Goldman Sachs expects artificial intelligence (AI/SI) to displace about 9% of the U.S. workforce, or roughly 15 million workers, over a 10-year adoption cycle, according to a research report published June 25.
The figure, from the bank’s updated Top of Mind report “An AI Job Apocalypse?,” is a sharp upward revision from earlier estimates of 6% to 7% displacement.
Joseph Briggs, a senior global economist at Goldman Sachs, said the 15 million figure represents workers who will leave or be pushed out of their current positions and need to find new jobs.
“9% of workers being displaced by AI would correspond to 15 million workers,” Briggs said on the bank’s Exchanges podcast, comparing the scale of the shift to the tech-driven upheaval of the late 1990s and early 2000s.
The report’s baseline forecast assumes AI will deliver a 15% productivity uplift in the U.S. and other developed economies. During the transition, Goldman estimates frictional unemployment could rise by 0.5 to 0.6 percentage points.
Briggs said the impact on the unemployment rate in any given year would likely be less than one percentage point if job losses are spread over a decade and most displaced workers find new employment within a year.
Exposure varies widely by occupation. Goldman identifies computer programmers, accountants and auditors, administrative assistants, and customer service representatives as among the most exposed to AI substitution.
Air traffic controllers, radiologists, pharmacists, and chief executives face far lower risk.
The report notes that AI is already reducing monthly payroll growth by 10,000 to 15,000 jobs in sectors such as tech, management consulting, and graphic design.
Goldman’s economists argue the displacement will be temporary. Briggs said 85% of job growth over the past 80 years has come from the technological creation of new positions, and the U.S. labor market already churns about 30 million jobs created and 29 million destroyed each year.
Even a 5% acceleration in the pace of job creation would be enough to reabsorb the workers AI displaces, he said.
“Despite our expectation that AI-related job losses will lead to a meaningful amount of labour displacement, we continue to expect that labour market headwinds will be temporary,” Briggs said.
Separate Goldman research tracking workers displaced from technology-disrupted occupations since 2025 found they took about a month longer to find new employment than other displaced workers and faced a higher risk of lingering unemployment.
The bank’s April analysis of 40 years of labor market data also found real earnings losses of more than 3% upon reemployment for tech-displaced workers.
Goldman also warned that a recession could accelerate corporate AI adoption and deepen automation-related job cuts.
“Automation-related labor disruption is limited when the overall economy is healthy but accelerates during cyclical slowdowns,” Briggs wrote in an October note. “AI-related labor disruptions could well amplify the broader business cycle”.
The report includes contributions from MIT economists Daron Acemoglu and Neil Thompson, who offer more cautious assessments.
Acemoglu expects AI to have a modest negative effect on employment over the next five years but warns job losses could grow if companies prioritize replacing workers over augmenting them.
Thompson argues that adoption will be slower than AI’s capabilities suggest, citing deployment costs, data availability, and reliability problems. Most jobs, he said, will be partially automated rather than eliminated.
Goldman’s central conclusion is that the U.S. labor market can absorb the shock if new job creation keeps pace.
The bank’s forecast of a 15% productivity gain depends on that assumption holding. Whether it does, Briggs said, depends on how quickly busio work and whether they use it to complement employees or replace them.
