In every boardroom there is someone who speaks the fastest and most confidently about AI. That person has seen a pilot, spoken with a vendor, or read an article that sounded convincing. That is not a problem in itself. The problem arises when that voice, without anyone deciding it, determines the order: which department gets AI capacity assigned first, which tasks get investigated and which do not. Not because that person knows the most, but because that person speaks the loudest.
That is a decision-making problem, not a technology problem. And it is precisely the type of shift on which a strategy quietly becomes outdated: the assumption that priorities in the boardroom rest on facts, while in practice they rest on persuasiveness.
AI taking over work does not happen as a project with a start date. It happens in parts: a task that is fully handled by AI, a task where AI makes a proposal and a human approves or rejects it with reason, and a task that remains human work. Those three categories run through every department, and the distribution differs per company, per process and per moment.
As long as that distribution is unclear, someone fills the gap with an opinion. Companies where that gap has been filled with a picture of the tasks themselves do not have that problem because the discussion has been resolved — they have it because the discussion is no longer necessary. The question shifts from "who is right" to "what does the task list say".
The stress test we use places two things side by side. On one hand a picture from outside: sector data, regulatory clocks, public signals about what competitors are already doing. On the other hand a self-plot of the company's own management team: where does each board member think AI is already taking over tasks, and where does one think it is not yet.
The point is not that one picture should beat the other. The point is that differences between those two pictures become visible, and that a difference says something. If the outside world is already far along on a task that is still internally considered "human work", that is a signal to investigate, not to ignore or force through. What you do with that — assess a task on an accelerated basis, involve a team earlier — remains a decision made by the board itself. The stress test delivers the contrast, not the conclusion. What such a board member who does not share the picture should then do with it depends on why that person has a different picture: different information, different experience, or a different interest.
The self-plot is a snapshot of opinions, not of facts. If three of the five board members think the same thing about a task, that does not mean they are right — it means they agree. Consensus within a board is not confirmation from practice.
The outside picture has its own limit: sector data and public signals lag behind reality, and a competitor that announces something does not necessarily have it working yet. Both pictures together give an indication of where the conversation should go, not a verdict on where the conversation should end.
And an outcome mainly says something when there is a difference. If the outside picture and the self-plot run roughly parallel, that is reassuring, but it does not confirm that the assumption is correct — it only confirms that no one has contradicted it. That is a different kind of certainty than is often assumed, and it helps to know that difference before basing a decision on it.
Every assumption in the strategy gets a status: does it still hold, or has it quietly lapsed without anyone withdrawing it? That is a different question than "who is right about AI", and precisely for that reason less sensitive to the loudest voice. The question is no longer what someone thinks about the future, but when a specific assumption was last checked and on the basis of what fact. How often that needs to happen again differs per assumption and per pace of the market — you can read more about this at how often an assumption about AI needs to be reconfirmed.
This also prevents another pitfall: confusing a successful trial with a durable strategy. A pilot that works in one team does not yet say anything about the rest of the company, and that difference is addressed in why an AI pilot is not yet a strategy. Whoever does not make that distinction explicit risks letting a chance success take the place of a well-founded order of priority.
This is not a tool to determine who has to go or whose job is redundant. What an employer does with the outcome of such an analysis with regard to personnel falls under its own legal requirements and its own responsibility; we do not make that assessment and we do not replace it. What is delivered here is a factual picture of tasks: which part of the work AI can already handle, which part can be done with oversight, and which part remains human work. That question, applied to the tasks of your own company, is precisely what the FTE TO AI work scan answers, per task and not in generalities.
If you want to know whether the order in your own boardroom rests on facts or on persuasiveness, the first step is not a large investigation but a brief inventory: which assumptions does your strategy carry, and when was each of them last confirmed. That is exactly what the free assumptions check offers — a short round in which you name your main assumptions and see per assumption when it was last tested. The full stress test, with the outside picture and the self-plot 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.