A performance measurement is built on an assumption: whoever does the work determines how you measure it. A salesperson gets a revenue target, a support employee a handling time, an analyst a turnaround time per report. As soon as part of that work is done by AI — sometimes fully, sometimes with a human who approves or rejects — what there is to measure shifts. Not because the KPI was wrong, but because it measures an action that no longer happens in the same way.
The three categories in which AI takes over work run right through every performance measurement. Work that AI fully takes over calls for a different metric than turnaround time: that time no longer exists as a bottleneck. Work where a human maintains oversight — approving or rejecting, with reason — calls for a metric that captures the quality of that judgment, not the speed of the old action. And work that remains human work keeps its old KPI, but gains weight because it remains after the other has shifted.
Some organisations have already processed this, others have not. The difference rarely lies in the technology itself, but in who set the KPIs and when. A KPI drawn up five years ago for a role assumes a way of working that has now been partly taken over. If nobody has revised that KPI, the organisation is still measuring something — just no longer what is actually happening.
You notice it first in the reports that no longer match the feeling on the floor. A team hits its targets, but nobody can explain why output has risen so quickly. Or a target is structurally missed, while output has actually increased — because the metric still measures the old action, not the new one. Both are signals that the measurement is lagging behind the work.
Three shifts recur in virtually every organisation where AI work has been implemented, in varying order and speed:
None of these shifts happen everywhere at the same speed. In an organisation where administration has already shifted, the measurement of customer contact sometimes still lags years behind. That is not a contradiction — it is precisely why performance measurement must be examined per task category, not per department.
A KPI that no longer aligns with the work undermines not only the measurement but also who is entitled to decide based on it. If a team leader is held accountable for a figure that no longer accurately reflects the work, the question of who is actually responsible for the outcome also shifts — the human who approves, or the system that supplies the work. That touches on decision rights within the organisation, and that is exactly where an ambition gets stuck if the layer beneath the KPIs has not been revised.
This topic sometimes touches the workforce — who still does which work, and what that means for a role. Separate statutory requirements apply to that, apart from what is described here.
Performance measurement does not stand alone. An ambition that assumes AI work must also answer how you weigh growth against margin when capacity shifts, what that means for the organisation's financial resilience as work changes shape, and how a vision is made measurable so that the KPIs under that vision actually align. Without that coherence, you measure one part, while the ambition concerns the whole.
The underlying question — which work in your company can genuinely be taken over by AI, and where that requires oversight rather than takeover — is answered per task by FTE TO AI's work scan.
The question of whether your performance measurement is ready for work that is shifting cannot be answered in a single conversation, but it can be started. The free readiness check gives you, with eight short questions — one per dimension — a picture of where you are furthest along and where you are least far along. The full ambition assessment, with the four layers from vision to decision rights, is under construction.
Vertel wat u wilt bereiken, dan kijken we samen wat daarvoor moet staan.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.