Goldman: AI could displace 1 in 11 workers — but mostly during a recession
Goldman Sachs raised its AI job-displacement estimate to 9% of the workforce, but says the real wave of layoffs is likely to hit only when the economy next slows down.
What happened: Goldman Sachs economist Joseph Briggs published a note raising his estimate of how many U.S. workers AI could eventually displace to 9%, or about 15 million people, up from a 6-7% estimate in August 2025. But the headline finding isn't the number, it's the timing. Briggs says automation-driven job losses have historically been mild during healthy economies and have instead hit hard and fast during recessions, and he expects AI to follow the same pattern. Right now, AI-linked layoffs tracked by Challenger have actually fallen from their spring peak.
Why it matters: This reframes the AI jobs debate. The absence of mass layoffs today isn't proof AI won't cost jobs, it's proof the economy is still growing, according to Goldman's own analysis. Briggs found that in past tech transitions, routine-job employment held steady in good times but dropped in sharp, permanent steps during downturns like 2001, 2008 and 2020, because companies use slumps as cover to restructure. If that pattern repeats, workers in exposed roles may feel fine for years, then lose jobs suddenly and in bulk whenever the next recession lands.
How it works, plainly: Goldman's model assumes AI eventually lifts U.S. productivity 15%. Historically, each 1-point gain in tech-driven productivity has pushed unemployment up 0.3 to 0.4 points at a peak, and raised gross job losses by 0.5 to 0.6 points, which multiplied out gives roughly the 9% figure. Briggs also looked at 25 past occupations hit hardest by automation, like bank tellers and file clerks: employment in those jobs fell 25% on average over five years once decline began, mostly from people leaving or losing jobs rather than companies simply hiring fewer people.
The rollout: For now, only about 23% of U.S. businesses have built AI into regular workflows, which Briggs says explains why disruption looks modest today. Even Anthropic CEO Dario Amodei, who warned AI could wipe out half of entry-level white-collar jobs, has softened to suggesting workers' roles may expand rather than disappear, while still cautioning AI is moving faster than past technologies and could behave unpredictably. Goldman's own economists have also shifted their estimate twice in a year, from skepticism of very high unemployment claims to warning of amplified job losses in the next downturn.