An assumption in a strategy rarely gets an expiry date. It says: "this work will remain human work" or "this process requires our people," and that was true at the moment it was written down. The question is not whether that assumption will ever become incorrect. The question is when, and who notices before the plan quietly keeps building on it.
Market assumptions wear out gradually. A competitor slowly gains ground, a price drops over quarters. An assumption about AI can tip over on a different timeline. A task that was entirely human work last year may this quarter be largely takeover-ready, partly with oversight that approves or rejects with reason, partly not yet. That does not happen evenly everywhere. In a company where the underlying tasks are already structured and digitally available, the assumption shifts faster than in a company where the work largely sits in people's heads and conversations. The difference is not in ambition, it is in what is already measurable and transferable.
That is also why a fixed retesting frequency does not work. An annual strategy cycle assumes that all assumptions age at the same pace. Some do. The assumption about AI usually is not one of them, because the underlying technology and the regulation around it do not wait for the calendar of your strategy process. How you keep a strategy current without revising it every quarter describes why the assumption, not the calendar, determines the moment of retesting.
There is no universal answer to "how often." There are factors that determine the shelf life of a specific assumption. How directly does the task touch text, data, or images that a model can process. How fast are the vendors and regulators moving in this specific domain. How visible is that movement to the people who carry it out daily, versus to the leadership team that wrote the plan. An assumption about a task that is already partly automated deserves a shorter shelf life than an assumption about work that still rests entirely on human judgment. That distinction is exactly why a list with all assumptions on the same date does not work: every assumption has its own clock, and that clock is set by what is happening in that specific part of the organization.
That also immediately states the limit of this method. Setting a retesting moment is not a prediction of when the assumption will tip. It is an estimate of when looking again is worthwhile. Sometimes the retesting moment reveals that nothing has changed, and that is a valid outcome, not a missed opportunity.
An assumption check can give a false sense of certainty if the underlying data is itself already outdated, or if the person assessing the assumption is the same person who originally formulated it. Self-confirmation is a real risk: someone who believed an assumption when they wrote it is not the most critical assessor a year later. That is why the stress test places an outside view, sector data, regulatory clocks, public competitive signals, alongside a self-assessment from the leadership team itself. Differences between those two pictures are a signal, not a conclusion. They point to where further investigation is needed, not to what must change immediately. And who in the organization is the first to see that AI is taking over work is often not the leadership team itself, which means an assumption check without signals from the shop floor misses part of the picture.
Also important: this is not about personnel decisions. An assumption that no longer holds says something about the strategy, not about what an employer does with its staff. Decisions that affect the workforce are subject to their own legal requirements, separate from what this method produces.
Another risk is confusing a pilot with a structural shift. A successful experiment in one team does not mean the assumption no longer applies to the entire organization, just as a failed pilot does not prove the assumption still holds. Why an AI pilot is not yet a strategy is about exactly that difference. Whoever determines whether a signal on the shop floor is a real shift or noise finds guidance in how you distinguish an AI hype from a shift that lasts. Both questions belong to the same discipline: not every signal is a changing assumption, but every ignored assumption is a risk waiting to happen.
The underlying question, which work in this company can genuinely be taken over by AI at this moment, cannot be answered with a general estimate. That differs per task, and the work scan from FTE TO AI maps that out per task.
An assumption check is not a task for one role. The director who wrote the assumption, the manager who sees the task carried out daily, and the person who tracks external signals each look at a different part of the picture. Who should take part in a strategy review describes why that group is larger than leadership alone, and why an assumption check without that spread produces a narrower picture than needed.
You can start today with an overview of the assumptions your strategy carries, and for each assumption establish when it was last confirmed. That overview alone shows which assumptions are standing still while the rest of the strategy keeps moving. FTE TO AI offers a free assumption check for this: a short round in which you name your key assumptions and see for each one when it was last confirmed. The full stress test, with the external view and the self-assessment side by side, 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.