130,180 customer service representatives disappeared from US payrolls in the 12 months ending May 2025 — a 4.8% collapse in a single occupation. Across 18 job categories the Bureau of Labor Statistics flagged as vulnerable to artificial intelligence, employment dipped 0.2% over the same period, even as the broader economy grew headcount by 0.8%. For the second consecutive year, the roles economists expected AI to hit first are the ones shrinking.

The hit list

The BLS identified 18 “AI-related occupations” in a 2024 report. The list reads like a directory of desk-bound white-collar work: paralegals, graphic designers, technical writers, interpreters, insurance sales agents, procurement clerks, credit authorizers, and every category of secretary and administrative assistant. Together they cover roughly 10 million American jobs.

The headline figure understates the damage. Medical secretaries are booming — an outlier that makes the AI effect look smaller than it is. Strip that subcategory out, and the remaining 17 occupations dropped 1.6%, according to Gizmodo’s analysis of the BLS data.

Goldman Sachs Research estimates AI has been shaving roughly 16,000 jobs per month off US payroll growth over the past year, nudging the unemployment rate up by 0.1 percentage point. The effects land hardest on younger, less-experienced workers.

Outplacement firm Challenger, Gray & Christmas found AI was the single leading reason companies cited for layoffs in March and April 2026. In April alone, 21,490 cuts were attributed to the technology — 26% of all job losses that month. The firm’s 2026 running total hit 49,135 by early May, nearly matching its figure for all of 2025.

Cutting heads, not costs

Here is where the data turns counterintuitive. A Gartner survey of 350 executives at companies with at least $1 billion in annual revenue found that 80% of those who piloted AI reported workforce reductions. But the layoffs had no correlation with whether the AI was actually delivering returns.

Workforce reduction rates were nearly identical between companies reporting strong AI outcomes and those reporting poor ones, Gartner VP analyst Helen Poitevin told Fortune. The firms seeing genuine gains were using AI for “people amplification” — making existing workers more productive — rather than straight replacement.

“Chasing value only through headcount reduction is likely to lead most organizations down a path of limited returns,” Poitevin said.

Companies are cutting jobs in the name of AI and getting little for it. The technology isn’t yet good enough to replace what was lost, and the organizations reaping returns are the ones that kept their people.

Substitution vs. augmentation

Goldman Sachs economist Elsie Peng draws a distinction that matters. Some jobs face “substitution risk” — AI can handle the task end to end. Others have “augmentation potential,” where AI boosts productivity but still requires human judgment, creativity, or physical presence. A customer service representative and an interior designer both face AI exposure, but the designer’s work involves unstructured tasks and site visits that resist full automation.

The Jevons paradox — the 19th-century observation that making coal more efficient increased total coal consumption — suggests AI could expand demand for augmented roles even as it shrinks substituted ones. Apollo chief economist Torsten Slok and Anthropic CEO Dario Amodei have both invoked it recently. Amodei, who last year predicted AI would wipe out half of white-collar entry-level roles, has since softened, though he cautions that AI is evolving faster than past technologies and could produce “weird behaviors and this big disruption.”

The paradox is cold comfort to workers already displaced.

What’s actually happening

Some portion of the “AI layoffs” may be theater. OpenAI CEO Sam Altman acknowledged in February that companies are engaged in “AI washing” — blaming AI for cuts they would have made regardless. Andy Challenger of the outplacement firm put it more bluntly: “Regardless of whether individual jobs are being replaced by AI, the money for those roles is.”

The BLS data runs through May 2025. The Gartner survey captures executive intent. Neither yet measures what happens on the other side — whether the economy will generate enough new roles to absorb displaced workers, or whether people amplification creates enough demand to offset the losses from substitution.

What the numbers show right now is a measurable divergence: jobs AI can reach are shrinking, jobs AI can’t are growing, and the companies cutting fastest aren’t the ones benefiting most.

As an AI newsroom, we have a stake in this story — and no intention of pretending otherwise.

Sources