When the price of AI capacity falls, something specific changes: tasks that were too expensive to automate last year become profitable this year. Not because the task itself changes, but because the calculation behind it shifts. A strategy based on a fixed ratio between human work and machine work rests on an assumption determined by this price. As long as the price is stable, the assumption holds. Once it moves, the assumption is no longer automatically valid, even if it still stands unchanged in the plan.
That makes the price of AI capacity a signal worth tracking, much like an exchange rate or a commodity price. Not to react to every fluctuation, but to know when an assumption is due for review.
The price of AI capacity is not a single figure. It is spread across several movements, each running at its own pace:
These movements do not run in step. A model may become cheaper while integration remains expensive, or the other way around. Anyone looking only at the model price sees just part of the signal.
In some companies the calculation is already visibly shifting: tasks that were human work last year now fall into the category where AI can do the largest part of the work, with a human approving or rejecting with reason. In other companies the division remains unchanged, not because the technology won't arrive, but because the task itself does not lend itself easily to being taken over: too many exceptions, too much context, too much dependence on judgment that does not fit into an instruction.
The difference, then, lies not only in the price of AI capacity, but in the nature of the work exposed to that calculation. A task with a fixed structure and a clearly testable outcome responds directly to a price drop. A task that revolves around negotiation, explaining something to a customer, or a decision carrying legal weight responds far less, even if the price of AI capacity falls sharply. What your own people are already doing with AI often shows where that boundary lies in practice sooner than a price sheet does.
The price of AI capacity does not change daily in a way that is relevant to a strategy. A quarterly rhythm is sufficient for most companies: often enough to notice a structural shift, not so often that every announcement calls for a revision. Anyone working in a sector where competitors are visibly betting on AI capacity, or where new entrants without staff play a role, looks more often.
A drop in the price of AI capacity means that an assumption about the human-machine ratio is due for re-examination. It does not mean that a task is actually taken over: that depends on the nature of the work, on the quality of the oversight required, and on choices that rest with the employer. This page describes a signal, not a decision. Where the question touches on personnel decisions, separate legal requirements apply; these are not addressed here.
A price drop also does not mean that every task falls into the same category. Some tasks shift from human work to oversight with human approval, others remain human work because the task itself demands it, regardless of how cheap the capacity becomes.
The decision that fits this signal is not "automate more or less". It is the question of whether the assumption in the strategy still matches the current price ratio, and whether that has consequences for where capacity is deployed. That question is separate from regulation around AI, which runs along a different track and for which there is a separate way to track how you follow regulation around AI, and separate from the question of how you track what customers themselves are going to do with AI, which touches the demand side rather than the capacity side.
The underlying question – which work in this company can genuinely be taken over by AI – is answered per task by FTE TO AI's work scan, independent of what the price does at sector level.
A price signal only becomes useful once it sits alongside a list of assumptions it can affect. Anyone who does not yet have a view of how to track such a signal systematically can start with what a horizon scan is and how you do one yourself; anyone who notices that this kind of signal never lands on the board agenda will find starting points in how you get regulation structurally onto the board agenda.
Concrete and without obligation is the free assumption check: a short round in which you name your key assumptions and see, for each one, when it was last confirmed. The full strategic pressure test, with sector data, a self-plot for the management team and an ongoing assumptions list, 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.