AI could displace one in 11 workers, Goldman says, but the layoffs are likely waiting for a recession

The AI layoff scare that flared this spring has cooled off, but Goldman Sachs thinks the bigger wave may simply be waiting for a recession. Joseph Briggs, one of the bank’s economists, said Thursday that concern about immediate AI job losses has eased after peaking earlier in the year. His view is less comforting than it sounds: if AI eventually boosts productivity by 15% in the U.S. and other developed economies, as Goldman assumes, then as many as 9% of workers could be displaced during the transition, with slower hiring shaving off another 3%. That works out to roughly one worker in 11.

The timing is the point. Briggs dug through decades of technology-driven labor shifts and found a pattern that should make executives and workers equally uneasy: when the economy is healthy, automation disruption stays muted; when growth slows, it gets sharper. He wrote that AI-related labor disruption could end up amplifying the business cycle, not just changing it. Goldman has now made the same basic argument twice, a year apart, through two different teams.

That earlier warning came from U.S. economists David Mericle and Pierfrancesco Mei, who said in October 2025 that “jobless growth” may be the new normal. Their message was blunt: the full labor-market effects of AI might not show up until a recession arrives. Briggs landed in the same place. He said companies tend to automate and cut labor costs when the economy starts to slow, and history backs that up in routine jobs like clerks, secretaries, factory machine operators, mechanics, and bank tellers.

The data behind the note is messy but pointed. Across 25 occupations disrupted by past technology, employment fell an extra 4% over five years for every one-point rise in the national unemployment rate. Briggs also said that in those occupations, separation rates rose much more often than hiring rates fell, which suggests workers are losing jobs faster than firms are merely slowing hiring. He still calls the 9% figure a temporary displacement estimate, not a permanent unemployment forecast. That’s a nice distinction if you’re the economist; it’s less soothing if you’re the worker whose role gets “reallocated.”

For now, the market signal is still mild. AI-linked layoff counts in Challenger’s tracking have dropped from their spring peak, and news coverage of AI job losses is back near the level seen at the start of 2026. Goldman also says only 23% of U.S. establishments have built AI into regular workflows, which leaves a lot of the productivity story still ahead. In other words, the loudest phase of the panic may have passed, but the bill may not be due yet.