• Goldman Sachs (GS) estimates AI could displace roughly 9% of the US workforce—about 15 million people—over a 10-year adoption cycle, but the unemployment rate would rise by less than 1 percentage point per year if losses are gradual.
  • The bank projects a 15% productivity boost from AI, which would ultimately lift GDP, but warns that job losses could accelerate during an economic downturn.
  • New IMF research shows hiring demand weakening in AI-exposed occupations, though the effect has stabilized; experts remain divided on the pace of displacement.

A Looming Transition

Goldman Sachs economists have raised eyebrows with a new estimate: artificial intelligence could temporarily displace around 9% of the US workforce—roughly 15 million workers—over an assumed 10-year adoption cycle. The research, led by economist Joseph Briggs, stresses that this is not a forecast of a sharp jump in unemployment. Instead, it reflects cumulative worker displacement as businesses integrate AI and reassign tasks. In a July follow-up, Briggs said the annual increase in the unemployment rate would likely remain below 1 percentage point if job losses are spread across the decade.

The headline figure has been widely misinterpreted. Goldman’s estimate is not that unemployment will rise by 9 percentage points; rather, it means that about 9% of workers may need to find new jobs over the transition. The bank also projects that AI adoption could ultimately boost economy-wide productivity by about 15%, a cumulative gain that would lift GDP. But the path to that higher output could be uneven, with some workers facing prolonged job searches.

Historical Parallels and Caveats

Goldman’s analysis draws on historical patterns. Each 1% increase in technology-driven productivity has typically raised job destruction by 0.5–0.6 percentage points over the following two years, according to the bank’s updated estimates. That relationship underpins the 9% displacement figure. However, the translation from displacement to unemployment is not one-to-one; new hiring and workers finding other roles can absorb much of the shock. The bank notes that the unemployment effect of past technological shifts has usually faded within two years.

Still, the report cautions that a recession would complicate the adjustment. If AI-related layoffs coincide with an economic downturn, displaced workers would struggle to find new jobs, amplifying labor-market weakness. “The risk is that the transition becomes much more painful if it happens during a downturn,” Briggs said in a note to clients.

The “25% of jobs within five years” claim that has circulated in some summaries could not be independently verified and should be treated cautiously. It does not represent a prediction that overall employment will fall by a quarter.

Mixed Signals from the Labor Market

New research from the International Monetary Fund (IMF) adds nuance. An October working paper analyzing over 700 million online job postings across 110 economies found that labor demand declined by nearly 10% in occupations most exposed to AI. However, the effect stabilized even as AI capabilities improved—a sign that the disruption may be less accelerating than some fear. The IMF data measures job-posting demand, not total employment.

Meanwhile, Goldman Sachs’ own third-quarter earnings, released on October 13, showed robust financial performance: net revenue of $18.68 billion, up 23% year over year, and diluted earnings per share of $17.01. The bank’s headcount increased 3% sequentially, reflecting campus hiring—evidence that AI adoption does not immediately translate into falling total employment at the firm itself. CEO David Solomon also moved to quell succession speculation, stating there was “no succession transition or timeline underway.”

A Conditional Forecast

Experts remain divided on the ultimate impact. Neil Thompson of MIT expects uneven and often partial automation, with reliability and implementation costs slowing adoption. Daron Acemoglu, also of MIT, anticipates a relatively small net negative employment effect over the next five years but warns of larger losses later if investment continues to favor replacing workers over complementing them. Goldman’s Briggs, by contrast, argues that even a 5% acceleration in new-job creation could be sufficient to reabsorb displaced workers.

For now, the Goldman report is best understood as a conditional scenario, not an observed job loss. The key variables—adoption speed, productivity gains, and the pace of new job creation—will determine whether the transition is manageable or painful. Investors and policymakers will be watching closely for signs of which path the economy takes.

Update: This article was updated on October 22, 2026, to include the latest IMF research and Goldman Sachs’ third-quarter earnings.