A banker assessing a credit relationship or financing works with a cost structure as a prediction. Not as a photograph of last year, but as a basis for a repayment capacity that must hold up in the coming years. If that structure changes because part of the work is done by AI instead of by people, the predictability of the figures on which he bases his assessment changes. His risk is not that costs fall. His risk is that he is assessing a cost structure that is already shifting while the reporting still has the old form.
What he stands to gain is simpler: a borrower who can show which part of the costs has become variable because work has been transferred to AI, and which part remains fixed because it requires oversight or is human work, gives him a structure he can calculate with. Unclarity about that costs him more than a lower figure. Unclarity costs him confidence in the rest of the file.
The banker's question is not about technology. He asks: which part of this cost structure is subject to change, and has that change already happened or is it still expected? A cost item that AI has taken over is a different kind of cost item than one that is still human work: the first can move faster with volume, the second cannot. Whoever does not keep these two categories separate in the explanation of the figures gives the banker a structure he cannot weigh.
The answer he does not accept is a figure without substantiation of its origin. "Costs are falling because we work more efficiently" is not an answer on which a credit assessment can rest. Which tasks have been taken over, with what oversight, and since when — that is the layer he needs. Without that layer he sees a trend without a cause, and a trend without a cause he cannot extrapolate.
In some companies this explanation already exists: the reporting to financiers distinguishes costs that have structurally fallen because a task is now largely done by AI, from costs that have merely shifted to oversight instead of execution. In other companies this explanation does not exist, and a decreased cost item is reported as the result of "efficiency" without further clarification.
The difference does not lie in the size of the company or the sector. It lies in whether someone at the boardroom table has explicitly abandoned the assumption underlying the old cost structure. A structure in which certain costs were budgeted as fixed because they were always human work stops holding up the moment part of those tasks is transferred to AI. Companies where this has been noticed and recalculated can show it to a banker. Companies where this has not yet been noticed leave an old assumption in the figures without anyone having withdrawn it.
The banker clashes most often with the commercial director. That person looks at revenue and growth capacity and sees mainly room in AI: processing more volume without proportional cost growth. The banker looks at the same shift and sees something else: a cost structure that has become less predictable as long as it is not established which part has been permanently transferred and which part becomes human work again as soon as the quality of oversight requires it. Growth resting on an unconfirmed assumption about automation is, for the banker, not growth but a risk that has not yet been named.
He also clashes with the operations director, who mainly sees the process: tasks have been transferred, it works, done. The banker asks about the durability of that over a longer term than an operations director usually has in mind. What works today with the current level of oversight may look different at the next scaling step, and the banker wants to know whether that difference has already been thought through.
These clashes are not a disagreement he settles himself. Whoever wants to adjust this cost structure based on what AI takes over has their own statutory requirements to meet for that; the banker judges the figures, not the path to them.
Underneath all these questions lies one question: which work in this company can genuinely be taken over by AI, which part is already happening with oversight, and which part remains human work as long as no one has shown otherwise. That is the same question that comes up in the question a cfo looks at when AI takes over work, in the trade-off a commercial director makes between growth assumptions and automation, and in the way a strategy review determines who should have a say in these assumptions. The work scan of FTE TO AI answers that question per task, not as an estimate but as a count.
If the cost structure presented to a banker rests on an assumption about human work that has since shifted, that is not a detail for the next annual report. The free assumption check is a short round in which you name the key assumptions underlying your strategy and see, for each assumption, when it was last confirmed. The full proof press, with a self-plot from the management team alongside the outside view, 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.