A strategic plan contains sentences like "this process requires specialised people" or "this reporting takes a team a week". Those sentences were true at the moment they were written. They do not become untrue at the push of a button. They become untrue because AI has, at some point over the past quarter, started taking over a task that was previously considered a fixed part of the work, or because that task has become partly takeable over with human oversight in place. Nobody deletes the sentence from the plan. It remains there, and the board still acts on it, while the ground beneath it has already shifted.
That is the phenomenon this is about: not a risk that presents itself, but an assumption that quietly loses its validity. The difference between the two lies in what you need to do about it, and what is the difference between an AI risk and an AI assumption goes into that further.
Recording an assumption in the way meant here is not writing down an opinion. It is noting three things: what exactly the assumption states, on the basis of which fact it is now judged valid or lapsed, and when that judgement was last made. Without that third line, a list of assumptions is a snapshot that will be just as meaningless within a quarter as the plan it sits next to.
The reason an assumption lapses rarely lies in an announcement. It lies in an accumulation of small signals: a supplier adding a feature that partly automates the approval of a task instead of its execution, a competitor dropping a feature without explanation, a sector in which oversight tasks shift from executing to checking. Each signal on its own proves nothing. Together they form the burden of proof with which you can withdraw or confirm an assumption. How often that check needs to take place before an assumption may still be considered current is a question with an answer of its own, set out in how often must an assumption about AI be reconfirmed.
The difference between companies that actively maintain their assumptions and companies that do not rarely lies in knowledge of AI. It lies in who within the board is allowed to let an assumption lapse. In one company that is a fixed procedure: an assumption is periodically tested against external evidence, independent of who originally introduced it. In the other company an assumption remains in place until someone with enough weight in the meeting puts it up for discussion, and that only happens once the consequences are already visible.
That second pattern is often linked to who speaks most forcefully about AI in the boardroom, not to who holds the most recent evidence. how do you prevent the loudest voice from determining what AI takes over first describes where that distortion arises and how a record actually protects against it: an assumption with an expiry date and an evidence rule does not depend on who is most forcefully convinced.
This way of recording does not say whether a task is actually being taken over, and says nothing about when that happens. Nor does it say anything about what a company should do with that information regarding staffing; that is a decision subject to its own legal requirements and is not answered here. The method also gives no certainty about the company's own situation: external evidence -- sector data, regulation, what competitors publicly show -- tells you what has become possible, not what is already happening in this specific company.
That is why the record only has value alongside a second picture: how the board itself assesses which tasks AI can take over, which are takeable over under oversight, and which remain human work. That self-assessment is the other half of the stress test. When the outside view and the self-assessment diverge, that is rarely a sign that someone is mistaken. More often it points to a real difference in information position between board members, and why do the views within a board diverge shows where that difference comes from.
The easiest mistake to make is using the organisation itself as the only source for determining whether an assumption still holds. A board that only looks inward rarely sees how quickly the outside world is changing, precisely because internal processes generally shift more slowly than the market around them. Looking outward -- at what comparable companies have already adjusted, what regulators already require, what already exists as a feature in software -- is a different way of working than reasoning from your own plans. what is outside-in thinking in practice describes what that looks like without it becoming an exercise in importing someone else's strategy.
The underlying question -- which work in this company can genuinely be taken over by AI, which part under oversight, and which part remains human work -- is answered per task in the work scan from FTE TO AI.
A list of assumptions does not start with a system. It starts with a short, honest summary: which three or four assumptions does the current strategy carry, and when did someone last check whether they still hold. Anyone who cannot answer that already knows something from it.
The free assumption check is a short round in which you name your key assumptions and see, per assumption, when it was last confirmed. The full stress test, with sector data, regulatory clocks and the board's own self-assessment placed 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.