The Three Speeds of AI Transformation
Part 1 of 2 in the Transformation Stack series
TL;DR
AI transformation involves three rates of change operating simultaneously
Most leaders are treating a synchronisation problem as a rollout problem and that misreading explains most of the friction they are experiencing.
The friction is almost always located between the people speed and the organisation speed
The most useful diagnostic question is not “what are we doing wrong?” It is “which of the three speeds is acting as the constraint?”
There is a pattern that appears with remarkable consistency across organisations two or three years into serious AI adoption. Despite genuine investment, working technology, and real productivity gains among regular users, the organisation as a whole does not feel meaningfully different. The capability that was supposed to accumulate is not showing up where it was expected.
A CEO at a retail company described it precisely last year. Eighteen months of AI pilots, genuine enthusiasm at the start, a steady stream of demos. Zero systematic capability. He understood the headwinds, the technology moves fast, people have understandable concerns, bottom-up discovery takes time. But he kept asking his team the same question: when do we move from experimenting to actually running the business differently? Nobody had a satisfying answer.
His situation is common. The reason most leaders struggle to answer that question is not that they are moving too slowly, it is that they are misreading the nature of the problem.
AI transformation is a synchronisation problem. Most organisations are treating it as a rollout problem. That misreading explains most of the friction.
Three Rates of Change, Moving at Once
What makes AI transformation structurally different from most previous technology transitions is that it involves three distinct rates of change, each responding to a different kind of leadership attention.
Technology evolves in days. New capabilities arrive continuously, often before the previous wave has been fully absorbed. By the time a tool has been approved and rolled out, a more capable alternative may already exist. An organisation’s relationship to this speed is one of selection and pacing: deciding which capabilities to adopt, at what moment, and with what degree of organisational preparation.
People adapt in weeks. Building genuine AI fluency across a workforce takes time, repeated exposure, and conditions that most training programmes do not create. Organisations moving fastest here have embedded AI into the texture of daily work, rather than treating fluency as a skill to be acquired separately and then applied. This speed responds to deliberate investment, but it does not respond to urgency alone.
Organisations change in months, if not years. Processes, decision structures, role definitions, performance frameworks, and cultural expectations do not move at the speed of a technology deployment. An organisation can adopt a new AI capability in weeks and still be running the same underlying operating logic it ran before. In many cases, that is exactly what is happening.
Traditional change management implicitly assumed these three moved at roughly the same pace. That assumption shaped most of the transformation playbooks still in use today. AI breaks it, which is why those playbooks keep producing results that feel incomplete.
Where the Friction Actually Lives
The friction that leaders notice - the sense that adoption is happening without transformation - is almost always located in the gap between the second and third speeds. People are learning but the organisation is not yet changing around what they are learning.
Picture the scene. A compliance team finishes an AI rollout and cuts regulatory report drafting time by over 70%. The work is faster, measurably so, and the team knows it. Then they look at their performance reviews. The targets have not changed. The metrics still measure what they measured before. The time they saved has nowhere to go within the structure they are evaluated by. Three months later, usage has drifted back toward old habits - not because the tool failed, but because the organisation never changed around it.
The technology had moved. The people had moved. The organisation had not, and that gap absorbed all the gain.
This pattern repeats reliably across sectors. Its symptoms are recognisable:
Employees with genuine AI fluency find their new capabilities constrained by processes designed before those capabilities existed.
Teams with high-value applications cannot scale them because the approval structures or data infrastructure are not yet in place.
Leaders who have invested in skills programmes find those skills producing activity rather than change.
None of these are technology problems. They are organisational design problems that AI has made visible.
The Diagnostic That Changes What You Do Next
The most common error is treating the three speeds as a single problem. A cultural challenge managed as a technology deployment, or an organisational design problem addressed through individual training, will produce exactly the frustration that characterises most stalled AI programmes - not because the effort is wrong, but because it is aimed at the wrong layer.
The practical implication is a question of diagnosis before action. When a significant AI initiative is not producing the results expected, the most useful first question is not “what are we doing wrong?” It is “which speed is acting as the constraint?”
That diagnostic question changes what kind of leadership attention is needed:
If the technology layer is the constraint, the answer is selection and pacing - choosing the right tools at the right moment, not deploying everything at once.
If the people layer is the constraint, the answer is structured capability building embedded in actual work, not standalone training delivered and forgotten.
If the organisation layer is the constraint - and this is by far the most common finding - the answer is structural: redesigning processes, redefining roles, updating the performance frameworks that determine what people are actually rewarded for doing.
Misidentifying the constraint is one of the most expensive mistakes in transformation. It redirects budget and energy toward a layer that is not the source of the problem, while the actual constraint continues to hold everything back.
For most organisations engaged in AI transformation today, the honest answer to that diagnostic points to the third layer. The technology is capable enough. The people, given the right conditions, can develop. It is the organisation itself - its structures, its habits, its unexamined assumptions about how work should be organised - that is moving slowest, and asking the least of itself.
In your most important AI initiative right now, which of the three speeds is acting as the real constraint - and is your leadership attention focused accordingly?
This framework is part of a broader playbook I’m developing for leaders navigating AI, the AI Strategy Playbook: From Zero to 100. If you’d like early access to the book and related materials as they become available, you can join the waitlist here.