There is a fantasy at the centre of most transformation programmes. It is rarely stated directly. But if you have sat in enough boardrooms, you have heard it between the lines. It goes roughly: if only we could run this without so many people in the way.
Let us be honest about something most strategic planning documents are not honest about.
People are complicated. They have opinions that were not requested. They remember things from 2019 that you would rather everyone had forgotten. They ask but why at exactly the moment when you were hoping the room had moved to how. They know things that do not fit the model. They feel things about the strategy that no engagement survey was subtle enough to detect. They talk to each other after the all-hands.
This is, by any reasonable measure, a significant inconvenience.
And somewhere in the long history of management theory, and the longer history of people finding other people difficult to manage, a very appealing idea took root: what if we could reduce the variable? What if the organisation could be designed so that the thinking happened in a small, trusted, aligned group at the top — and everyone else just executed?
Technology, in every generation, has offered this fantasy another chapter. And AI is writing its most compelling chapter yet.
The things leaders actually say
You will recognise these. You may have said them.
The AI will synthesise all the input we need.
Translation: we do not need to ask people things directly any more. We have their emails. We have the data. We have the outputs of the tools they use. We know enough. (We do not know enough. We know what the tools were designed to capture, which was not this.)
We don't need more feedback. We need better adoption.
Translation: the thinking is done. The problem is the humans who have not yet understood that the thinking is done. (The thinking is not done. The thinking has not yet encountered the people who will have to make it work.)
The survey showed 73% alignment — that's enough to move.
Translation: 27% of people have thoughts we are choosing not to engage with, on the grounds that 73% did not share them loudly enough. The 27% may include everyone who knows the thing that matters. Surveys are not designed to surface the thing that matters. They are designed to produce a percentage.
The data doesn't lie.
The data contains what was measured, measured in the way it was designed to be measured, by people who decided what to measure. The data is a selection. The selection is a point of view. But a point of view dressed in a spreadsheet is much more comfortable than a person saying something inconvenient in a meeting.
Once the system is in place, it will be self-correcting.
This one is particularly poignant. It imagines an organisation as a closed system, in which the designed inputs produce the designed outputs and the deviations are edge cases rather than signals. It has never described a real organisation. It will not describe yours.
The taxonomy of the problem human
Every leader has encountered all of these. Every leader has, at some point, wished they had not.
The Rememberer.
This person has been here longer than the strategy. When you present the new direction, they begin a sentence with "we tried something similar in —" and the room shifts in its seats. You need them to stop doing that. Unfortunately, what they remember is often exactly what you need to know.
The Why-Asker.
Three meetings into the implementation planning, they are still asking foundational questions. But what are we actually trying to achieve? It is maddening. It is also occasionally the question that reveals that three meetings of implementation planning were based on a shared misunderstanding of the goal.
The Customer-Knower.
This person speaks to customers in a way that no CRM can replicate. They know which customers are about to leave before the churn model does. They know which product feedback is signal and which is noise. Their knowledge is not in the system. When they leave, it leaves with them, and no amount of data migration brings it back.
The Quiet Non-Believer.
Says nothing in meetings. Nods at the appropriate moments. And then, quietly, continues doing things the way they were done before — not out of malice, but because they have seen too many strategies to give this one their full credence until it proves itself. They are contagious. And they are not always wrong.
The Open Dissenter.
The most valuable and most endangered species in any organisation. The one who says, in the meeting, what everyone else only says afterwards. They are rarely listened to in the moment. They are frequently vindicated six months later. By then, they have usually left.
The dream, stated honestly
The dream — fully stated, for once — is this: a small council of strategists and executives, working with excellent AI, producing clean and confident outputs, directing a workforce that executes without interpreting, questioning, or generating friction. An organisation chart as an elegant machine. Decisions flowing efficiently from the informed top to the executing bottom. No meetings after the meeting. No surprising dissent. No one remembering 2019.
This organisation exists, in partial form, in many places. And here is what it cannot do.
It cannot notice what is changing before it is in the data. It cannot challenge a direction that is heading for a cliff. It cannot adapt when the plan hits reality. It cannot generate the thinking that was never in any training corpus.
Nokia had the intelligence. It was distributed across the organisation in the heads of people who could see what was coming. It did not reach the decision that needed it. The efficient, human-light organisation is not a new idea. It is a very old one that keeps producing the same result.
What the idea of the problem human actually is
The idea is this: that people are primarily a source of friction, error, resistance, and cost. That they bring bias where you want objectivity, emotion where you want analysis, stubbornness where you want alignment, and opinions where you want execution.
This idea has surface plausibility. People do all of these things. They also bring the understanding that forms before the data, the knowledge that exists before the document, the signal that arrives before the trend. They bring the intelligence that no AI was trained on, because it had not been expressed yet.
Remove them from the thinking, and you remove the one part of the system that can surprise you productively. What remains is a very fast, very confident, very well-resourced machine for optimising the answer to last year's question.
The organisations that will outthink the rest are not the ones with the fewest inconvenient humans in the loop. They are the ones that built the layer, the operating system, to hear them properly — and had the unusual courage to act on what they found.
Related reading
— What AI cannot access — the structural argument for what no model can reach
— The missing layer — why the human intelligence layer is absent from every AI stack
— Get your deliberative house in order — what it looks like when you do want the full range of human intelligence
— The collective thinking problem is getting worse — not better with AI