The supervisory board does not approve an AI project. That is not its role and it should not be: the executive board decides, the supervisory board oversees. But AI that takes over work touches precisely the domain that oversight is meant to cover, namely whether the plan the executive board presents still rests on assumptions that hold true. A strategy approved last year was based on a cost structure, a staffing need, a competitive position. If AI takes over work within parts of that structure — sometimes fully, sometimes with human oversight that approves or rejects, sometimes not at all — then the question for the board is not "is this project good" but "is the plan we approved still the plan that is actually in place".
What the board stands to gain from this is timely insight into erosion. What it stands to lose is credibility if it later turns out to have signed off on an assumption that had already stopped holding true months earlier. Oversight that only asks about AI once it becomes visible in the annual accounts asks too late.
The board has no role in the question of exactly which task is taken over by AI — that is execution, and it falls to the executive board and, where staff are affected, to the applicable statutory requirements on employee participation and labor law. What the board can and should ask is: which assumption underlying this strategy has recently been checked, and by whom. That is a different question from "does it work", and a different question from "is it permitted". It is the question of whether the plan still matches the world in which it is being carried out.
That question does not accept an answer based on assumptions. "We believe our people continue to add value here" is not an answer the board can assess against; it is an assumption that itself still needs confirming. A board that seriously assesses AI decisions does not ask for a demonstration of the technology, but for a list: which assumptions underlie this plan, and when was each of them last checked against what is actually happening in the company.
The friction usually does not arise over the technology, but over pace and over who signals it first. An executive board that has built a strategy has an interest in it holding up — for an executive, admitting change is a different conversation than preventing change. The supervisory board has the opposite interest: it must see early when an assumption starts to shift, even when that is uncomfortable for the person who made the plan.
That is the same tension that shows up elsewhere at the table. What the CEO keeps watch over when AI is set to take over work is often the story told to the market and the organization; what the CFO keeps a close eye on during the same shift is the cost base and capacity planning; what a commercial director sees changing is the promise made to the customer. The supervisory board does not sit in any of those chairs, but it hears the cross-section: it is the body that can ask whether these three pictures still fit together, or whether they have drifted apart without anyone naming it. That happens more often than a board finds comfortable, and it is exactly why disagreement over priorities within the executive board so often traces back to an assumption that one executive has already let go of while another is still holding on to it.
In some companies, there is already an overview of which tasks AI can take over, which fall under oversight, and which remain human work, and that overview is reviewed periodically. There, the board can ask its question and get a concrete answer. In other companies, that overview does not exist, and the question is answered with an impression: "we're working on it", "it's progressing". The difference does not lie in the sector or in the size of the company. It lies in whether someone has ever written down the assumptions underlying the strategy as assumptions, rather than as established facts. Where that has happened, the board's question is a fifteen-minute conversation. Where it has not, every AI decision is a fresh discussion without common ground.
The underlying question — which work in this company can genuinely be taken over by AI, and which part remains human work or falls under oversight — is mapped per task using the work scan from FTE TO AI, independent of what the board subsequently decides to do with it at governance level. That is, in fact, exactly the distinction: the scan delivers the factual picture, the board remains the one that judges whether the plan still fits it. Those who want to know more about what a strategic assumption precisely is and how to distinguish it from a fact will find that worked out under what a strategic assumption is and how to check whether it still holds.
A board that wants to know whether it is still assessing against an up-to-date strategy can start with a free assumption check: a short round in which the key assumptions underlying the current plan are named, and for each assumption it becomes visible when it was last confirmed. This is not a judgment on the plan and not advice on staffing — it is a status report, drawn up before the next AI decision reaches the agenda. The full proof print, showing the outside view alongside the executive board's own self-plot, 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.