AI Obsolescence Anxiety in High Achievers

The fear of becoming obsolete is usually described as being about the job. For someone whose sense of worth was quietly built on being the most capable person in the room, it is about something else. An earlier version of this article made that case with survey figures that should not have been used, and without the two experiments that actually bear on it.

What Had to Come Out First

An earlier version of this article reported that by early 2026 roughly four in ten workers named AI-driven obsolescence as a primary career fear, that the share had nearly doubled in a year, and that a majority believed AI would make the workplace feel less human.

Those figures came from a survey published by a consultancy that sells advice on artificial intelligence, reported onward by a magazine. This publication does not use numbers that a company put into the world to market itself, and that applies to the reference list as well as the body text. The figures are gone, and no replacement figure is offered, because none was looked up.

Also removed: a claim that the skill demands of AI-exposed roles are shifting dramatically faster than a year ago. No source for it was found.

So this article no longer says how many people feel this. It does not know.

The Acronym, and What It Names

There is a term circulating for it: FOBO, the fear of becoming obsolete. The World Economic Forum used it in a 2023 piece, which establishes that the word is in use and nothing more than that.

What it points at is distinguishable from the fear of being dismissed. It is the prospect of becoming irrelevant while still employed: the same chair, the same title, and the gradual discovery that the thing you were uniquely good at can be done faster and without you.

Two Experiments That Say Something Checkable

Here the research is genuinely useful, and it sharpens the article’s claim rather than confirming it.

Brynjolfsson, Li and Raymond (2025) studied the staggered rollout of a generative AI assistant in the Quarterly Journal of Economics, “using data from 5,172 customer support agents”. Productivity rose “by 15 percent on average, with substantial heterogeneity across workers”. The distribution of that gain is the part worth reading twice: “Less experienced and lower-skilled workers improve both the speed and quality of their output while the most experienced and highest-skilled workers see small gains in speed and small declines in quality.”

Noy and Zhang ran a preregistered experiment assigning “occupation-specific, incentivized writing tasks to 444 college-educated professionals”, randomly exposing half to a chatbot. Time taken fell by 0.8 standard deviations and output quality rose by 0.4. And: “Inequality between workers decreases, as ChatGPT compresses the productivity distribution by benefiting low-ability workers more.” The work was published in Science in 2023; the version opened here is the working paper, which carries the note that it was not yet peer reviewed.

What both findings describe is the compression of an advantage. Not the disappearance of expertise, and not the obsolescence of the expert. The gap between the best and the rest narrows, which is a different and more specific thing than the article was claiming.

The limits matter and are easy to state. One study covers customer support agents at a single company; the other covers writing tasks in an online experiment. Neither studied executives, and neither studied identity. Both measured output.

One further detail from Noy and Zhang, because it fits the subject exactly: exposure to the tool “increases job satisfaction and self-efficacy and heightens both concern and excitement about automation technologies”. Both at once, in the same people.

Why This Might Land Harder Higher Up

From here on this is a reading rather than a finding. No study cited above establishes who feels most threatened, and the two that measured anything measured productivity.

Deci and Ryan identified competence as one of three basic psychological needs. That is a claim about everybody. What distinguishes the case this article is about is narrower: for some people, competence has also become the route by which worth gets established, which is a different job from the one a need does.

Crocker and Wolfe described that arrangement in Psychological Review in 2001 as contingencies of self-worth, the specific domains in which a sense of being worthwhile is staked on how things turn out. In those domains, they write, “people typically have the goal to prove themselves rather than the goal to learn, grow, or improve themselves”. They describe it as “a trade-off between increased motivation and increased emotional vulnerability”.

Put the economics next to that and the shape is clear enough. If the advantage of being the most capable person in the room is being compressed, and your worth was staked on holding that position, then a tool that helps a junior colleague produce a competent draft is not a change in the tooling.

Three Threats Wearing One Coat

1. The work stops being proof

For a long stretch, being the person with the answer was both the job and the evidence. Producing a competent answer quickly was proof of something: intelligence, diligence, seriousness. When that output is cheap, it stops functioning as proof, and whatever the proof was supporting has to stand on something else.

2. Scarcity, not effort

Standing built on scarce expertise depends on the scarcity, not on how hard the expertise was to acquire. This is where the two experiments above apply most directly: what they measured was precisely the narrowing of a gap. Twenty years of accumulated judgment still exists. What has changed is how far ahead of the next person it puts you on a given task.

3. Mastery stops working as a coping strategy

The usual way of handling this kind of anxiety is to master the domain: more preparation, more expertise, more control. A domain that keeps moving does not reward that strategy, and what remains when the strategy stops working is the anxiety it was managing. That is a reading, not a measurement.

How It Shows Up

Descriptions, offered because they are recognisable. Nobody has measured their distribution.

Working longer to re-establish the point. Effort that the task does not require, serving to demonstrate daily that you still matter.

Private use, public scepticism. Using the tools quietly while dismissing them in meetings. The secrecy is the informative part.

Loud dismissal. Insisting the technology is overhyped. Sometimes that is accurate judgment, and sometimes it is doing a second job.

Joyless upskilling. Learning driven by dread rather than interest, which tends to be learning that does not stick.

What Is Actually True Here

An earlier version of this article claimed that the things which “have always distinguished exceptional leaders are stubbornly resistant to automation” and listed four of them. No evidence was offered for that, and the word “always” was doing unearned work. It has been removed.

What can be said: the two experiments measured task output. Accountability for a decision, and the willingness to put a name to it, were not among the things measured, and no result here tells you they are safe.

An earlier version also wrote that the panic is “old information about you” and that “the wound is decades old”. That is a statement about a reader’s history that this publication cannot make, and it is gone. What stands without it is a question rather than a diagnosis: whether anything would be left of your sense of your own worth if the output stopped being scarce.

What to Do With It

None of this has been tested as an intervention.

Separate the competence from the worth

Slow, and the only one of these with a mechanism behind it rather than a result. Crocker and Wolfe’s point is that the domains are specific, which means they can be named before a technology names them for you.

Reorient from having answers to choosing among them

The two experiments found gains concentrated in drafting and speed. Noy and Zhang describe the tool as restructuring tasks “towards idea-generation and editing and away from rough-drafting”. Whether that reorientation protects a career is not something either study tested; it is where the measured effects point.

Say it to someone

An earlier version recommended a coach or therapist and claimed professional support “can make the difference between rumination and resolution”. The recommendation is gone, because this publication does not point readers at other people’s offerings and has no basis for a claim about what they deliver. The plainer version: a fear held silently does not get corrected by anything.

And if the anxiety has stopped looking like a question about work and is affecting how you sleep, eat or function, that belongs with a licensed professional. Nothing here substitutes for one, including the assessment below.

The Mirror, Not the Threat

Treating the technology as the antagonist is tempting and probably mistaken. The question it forces into the open was available beforehand: what is left of me when what I produce is no longer scarce?

That question is worth answering for its own sake, and the two experiments above suggest it is not hypothetical. What the answer is depends on where the worth was staked, which is a matter of fact about you rather than a matter of opinion about the technology.

One response to that fear is to hold on tighter, and staying in a role because leaving feels dangerous carries its own cost.

Where Your Worth Is Currently Resting

The Executive Identity Profile is a confidential assessment of where your sense of worth currently rests across five dimensions, and which of them is carrying more than it should. It takes about ten minutes, and it opens a conversation rather than closing one. Begin your profile here.

Continue Reading

→ The Executive Confidence Gap: Why High Achievers Feel Less Certain Than They Should
→ The Sunday Dread: What It Really Means When Successful People Fear Monday
→ The Need to Be Needed: When Your Leadership Becomes a Dependency

References

Brynjolfsson, E., Li, D., & Raymond, L. (2025). Generative AI at work. Quarterly Journal of Economics, 140(2), 889–942.
Noy, S., & Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Science. The version consulted here is the 2023 working paper.
Crocker, J., & Wolfe, C. T. (2001). Contingencies of self-worth. Psychological Review, 108, 593–623.
Deci, E. L., & Ryan, R. M. (1985). Intrinsic Motivation and Self-Determination in Human Behavior. Plenum.
World Economic Forum (2023). “Is AI making you suffer from FOBO? Here’s what can help.” It is listed only because it shows the term was in use.

The Obsolescence Anxiety: What AI Is Really Threatening in High Achievers