When AI use expands, what must managers make visible so employees feel secure?
Gallup’s three-year panel found that frequent AI users report job-displacement concern at more than double the rate of infrequent users — and that feeling respected by the organisation is linked to less of it. What a manager has to make visible after AI use expands: the changed role, the owner of exceptions, the support, and the purpose.
A manager check-in after the team’s AI use has grown. The dashboard shows more people using the tools, more often. The check-in covers the numbers.
What it does not cover is which part of each person’s job has changed, who handles the cases the AI gets wrong, and what support someone gets when the new way of working does not yet fit their task. Those three things are not on the dashboard, and they are exactly what people use to judge whether the change is being done with them or to them.
Increased AI use does not, on its own, make employees feel more secure about their jobs. Evidence from a large US panel suggests the opposite pattern — and that what an organisation and its managers make visible matters more than how often the tool is opened.
Does using AI more make employees less worried about their jobs?
No — in Gallup’s data, frequent AI users were more worried, not less.
Gallup’s article “Using AI More Does Not Reassure Workers — Managers Do” (9 September 2026) draws on nearly 30,000 worker observations across four waves, from 2023 through the first quarter of 2026, tracking employees over time. Workers who use AI daily or several times a week reported acute displacement concern at more than double the rate of less frequent users.
Two limits matter before using this. The panel is nationally representative of employed adults in the United States, so it is not EU or Czech and Slovak data. And it describes associations in a panel, not a controlled experiment. It is strong evidence about a pattern, not a law about every team.
What is linked to lower fear of being replaced by AI?
Feeling respected and cared for by the organisation is linked to lower displacement concern, most of all among the heaviest AI users.
In the same data, employees who strongly agree that their organisation cares about their wellbeing are about six to seven points less likely to report displacement concern. Among workers giving the highest respect rating, the association between frequent AI use and displacement concern was 6.8 points smaller.
Gallup also names what managers do in practice: explain the purpose of the change, connect it to what the team becomes able to do rather than to headcount efficiency, and make room for people to raise concerns directly.
None of that is a communication campaign. Each item is a management behaviour that either happens in the weekly work or does not.
What should a manager make visible after AI use expands?
Four things, each tied to the actual work rather than to the tool:
- What changed in the role. Name the specific task: which steps AI now prepares, which steps the person still owns, and which judgement they are now expected to exercise. “Use AI more” is not a role description.
- Who owns the exception. When the output is wrong, incomplete or ambiguous, someone decides what happens next. If that is the person doing the work, say so and give them the authority. If it is someone else, say who.
- What support exists. Time inside the working week to practise on real cases, and a named person to ask. A course link on top of an unchanged workload does not count.
- What the change is for. Say it plainly and truthfully. People compare the explanation with what happens next, so an honest purpose builds more trust than a comforting one that the next reorganisation contradicts.
A usage figure can sit next to these four. It cannot replace any of them.
How do you get employees to actually use AI tools without turning it into a mandate?
Design the change around the work, and let use follow from it.
A target for daily use tells people how often to open a tool. It does not tell them which task is now different, what good looks like, or what happens when the output is wrong. When those are defined — one workflow at a time, with the people who do it — use rises because the tool has a place in the job. When they are not, rising use can come with rising worry, which is the pattern Gallup describes.
This is also why we treat manager behaviour as part of the change design rather than as a soft extra. In our work, the question for each changed workflow is not only what the AI may prepare, but who explains the new role, who owns the exceptions, and how the team practises the new way of working on its own cases before it becomes the default.
What is the useful measure, then?
Measure whether people can run the changed work, not only whether they open the tool.
Can the person describe which part of the task changed? Do they know who owns the exception? Have they practised it on a real case? Did the workflow result — time, quality, rework — move against its starting point? Those questions take longer to answer than a usage chart. They are also the ones that tell you whether the change will hold.