination.
Kahneman explains why the sequential model feels so compelling despite its inadequacy. The planning fallacy focuses attention on the specifics of the plan rather than the base rates of similar projects. WYSIATI means the coherent story of eight sequential steps suppresses awareness of what the model leaves out. And narrative fallacy means successful transformations are retrospectively reconstructed as having followed a logical sequence, even when the actual path was recursive and full of surprises. Kotter’s model is System 1 candy: coherent, simple, compelling. System 2, if activated, would note that the base rate of transformation success has not improved since the model was published.
3. Why Urgency Produces Compliance, Not Learning
Kotter’s Step 1, “establish a sense of urgency,” is the most widely applied and the most consistently counterproductive element. The logic is seductive: people will not change unless they believe the status quo is unacceptable. Therefore, manufacture urgency, surface threats, create the burning platform.
Seligman would recognise this as a formula for producing learned helplessness. People who have been through previous change programmes have learned that the urgency narrative is a management technique. Responding to it with genuine emotional commitment is a mistake experienced professionals do not make twice. Heifetz provides the more precise diagnosis: urgency works for technical problems (the server is down, the regulator is at the door) but produces fight-or-flight responses for adaptive challenges, narrowing attention, triggering defensive routines, and preventing the exploratory learning that adaptation requires. The burning platform does not produce transformation. It produces compliance disguised as transformation.
Dweck adds the mechanism. Under threat, people with a fixed mindset retreat to performance goals that protect their sense of competence. Urgency framed as threat (”if we do not adopt AI, we will be left behind”) activates this defensive posture. The same urgency framed as opportunity (”AI creates possibilities we have not imagined, and your domain expertise is what makes those possibilities real”) invites the learning goals that growth mindset enables. Kotter’s model does not distinguish between these framings. The difference between them determines whether the organisation learns or merely complies.
Step 4, “communicate the vision,” rests on the same information-deficit assumption. If people understood, they would support the change. Argyris would point out that the obstacle is rarely information deficit. It is the gap between espoused theory and theory-in-use. The leader who communicates “we embrace AI innovation” while the governance framework requires six approvals before anyone can use a tool is sending two messages simultaneously: one in words and one in structure. People are extraordinarily skilled at reading the structural message and ignoring the verbal one.
4. Why Culture Changes Last, but Not the Way Kotter Thinks
Kotter’s insight that culture changes last, not first, is genuinely valuable. Culture follows practice; it does not precede it. On this, Kotter and the series thinkers agree.
But Kotter treats culture as sediment deposited by successful practice. Giddens shows it is far more actively reproduced: structure operates through three dimensions simultaneously, signification, domination, and legitimation, and all three must change together. If you change the practices but leave the power structures unchanged, if the people who control budgets and promotions still reward the old way of working, the new practices will not survive. If the legitimation remains unchanged, if “real engineering” still means writing code, the new practices will be experienced as illegitimate regardless of demonstrated success.
Bourdieu goes deeper. Culture is not just structural; it is embodied. The habitus was formed through years of professional socialisation. The senior developer’s body knows how to write code. Their reflexes are calibrated to a world where coding is the work. New practices bounce off the habitus until the embodied dispositions themselves shift, and habitus change requires sustained immersion in new conditions, not a demonstration project followed by a communication campaign. Weber would observe that the eight-step model uses bureaucratic methods to produce change within a bureaucratic system. This can reconfigure the iron cage. It cannot open it.
5. What Must Be Added
The synthesis is not a new model to replace Kotter’s. It is a set of corrections that make his practical instincts more honest.
From Argyris: surface the defensive routines that make vision communication performative. The change programme that cannot make the undiscussable discussable will produce compliance, not learning.
From Weick: place action before understanding. Small wins that accumulate into pattern change through sensemaking, not through cascaded implementation.
From Stacey: participate in the interaction, do not stand above it. Plans are gestures, not blueprints. Include the people already doing the adaptive work informally. From Heifetz: distinguish technical from adaptive. Give the adaptive work back to the people who must live with its consequences.
From Giddens and Bourdieu: change all three dimensions of structure simultaneously, and understand that the habitus changes through prolonged practice, not through communication.
From Bateson: recognise that execution within the wrong frame is still the wrong frame. The eight steps are Learning I tools. AI transformation is a Learning II challenge.
From Stafford Beer: if the programme is producing programme artefacts but not changed practice, then producing artefacts is its purpose. Redesign the information architecture, not the communication plan.
Lead, yes, but from within the interaction, not above it. Communicate, yes, but only after ensuring the structures say the same thing as the words. Generate wins, yes, but measure the learning, not just the outcomes. And anchor in the culture, yes, but understand that culture is the practical consciousness that is either changing, moment by moment, in the daily practice of the organisation, or it is not.
(An Organisational Prompt is something you can do now....)
Organisational Prompt
Kotter’s model assumes the leader designs the change. The thinkers in this series argue that the leader participates in change that is already happening, whether the leader knows it or not.
Identify the most recent executive announcement about your AI transformation. Now find three unplanned consequences of that announcement: a team that started doing something different, a workaround that emerged, a conversation that shifted. Not the planned outcomes, the metrics, the dashboards. The unplanned consequences. These are not noise. They are the emergent strategy responding to the leader’s gesture. They are what the plan actually produced, as opposed to what it intended to produce. If you cannot find them, you are not close enough to the work.
Further Reading
John Kotter, Leading Change (revised edition, 2012). The original statement of the 8-Step Model. Read it for the practical observations about why change fails. Read the thinkers in this series for what it leaves out.
John Kotter, Accelerate (2014). Kotter’s refinement into a “dual operating system” running the eight steps continuously through a network alongside the hierarchy. Worth reading alongside Stacey for what the “network” model still assumes about control.
Ronald Heifetz, Alexander Grashow, and Marty Linsky, The Practice of Adaptive Leadership (2009). The essential complement. Where Kotter tells you how to run a change programme, Heifetz tells you what to do when the programme encounters an adaptive challenge it was not designed to handle.
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.