A team of ten people in a professional services firm starts using an AI tool to help with case write-ups. The tool was introduced as augmentation. Each person still does the work, and the tool helps them get drafts together faster.

In the first month, the team is mostly relieved. The tool takes care of the parts of the write-up that people enjoyed least.

By month two, the picture is more mixed. The team leader has noticed how much faster the tool makes things, so the volume of cases the team is expected to handle has gone up. Pace has risen across the week. Changed workload has surfaced.

By month three, the day-to-day work has shifted again. The drafts come out of the tool quickly, but they need careful checking. The work is now mostly checking and refining, not producing. Some of the team are good at this. Some are not. Changed skill mix has surfaced.

At the end of month three, the team leader starts looking at the tool’s usage data. How many cases each person processed. How long each one took. Quietly, that data begins to shape how individual performance is talked about. Surveillance drift has started.

The original decision was to introduce the tool as augmentation. Three months in, three different patterns are visible at the same time. None of them was named in the adoption plan. They unfolded as the tool got used.

This is what the patterns look like in practice. They do not always show up at once, and they are rarely predictable from the moment a tool is brought in.

Concern about job security

This is the most common concern, and often the loudest. Even where no replacement is planned or likely, people notice that AI is moving into work that used to be theirs, and they wonder what comes next. People in roles that look more vulnerable from the outside, such as administrative, customer-facing or content-producing roles, often feel this more strongly.

Loss of the meaningful parts of the job

People often did not only value the outputs they produced. They valued the thinking, the craft and the judgement that went into making them. When AI takes over the parts of the job that felt most like their own work, the job can feel empty even when it still exists.

For example, someone whose strength was in producing strong analysis may now feel they are reviewing and tidying AI output rather than doing the analysis themselves. The output looks similar from the outside. The work feels different to do.

Worry that experience no longer counts

This is the worry that years of experience in doing something well are being made irrelevant. Someone who built their career on being fast and accurate at a particular task now sees AI being faster and more accurate. The career path they were on may no longer exist in the same shape, and the skills they invested years in may not carry the same value any more.

Being judged by AI

People increasingly work alongside tools that watch their work and score their performance. The same tools suggest what they should do and produce data on how they compare to others. The worry is that an employer now has an AI-shaped picture of them that they cannot see, cannot challenge and cannot control.

It is one thing to be assessed by a manager who knows the work. It is something else to be assessed by a system whose reasoning is hidden.

Being the last human in the chain

When AI does most of the work and a person is there to catch errors, the work becomes a kind of supervision job. It is less varied than the original job. There is less to do that feels creative, and the moments of decision are mainly about whether the AI got something right rather than what to make from scratch.

People know when their job has shifted in this direction, and they often resist it, because the work itself has lost something they valued.

Having to use tools they do not trust or understand

Sometimes people are asked to use AI tools they did not choose. They may not fully understand how the tool works, and they may not be sure the tool is producing the right output. Even so, they are expected to take responsibility for what comes out of it. If something goes wrong, it is their name on the work, even though the tool produced it.

This is a particular kind of stress. It is not the stress of having too much to do. It is the stress of being held responsible for something they cannot fully check.

A team of six is six months into using an AI tool that was brought in to help write client reports faster. The tool works. The reports go out quicker than they used to.

In a regular meeting with their manager, a team member raises some of what they have been noticing.

They mention that report writing used to be the part of the job they enjoyed most. Now they are checking AI drafts rather than writing from a blank page. They are not asking to go back. But the work has changed shape, and they want it acknowledged that the change is real.

They mention that the tool produces data on how many reports each person edits and how long each one takes. They have not been told that data is being looked at. But they are aware that it exists, and they have started to wonder.

They mention that on a recent report, the AI output included a small but important error that they only caught on the second look-through. Their name went on the final report. If they had not caught it, it would have been their problem.

The line manager listens. None of these are formal complaints. None of them are unreasonable. None of them were in the documents that justified bringing the tool in.

The picture in this conversation is different from the picture in any rollout report. The reports are going out faster. That is true. The team’s relationship to the work has also changed. That is also true. Both things are real at the same time.

A team is eighteen months into using AI tools across most of their day-to-day work. The efficiency gains are real. The team is producing more reports, more analysis and more client communication than it did before the tools were introduced.

From senior leadership, the view is straightforward. The investment has paid off. Output is up. Costs are down. The adoption is held up as a good example for other teams.

Inside the team, the picture is more complicated.

Pace has risen. The new baseline is fast, and people are working at that pace most days. Some of them are finding it hard to switch off in the evenings.

Some people on the team are using AI heavily. Others have quietly resisted, doing more of their own work by hand. The two groups are starting to be seen differently in performance conversations, even though no policy has been written.

A handful of small quality concerns have come up. None of them are big enough to be raised formally. But the team members know they exist, and they know they are the ones whose names go on the final work. The worry sits in the background.

A couple of people on the team have been asked to help bring the tools to other teams in the organisation. They are not deciding what to roll out. But they are the people the rollout happens through, and they end up feeling the discomfort of colleagues elsewhere who are unsettled by the change.

Both views of the adoption are accurate. From the top, the rollout is going well. From inside the team, several culture and wellbeing effects are also happening, even though they do not show up in the main numbers. Neither view on its own is the whole picture. The richer view is the one that holds both at once.