Goldman Sachs estimates that up to 9% of workers—approximately one in 11—could be displaced during the transition to AI. Economist Joseph Briggs derived the estimate from the bank's projection that AI will eventually raise productivity by 15% in the U.S. and other developed economies. Reduced hiring could weigh on employment by an additional 3%. Crucially, the displacement figure measures gross job losses rather than a permanent increase in unemployment: some workers may move quickly into new positions, while others may exit the labor force.
Briggs's historical analysis suggests that the pace of disruption will depend heavily on economic conditions. Companies have typically accelerated automation and workforce reductions during downturns, rather than periods of expansion. The share of Americans employed in routine occupations fell sharply during recessions in the early 1980s, 2001, 2008 and 2020, without subsequently returning to its previous level. Across 25 occupations heavily affected by earlier technologies, employment declined by an average of 25% over five years after the decline began. Briggs considers those cases likely upper bounds for AI because they represent particularly severe disruptions.
Current conditions remain less disruptive. AI-attributed layoffs have retreated sharply from their spring peak, and only 23% of U.S. establishments have integrated AI into regular workflows. Most anticipated productivity gains therefore remain ahead. Meanwhile, economic and labor-market resilience in 2026 may have limited companies' incentives to cut labor costs aggressively. A downturn, Briggs argues, could accelerate AI-related job losses and deepen the broader economic slowdown.
