There is a version of AI adoption that is going well in a lot of organisations right now. Teams have access to tools. People are being asked to identify where AI could help them. Use cases are being logged, prioritised, piloted. There is genuine engagement, genuine participation, and in many cases, genuine improvement.
This is worth acknowledging before making any kind of critical observation about it, because the criticism is not that it is wrong. It is that it is incomplete.
Thanks for reading The Operating Model Dispatch! Subscribe for free to receive new posts and support my work.
The pattern that is becoming clear, and that borrows from something much older than AI, is this: when each team optimises its own use of AI within its own boundaries, you get incremental improvement within those boundaries. What you do not get is the larger opportunity that only becomes visible when someone looks across all of them at once.
This is not a new problem. It is the same problem that process optimisation has produced for decades. A Finance team that improves how it produces its outputs will get better at producing those outputs. But Finance does not work in isolation. The information Finance produces is consumed by other teams, who make decisions based on it, who need it in particular forms and at particular times to do their own work effectively. And Finance is downstream of its own inputs: the systems, processes, and data that feed into it, which may themselves be unoptimised, inaccurate, or structured in ways that create invisible friction every time Finance has to handle them.
Optimise Finance in isolation and you get a better Finance function. You do not get a better system. You get a more efficient node in a network that nobody has looked at as a whole.
AI does not change this dynamic. It amplifies it.
When AI adoption is structured around team-level participation, you will get teams that are genuinely better at what they are already doing. That is real value and it should not be dismissed. But if Sales optimises how it uses AI, and Marketing does the same, and Operations does, and Finance does, and nobody looks across all of it, the organisation will still be leaving the most significant opportunities on the table. Not because the teams are doing anything wrong, but because the most valuable opportunities are not visible from inside any single team’s boundary.
The opportunities that tend to matter most are in the connections: where information moves between teams and something is lost in the translation; where a decision in one part of the organisation creates unnecessary work in another; where the same problem is being solved three times over because no one has ever seen all three solutions in the same room. These are not visible from inside Finance, or inside Sales, or inside any function that is genuinely and conscientiously optimising its own piece of the whole.
What surfaces them is a different scope of question. Not “how can your team use AI more effectively?” but “how does value actually move through this organisation, and where does AI change what is possible along that path?”
This does not require a giant transformation programme. It does not require months of consultants mapping every process before anyone is allowed to do anything. What it requires is a deliberate moment of looking at the whole before consolidating around the parts. It can be relatively lightweight. It is more about perspective than scale.
The reason most organisations skip this step is understandable. It is genuinely easier to engage people within their own context. The questions are more manageable, the scope is bounded, the feedback loops are faster. Team-level co-design feels more human and more practical, which it often is, on its own terms. But practical at the team level and optimal at the organisational level are not the same thing, and the gap between them tends to widen as AI adoption matures.
What tends to happen is that organisations that have done good team-level work find themselves, twelve or eighteen months in, with a set of well-optimised functions that are harder to integrate than expected. The boundaries that made adoption manageable become the constraints that make the next stage difficult. The cross-functional work that would have been easier to build in from the start now has to be retrofitted into structures and habits that have already formed around narrower assumptions.
The point is not to delay the team-level work. It is to make sure someone is holding the wider view at the same time. Not instead of the bottom-up engagement, but alongside it. The two are not in competition. The whole-system perspective makes the team-level work better, because it helps each team understand what it is optimising for and what its optimisation is in service of.
Co-design done well is one of the most important things an organisation can do in the current moment. The question is whether the scope of the conversation matches the scale of the opportunity.