AI anxiety at work — what it is tracking, and why reassurance slides off it

What this one is actually like

Most workplace worry has an event attached to it. A deadline, a review, a restructure with a date on it. You know what you are worried about and roughly when you will stop.

This one has no event. Nothing has happened. Your job is intact, your manager has not said anything alarming, and by every visible measure the situation is fine — which is precisely the problem, because you are not worried about anything visible. You are worried about a decision that may already have been made in a meeting you were not in.

There is a second thing that makes it worse, and it gets discussed less. You cannot raise it at work. Saying “I’m worried AI is going to make my role redundant” in front of your manager reads as either a request for reassurance or an admission that you think you are replaceable, and neither is a good look. So the one place holding all the relevant information is the one place you cannot ask. People take the question home and give it to a search engine instead.

It tracks ambiguity, not capability

Here is the thing that surprised me when I started collecting these situations. The people most wound up about this were frequently not the ones at the most automated employers.

Someone at a firm that has genuinely deployed AI across their function tends to be oddly calm about it. Not happy, necessarily — but the shape of the thing is known. They have seen what it does and does not absorb, the headcount question resolved one way or another, and they are dealing with a fact rather than a possibility.

The wound-up ones were often at companies that had announced a great deal and shipped almost nothing. Eighteen months of strategy decks, a task force, a pilot in one region that nobody talks about anymore, and no statement of any kind about what it means for the people doing the work. Maximum signal, zero resolution. That is where the pressure sits.

If anxiety tracked capability, this would be backwards. It tracks ambiguity.

Why the usual three responses do not work

Reassurance comes first, usually from someone who means well. They tell you the technology is overhyped, or that human judgement will always be needed. It lands for about a day. It slides off because it answers a general question about technology and yours is a specific question about an employer. Unfounded reassurance is also expensive later — the people who took “your job is safe” at face value in 2023 were not helped by it.

Doom-reading is the mirror image, and it can pass for diligence. You are researching, after all. But consuming more forecasts about a question that forecasts do not determine is just repeated exposure to the thing making you uneasy, and there is enough of it out there that you will always find something to confirm whichever mood you arrived with.

Panic-upskilling is the most respectable of the three and the hardest to argue with, because doing something beats doing nothing. Usually. Learning the tools is fine and often useful. It stops being useful when it is aimed at the wrong variable: if your exposure comes from not being in the room where decisions get made, no amount of tool fluency puts you in the room. People spend months on the lever that was easiest to reach rather than the one attached to their situation.

The common failure is the same in all three. They act on the feeling instead of on what the feeling is pointing at.

What actually moves it

Ambiguity does not have to be resolved in general. It has to be resolved in the few specific places that are carrying weight for you.

That is a much smaller job than it sounds. The unknowns producing this pressure are usually a handful: what your employer can actually deploy rather than intend to, what management has said and not said, whether the last change produced capability or theatre, whether you are inside or outside the decision, and how regulation cuts in your particular function. Most people can answer four of those from things they have already seen, and have simply never sat down and written them out.

What changes is not that the answers are good. Sometimes they are not. What changes is that you are dealing with a defined situation rather than an undefined one, and a defined bad situation is considerably easier to carry than an undefined ambiguous one. People consistently underestimate this, and it is the most reliable effect I have seen.

Where this stops being an information problem

Some of it will not be touched by any of the above, and it is worth being straight about that rather than tacking on a disclaimer.

A test that works reasonably well: does the feeling move when the facts change? If you learn something concrete about your employer’s plans and the weight shifts, you were dealing with uncertainty and clarity will help. If it does not shift, or if the feeling was present before AI became the subject, or if it is there on Sunday morning with no work in sight, then it is not tracking your job. Nothing you learn about your organisation will reach it, and spending months trying is the wrong use of the effort. That is a different problem, it is a common one, and it has its own kind of help — just not this kind.

Why the tool is called Depressure

It is an odd word and people ask.

The pressure in question is the ambient kind that comes from not knowing where you stand while everyone around you talks confidently about a transformation. Lowering it is not the same as being reassured. Reassurance lowers the feeling and leaves the situation exactly as it was, which is why it wears off by Tuesday. Removing an unknown lowers the pressure by changing what you are actually dealing with.

That is the whole idea, and the whole limit of it. It will not tell you that you are fine. It will tell you where you stand, including when where you stand is not good — which turns out to be the version people can do something with.