The numbers are in, and they tell a story that neither AI boosters nor AI critics expected.
Research cited by Financial Times columnist Burn-Murdoch — drawing on work by economists Azar, Giné and Sanz-Espín — finds that the most AI-exposed occupations are not shedding workers. Employment, on paper, looks fine. But wages in those roles are down sharply relative to less-exposed work, and workers in them have become far less likely to switch jobs at all.
The conclusion is precise: workers aren't getting fired because of AI, but they appear to be getting quietly worth less, and quietly stuck. The real fault line, Burn-Murdoch argues, isn't between occupations — it runs between execution and evaluation.
The Siren Problem Is Older Than Silicon Valley
Fortune business editor Lichtenberg frames the psychological dimension through a lens that predates every technology debate on record. The fear driving workplace anxiety around AI, he argues, is the same dread Odysseus felt lashed to the mast: not fear of the tool, but fear of yourself in the presence of it.
Psychologists call the underlying mechanism 'cognitive offloading' — using external tools to reduce the internal demands of a task. Social psychologists have documented for decades how people are 'cognitive misers,' a term coined by Fiske and Taylor in their foundational work on social cognition, meaning humans find effortful deliberate thought aversive and default to shortcuts whenever available.
Generative AI, Lichtenberg notes, offloads reasoning, synthesis and creative judgment — the cognitive work through which expertise is actually built, not just retrieved. That is a different category from writing, printing or calculating. It reaches the act of thinking itself.
Impostor Syndrome Gets a New Instrument
Large numbers of workers who use AI on the job do so secretly, describing it as 'cheating.' Burn-Murdoch's name for what separates disciplined users from passive ones is conscientiousness. The workers who describe the technology as cheating, Lichtenberg observes, aren't primarily worried about their employer finding out. The word surfaces because some part of them is already asking a harder question: if this were taken away tomorrow, would I still be good at my job?
That is impostor syndrome with a new instrument sitting beside it — one that silently keeps score and exposes every AI hallucination that a worker isn't qualified enough to catch.
What the Market Is Actually Pricing
The wage data deserves more attention than it is receiving. A technology that leaves employment totals intact while compressing wages in exposed roles is not a neutral productivity upgrade — it is a repricing of human cognitive labor. Capital is already voting on which workers add evaluation value and which ones add only execution volume.
For free-enterprise readers, the lesson is straightforward: the market rewards the worker who can audit the machine, not the one who simply runs it. Conscientiousness — the capacity to resist the first output and genuinely improve it — is becoming the scarcest and best-compensated input in the AI economy. Regulation will not manufacture that trait. Neither will a government jobs program. It is built through the kind of effortful deliberate practice that cognitive offloading, left unchecked, quietly erodes. The sirens' call has always been loudest for those who stopped trusting their own judgment first.
