Every strategy rests on a row of assumptions: about the market, about the competitor, about what the team itself can and cannot do. At the moment the plan was written, those assumptions were correct. No one has since agreed who keeps an eye on whether they still hold. That is not negligence. It is usually no one's task, because an assumption is not a line on an action list. It does not appear in a project plan, it sits underneath the project plan.
As long as the world stands still, this is not a problem. The world does not stand still. And the shift that is currently undermining assumptions the fastest is not a new competitor or a new law. It is that AI is taking over work that was previously fixed as human work.
An assumption such as "this role requires five years of experience and will remain scarce" or "this process can only be reviewed by a senior employee" was a fact at the moment it was written. The moment a task within that role can be done wholly or partly by AI, that assumption is no longer true, even though it still simply appears in the document.
Three situations run through almost every process: a task that AI can take over entirely, a task that AI can partly do provided a human approves or rejects the outcome with reason, and a task that remains human work. Which of the three applies changes by the month, by supplier, by process. An assumption that still stood firm in January may have lost its footing by September. Not because anyone made a decision, but because the underlying capability has shifted.
This touches on what an employer does with its staff, and its own legal requirements apply there; this page does not answer that question. What can be answered is whether the assumption on which a strategic decision rests still matches what is possible today.
At some companies it is already normal to revisit an assumption as soon as a signal comes in: a supplier offering a new feature, an industry report, a competitor doing something visibly different. There, it is fixed who owns the assumption, when it was last confirmed, and which signal warrants revisiting it.
At other companies, that record does not exist. The assumption appeared in a presentation from two years ago, the presentation is filed away, and no one has the task of looking at it again. The difference rarely lies in knowledge of AI. It lies in whether there is a place where an assumption has a status: valid, shaky, lapsed. Without that place, every insight disappears into the same drawer as the strategic document itself, and how do you prevent a strategic document from disappearing into the drawer is therefore not primarily a question about communication, but about ownership.
A strategic pressure test places an outside view next to the management's own self-assessment: what the sector, regulation, and visible competitive signals say, against what the team itself sees. Where the two diverge, an assumptions list emerges: for each assumption, whether it still holds, when it was last tested, and what would cause it to lapse.
This is not a measuring instrument with decimals. Sector data lag behind, public signals are partial, and a management team's self-assessment is an estimate, not a count. An outcome such as "this assumption is under pressure" is a signal to investigate, not proof. What the pressure test says nothing about: which decision a board should make. It shows where an assumption has started to shift, not what to do about it.
Nor can every assumption be tested in the same way. An assumption about market size can be substantiated with figures; an assumption about a team's resilience cannot. Anyone who wants to know how you get an assumption reduced to something you can measure will naturally run into assumptions that do not lend themselves to this. These then remain as a qualitative judgment, with the date on which that judgment was given, so that at least it is clear how fresh it is.
In almost every pressure test there are assumptions that contradict each other: the outside view says a task is already largely done by AI at comparable companies, the management's self-assessment says that for them this is still entirely human work. That is not a flaw in the method. It is precisely the point where a strategy quietly starts to shift, and what to do when two assumptions contradict each other is then the question that helps you move forward, not the question of which of the two is right.
Because AI capabilities shift per task and do not advance equally across companies, the distinction between a strategy and a plan is sharper here than elsewhere: a plan records what happens, a strategy records why, and what the difference is between a strategy and a plan explains why a plan can keep holding true while the strategy underneath it has already been abandoned. For an assumption that revolves around AI, it therefore makes sense to record separately why it no longer holds at a given moment; how you record why an assumption about AI lapses is about exactly that moment, not about what should happen afterward.
The underlying question for almost every assumption about AI is not abstract: which work in this company can genuinely be taken over by AI now, which work partly, with whom reviewing, and which work not. That question is answered per task with the work scan from FTE TO AI, separate from the question of whether, and how, a board acts on it.
The first step is not the full pressure test, which is under construction. The first step is the free assumption check: a short round in which you name your key assumptions and see, per assumption, when it was last confirmed. Anyone who first wants to know whether an assumption can be tested can start with what a strategic assumption precisely is and how you know whether it still holds.
Stel uw vraag. Vaak zit de echte vraag een laag dieper — daar mag ik naar vragen.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.