A strategic plan contains sentences that no one recognizes as an assumption, because they were written down as fact. "Our people do this work, so that will remain a cost item of that size." "This department is the reason we're faster than the competitor." Those aren't conclusions, they're assumptions that were correct at the time they were written. The question isn't whether they were correct then. The question is whether anyone has tested them since.
A fact is true today and true tomorrow, until proven otherwise. An assumption has a shelf life, even if no one has written that shelf life down. "This task requires five experienced people" was a reasonable assumption two years ago. Whether it still holds depends on what has changed about the task itself in those two years, not about the people carrying it out. That distinction is the difference between an AI risk and an AI assumption: a risk is something that could happen, an assumption is something you already take to be true without checking it anymore.
AI that takes over work doesn't affect every assumption equally. For work made up of structured steps, a task can by now largely be done by a system. For work that requires judgment, that happens partly, with a human approving or rejecting and giving a reason for it. And a third category of work barely changes, because the task revolves around something for which no training data exists or for which responsibility is fixed to a person.
The reason this has already been implemented in one company and not yet in another is rarely the technology. It comes down to who within the organization was first to distrust the assumption. A team leader who sees that a report now arrives largely automated knows that the assumption "this is work for three people" no longer holds. Whether that knowledge reaches the board is another matter.
Making an assumption measurable doesn't mean you make up a percentage. It means you record three things: what the assumption states precisely, what it was based on at the moment it was written down, and what signal would show that it no longer holds. That third point is where most strategic documents fall silent. They state what is assumed, not when that would be reconsidered.
That's why it's worthwhile to note a last-confirmed date for each assumption, just as a certificate has an expiry date. Not because the assumption automatically becomes false after that date, but because it tells you that no one has looked at it since. That's how you record why an assumption about AI expires: not as a judgment on whoever once made the assumption, but as the date on which the world beneath it changed.
This approach doesn't produce a prediction. It doesn't say when a task will be taken over, or whether that will happen. It only says whether the assumption your plan rests on currently still has backing in what is happening outside and inside the company. An assumption that is no longer backed isn't necessarily false. It may simply be that the backing hasn't been checked recently.
The method also says nothing about what you should do with that outcome regarding staffing. If a task turns out to be largely transferable, that's a fact about the work, not advice about the people currently doing it. Decisions on that front have their own legal requirements, and those lie outside this proof check.
There is one more limit: an outcome says nothing as long as the assumption itself hasn't been formulated sharply enough to be contradicted. "AI changes this work" is not an assumption you can test. "This task requires an experienced reviewer because the output is unstructured" is. Sharpening the assumption is often more work than testing it.
Within a management team, the problem rarely arises because someone defends an assumption that is clearly no longer true. It arises because two assumptions, each reasonable on its own, can no longer coexist. One department assumes a task setup that the other department has already abandoned. See what you do when two assumptions contradict each other for how that contradiction becomes visible before it ends up in a decision. Often that is the first signal that perceptions within a management team are diverging, and the question why perceptions within a management team diverge is about exactly that mechanism: not because someone is wrong, but because no one updated the same assumption last.
Beneath all these assumptions lies one question that is answered differently at every company: which work here can, in fact, be taken over by AI, which work partly with oversight, and which work not at all. That is precisely what the work scan from FTE TO AI maps out per task, independent of what the organization has assumed about it so far.
Reducing an assumption to something measurable doesn't start with a system, but with a list: which assumptions does your strategy carry, and when was each one last checked. That is also exactly what outside-in thinking in practice comes down to: not taking your own assumption as the starting point, but testing whether the world outside the organization still confirms it.
The free assumption check is a short round in which you name your key assumptions and see, for each one, when it was last confirmed. No report, no advice, just a date next to every assumption you already had. The full proof check, showing the outside view alongside the board's own self-assessment, is under construction.
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.