Both Sides Are Performing the AI Layoff

Three quarters of executives admit their company's AI strategy is more for show than substance ^[1]. Nearly half of American workers admit they have exaggerated their own AI fluency to keep up ^[2].

Performative AI use seems to line up with a trust gap: employees are unsure of the organization's AI plan, and unconvinced that leaders understand it themselves. — Law & Order: Criminal Intent

The executive side keeps a ledger. Challenger, Gray & Christmas counted 87,714 AI-related US job cuts in the first five months of the year — more than 2024 and 2025 combined — and AI was the stated reason behind roughly 40 percent of private-sector reductions in May alone ^[3]. Auditing the same stretch, Oxford Economics put AI's actual share of US job losses at 4.5 percent, with market conditions cutting four times as many people, and suspected firms of dressing up ordinary layoffs as a good-news innovation story ^[4]. Wharton's Peter Cappelli points at the gap between the headline and the fine print ^[4].

The headline is, ‘It’s because of AI,’ but if you read what they actually say, they say, ‘We expect that AI will cover this work.’ — Peter Cappelli

The timeline does the rest. Wells Fargo has put an AI efficiency drive atop 23 consecutive quarters of headcount reduction, a run that began years before generative AI was useful to anyone ^[5]. Nvidia's chief executive says such cuts were decided long ago and merely relabeled.

We're increasing our investments in areas like technology, including AI, as well as in advertising, while continuing to execute on our efficiency initiatives, which has resulted in 23 consecutive quarters of headcount reductions. — Charles W. Scharf

Then there are the walkbacks, which run in reverse when the work turns out not to have been absorbed. Commonwealth Bank of Australia reinstated 45 customer-service staff it had declared redundant ^[6]. Ford rehired hundreds of experienced engineers after its automated quality systems failed to deliver, its operating chief admitting the company had leaned on automation without getting the results it needed ^[7]. The beat that explains all of it is a stock chart. Indeed's parent Recruit Holdings hit a two-year low when investors feared AI would kill the job-matching middleman, then more than doubled once the company showed AI inside its product ^[8]. That is the incentive behind every AI label above it: the market pays for the badge, not for absorbed work. The people selling the badge know it is often just a badge. Sam Altman has a word for what his own customers are doing — AI washing, naming AI as the reason for what is really ordinary cost-cutting.

a smaller team using the tools we’re building can do more and do it better. — Jack Dorsey

Nvidia's and DeepMind's chief executives have said much the same ^[9]. Yet the same industry keeps selling the agents those cuts are justified with: OpenAI has staffed up to sell law firms automation of routine tasks and cuts to billable hours ^[10], and a startup of ex-OpenAI and ex-Cognition staff has raised $60 million to build agents that own whole jobs for weeks at a time ^[11]. The same premium the market pays the corner office for the badge is priced into wages. Pay is rising 37 percent in "professionalised" AI roles against 26 percent in "democratised" ones, a gap that gives every worker an incentive to look fluent ^[12]. Citigroup says more than 80 percent of its workforce now uses AI tools — exactly the kind of number that gets padded ^[5]. Fifty-eight percent of American workers fear becoming obsolete, 74 percent in tech and 73 percent in finance, the two professions the new agents target ^[13]. On the hiring desk the fakes meet head-on: employers run AI filters while rejecting AI-sounding résumés, and job postings now draw nearly triple the applications they did in 2017 ^[14]. Forty-four percent of Gen Z workers say they actively sabotage AI rollouts ^[1]. The worker side's dark exit is a loop researchers have now measured. Workers who passively accept AI output feel the tool is doing their thinking, lose confidence and ownership, and defer to it more ^[15].

If the AI solves a problem for you, you don’t think and you don’t learn. — Ethan Mollick

None of this means the gains are fiction. The Bank of England finds adopting firms posting genuine productivity gains, concentrated in software and IT consulting ^[16], and Goldman Sachs and Brookings find adoption so far associated with growth and higher employment rather than mass layoffs ^[17]. AWS's finance chief insists his own shift is toward new revenue and products, not away from headcount ^[18], and JPMorgan added staff in the same quarter Wells Fargo and Citigroup cut ^[5]. The tell is where the gains show up: in the Bank of England's own numbers, vacancies in high-AI-exposure roles fell 15 percent over three years, and administrative and customer-service postings by more than 20 ^[16]. The work is absorbed in vacancies and starting wages, never in severance. Which is why the bottom of the ladder is where to look. In management consulting the pipeline is closing from the ground up: postings fell 26 percent in a year, and entry-level inflows plunged 54 percent, with McKinsey's headcount down 11 percent from 2022 ^[19]. A management-consulting analyst says the firms keep the shrinkage quiet for a reason.

Because you still want to be seen as an aspirational brand for top talent. And if you start to talk about how entry-level hiring is going to decrease in the future, then some of this top up-and-coming talent isn't going to consider you as a potential employer. — Namaan Mian

Starting wages at AI-exposed companies are down 4.5 percent since ChatGPT, and the cut lands mostly on juniors ^[12]. Farthest down, the disappearance is absolute: entry-level remote tech work for refugees in Kenya's Kakuma camp has fallen by roughly half since 2022, as AI absorbed the transcription, translation and annotation that work was built on ^[20]. An experiment that cloned a real manager into an AI for four months got a bot that handled the rules but never pushed back, never disagreed, never carried the institutional context — sycophantic exactly where leadership needs spine ^[21]. The circuit closes at the one place it should not. OpenAI's own finance department now runs its monthly closes on a ChatGPT agent two to three times faster — but it keeps controller sign-offs, and its head of product finance hires only for AI proficiency plus the business judgment to challenge what the AI produces ^[22]. That second requirement is the whole story. Judgment is the one input neither side can fake, and it is the one being quietly destroyed: the executives relabeling old cuts and the workers inflating their fluency are both counting on a skill that is now trained nowhere, because the entry rung where it used to be learned is closing — quietly, while both confessions do the talking.