A CFO does not ask whether AI is taking over work. That is already happening, in parts, spread across the organization, without any project bearing that name. The question a CFO asks is more precise: which part of the cost structure in the multi-year plan is built on an FTE deployment that is no longer accurate. A forecast calculates based on assumptions about how many people do which work. If AI takes over a task, the assumption has not become wrong, it has expired. No one has withdrawn it, but it no longer holds.
That distinction is not trivial for a CFO. An error in a model gets corrected as soon as someone notices it. An expired assumption goes unnoticed, because the model simply keeps calculating. The business case, the cost calculation, the capacity plan for next year: these stand on figures that were still accurate three quarters ago and may no longer be so now.
In some organizations this is already being tracked: there is an overview of which tasks AI can take over, which tasks fall under human oversight where someone approves or rejects with reason, and which tasks remain human work. That overview is updated periodically, and the FTE capacity that becomes available as a result is factored into the capacity plan as freed-up hours, not as an eliminated position.
In other organizations that overview does not exist. There, the assumption simply continues in the model until someone asks about it, usually during a budgeting round or an audit conversation. The difference between these two situations does not lie in the sector or the size of the company. It lies in whether someone has been given the task of periodically testing assumptions, and whether that person has access both to the outside view — what competitors and the sector are already automating — and the inside view of how the organization's own leadership actually sees the work shifting.
A CFO does not accept an answer that relies on sentiment. "We have a feeling that AI is going to make a big difference here" is not a figure and not an assumption that can be dated. Equally unusable is the reverse: "this doesn't affect us," without anyone having checked which tasks in the organization's own processes fall under the category that AI can already take over or perform under oversight.
What a CFO does accept is a statement with a date attached. Not "this work is being taken over," but "this assumption was last tested in quarter X, and the sector data from that time supported it." That is not a prediction. It is a timestamp on something that would otherwise age unnoticed.
The CFO calculates in capacity and costs. The HR director calculates in people and obligations. These two readings of the same shift do not automatically align. Where the CFO sees that a task can be taken over and thus frees up hours, the HR director sees an obligation, a team, a conversation that needs to be had — and that conversation falls under its own legal requirements, which are not addressed here. What a CFO signals as capacity headroom is, for HR, a process with an entirely different timeline. See what an HR director watches for when AI starts taking over work for that side of the table.
There is also a clash with the operations director, who sits closer to execution and sometimes notices sooner that a task is already being partly taken over than the figure about it reaches the CFO — read about it in what an operations director watches for when AI starts taking over work. And the supervisory board member asks a different question than the CFO: not how much capacity is freed up, but whether leadership is tracking this structurally or only reacting once it is already visible, see what a supervisory board member watches for when AI starts taking over work.
Not some miraculous cost saving that can be captured in a single figure — that depends on which tasks in the company fall under which category, and that differs by process and by team. What a CFO does gain is a capacity plan that matches what is actually happening, rather than what was true three quarters ago. And an answer to the question from the audit committee or the supervisory board that does not rest on gut feeling.
The underlying question — which work in this company can genuinely be taken over by AI, and which part remains human work — is answered task by task with the work scan from FTE TO AI.
This is not a one-time exercise. How often an assumption needs to be retested depends on how quickly the underlying work is changing, something how often should you recalibrate a strategy explores further. And if the market is visibly shifting faster than the organization's own plan, that is a different conversation than a regular recalibration — see what to do if the market changes faster than the plan.
No figure in this piece replaces a look at your own numbers. What can help: a free assumption check, a short session in which you name your key assumptions and see, per assumption, when it was last confirmed. The full stress test, with sector data, leadership's self-plot, and the assumptions list 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.