This is the heart of my research, and it is a trickier question than it looks: when a generative AI enters a small company, what really changes in the way decisions get made?
The easy answer would be that the leader decides better, faster, better informed. That is too simple, and sometimes it is the opposite. What I see in the field, and what the figures hint at, is more subtle.
Start with a fact. In a Bpifrance Le Lab survey published in 2025, run with more than 1,200 leaders of small and mid sized companies, 58% see AI as a matter of survival within three to five years. But only 43% have formalised a strategy, and 26% actually use a generative tool. Between the fear and the act, a chasm. And one last figure, the most telling for me: in nearly three quarters of cases, it is the leader themselves who decides to commit the company, with no technical intermediary. Not a committee. One person.
That last point is the key to everything that follows.
Uncertainty does not disappear, it disguises itself
The first change is insidious. A generative AI answers with fluency and assurance, and that assurance gives the impression of solid ground. Yet the machine only produces the plausible. The leader who takes that confidence for certainty is in fact deciding on well presented sand.
Before, uncertainty was visible. You knew that you did not know, and that awareness made you careful. Now the tool dresses the blur in a clean answer, and caution lowers its guard. Deciding in the age of AI means first relearning how to tell a sure answer from an answer that is merely sure of itself. That is not a technical skill. It is a mental discipline.
The decision moves upstream
The second change runs deeper. Before, the leader decided at the end, after painfully gathering the information. Collection was slow, and the moment of choice came once the picture was complete.
Today information arrives instantly, abundant, already shaped. So the real work of deciding moves upstream: it becomes the work of framing. Which question I ask, which options I let the tool explore, which ones I forbid myself to delegate to it, which assumptions I test. The decision has not vanished, it has changed its moment. It now plays out in how the question is framed, as much as in the answer.
It is a reversal many live without naming it. They feel they are saving time, and at the same time a vague unease, the unease of deciding faster without always understanding better. That unease is right. It signals that the centre of gravity of the decision has moved.
In an SME, it all plays out in one head
The third change comes from the very nature of a small company, and it is what makes it a unique field of study.
In a large company, AI slots into processes, approvals, departments. There are shock absorbers. In a small company, the decision concentrates in the figure of the leader, with their intuitions, their relationship to risk, their blind spots. When that person adopts a tool that answers everything with assurance, there is no one to say wait, we are not sure here. The filter is them, and them alone.
That is why I chose this field for my thesis. Strategic decision making is laid bare there. You see how it really forms, without the fog of complex organisations. I work on it through a multiple case study of eight to twelve small companies, bringing together three frameworks that are rarely combined: organisational paradoxes, dynamic capabilities, and strategic decision making. The aim is not to measure adoption, but to understand how a leader makes sense of a radically new tool, and reshapes their practice around it.
What this means in practice
From this I draw three principles I share with the leaders I work with.
First, treat every AI answer as a proposal, never as a verdict. The nuance changes everything: you keep the stance of someone who evaluates, not someone who executes.
Then, look after your questions as much as your decisions. A poor question gives a fluent, useless answer. The time the tool saves must be reinvested in the quality of the framing, otherwise you accelerate towards the wrong place.
Finally, identify what you will never delegate, not even to a very convincing machine. The choices that commit the company for the long term, the ones that touch people, the ones you cannot undo. On those, AI informs, it does not decide.
My conviction, after months on this subject, fits in one sentence. Generative AI does not replace the leader’s judgement, it puts it under tension. It forces them to know, better than before, what belongs to the machine and what belongs to them. The leaders who come through this period best will not be the best equipped. They will be the ones who kept the clarity to decide when everything pushes them merely to approve.