A competitor loses staff, or simply stops hiring for a department that was still growing last year. The conclusion seems obvious: AI has taken over the work there. That conclusion comes too quickly. Staff turnover is an outcome of many factors at once — reorganizations, market conditions, individual departures, seasonal patterns in certain sectors. AI adoption is at most one of them. Using turnover figures as proof of AI maturity confuses a symptom with a cause.
At the same time, it isn't nothing. If a competitor hasn't replaced staff in a specific role — say, planning, first-line support, or contract review — for two years running while the volume of work stays the same or increases, that's an indication that calls for an explanation. The question isn't whether turnover says something, but what else needs to be in place before it means something.
AI takes over work in three flavors: fully automatic, partly with human oversight that approves or rejects, or not at all — it remains human work. On a dimension where AI takes over the task completely, staffing needs change structurally: fewer people can process the same volume, and that can show up in job openings or outflow. On a dimension with oversight, mainly the nature of the work changes, not necessarily the headcount — assessment is still needed, just distributed differently. And on dimensions that remain human work, turnover says nothing at all about AI; the usual reasons are at play there.
That means turnover figures are only useful if you link them to the question of which task within that department has changed. A decline in headcount in customer service could come from chatbots catching first-line questions, or from a shift in market share. Without that distinction, the figure is noise.
A few signals are more specific than raw turnover. Job postings that shift from executing to reviewing — no longer "processes applications" but "assesses outcomes of the system" — point to a shift toward the oversight category. Roles that simply disappear from the job listings without a replacement role elsewhere point more toward full takeover. And the absence of any change in job profiles, despite announcements about AI investments, is itself also a signal: it suggests that adoption is still limited to pilots.
These signals are externally visible — via job postings, LinkedIn profiles, annual reports — and that's precisely why they're suited to a benchmark. You don't have access to a competitor's internal decisions. You do have access to what an organization shows externally about how it structures its work. How to measure a competitor's AI maturity from the outside goes further into which public sources are useful for this and which aren't.
An important limitation: turnover and job postings tell you something about the staffing side of a shift, not about the quality of execution. A competitor may have scaled down a team without the underlying AI system functioning well — the work has been taken over, but not necessarily done better. Conversely, a competitor may keep a full team in place while AI already handles a large part of the preparatory work, simply because the freed-up capacity has been deployed on other work rather than resulting in outflow. Freed-up hours are not the same as fewer people; an organization can redistribute capacity without this showing up in the staffing figures.
It also applies that: if the topic touches on decisions about layoffs or reorganization at your own organization, separate legal requirements apply to those, independent of what an external benchmark shows. These signals are suited to assessing a competitive position, not as a basis for staffing decisions.
The reason to look at this at all isn't interest in AI as such. It's the question of whether a competitor is redefining a dimension on which you're currently still winning — speed, price, availability. A competitor that delivers faster with fewer people isn't automatically becoming cheaper; why fewer staff at a competitor doesn't automatically mean lower costs shows which assumptions in this regard often go untested. And if your own strongest sales argument rests on something that AI is also starting to automate at competitors, the question isn't whether you mind, but what you can still claim; what to do when your strongest point is simply replicated by AI addresses that situation specifically.
Staff turnover, then, is never the endpoint of a benchmark, at most a reason to look further: at job postings, at job descriptions, at what a competitor itself communicates about automation. Isolated figures without that context lead to how to score a competitor without making assumptions — a question that exists for exactly this reason.
The underlying question isn't what's happening at the competitor, but which work in your own company can genuinely be taken over by AI; that is mapped out per task by the work scan from FTE TO AI. As a first step in the comparison itself: name where you believe you're winning, and use the free dimension check to see which of those claims can be defended with evidence. The full benchmark, with the evidence matrix per dimension, is under construction.