- Goldman’s 9% estimate applies to the United States and represents workers changing positions over 10 years, not 15 million permanent job losses. [1] - The model assumes a 15% productivity lift after full AI adoption and applies historical evidence on technology-driven displacement. [1] - Goldman says AI is currently reducing U.S. monthly payroll growth by about 10,000 to 15,000 jobs in a limited group of sectors. [1] - MIT economists Neil Thompson and Daron Acemoglu say adoption costs, reliability, task composition and new-job creation make the longer-term effect highly uncertain.
[1] Goldman Sachs estimates that artificial intelligence could displace about 9% of the U.S. workforce—roughly 15 million workers—from their current positions over a 10-year transition. The forecast describes workers finding new jobs; it does not predict 15 million permanent job losses. [1]
The estimate appears in Goldman Sachs Research’s July 2, 2026 report, “An AI Job Apocalypse?” Joseph Briggs, who leads the firm’s global economics team, said the calculation combines a projected 15% productivity increase after broad AI adoption with historical evidence on how technology-driven productivity gains affect employment. [1]
The headline number is U.S.-specific #
The July report and its accompanying discussion focus on the U.S. labor market. Briggs identifies current effects in technology, management consulting and graphic design, then applies the longer-term estimate to U.S. workers. The report does not present the 15 million figure as a forecast for all developed economies. [1]
Goldman has a separate, broader framework covering the United States and other developed markets. That analysis examined more than 800 occupations and assessed task repetitiveness, the consequences of errors, the connections among tasks, and the value of AI-exposed work relative to prevailing wages. [2]
Goldman’s higher-risk examples include computer programmers, accountants and auditors, legal and administrative assistants, customer-service representatives, telemarketers, proofreaders and copy editors, and credit analysts. The bank’s framework also identifies occupations such as air-traffic controllers, chief executives, radiologists, pharmacists and members of the clergy as relatively less exposed. [2]
Those are exposure assessments, not observed job-loss counts. A job can contain automatable tasks while still requiring human oversight, judgment, client interaction or work that is not yet economical to automate. Goldman’s earlier analysis put potential displacement at roughly 3% to 14% under different assumptions. [2]
A small effect so far, according to Goldman #
Briggs estimated that AI is currently creating a drag of about 10,000 to 15,000 jobs on monthly U.S. payroll growth when several affected sectors are combined. He described the shock as narrow and said it has not yet produced a major effect across the broader economy. [1]
A separate April Goldman analysis estimated that AI had reduced monthly payroll growth by roughly 16,000 jobs over the prior year and raised the unemployment rate by 0.1 percentage point. That analysis distinguished between substitution, in which AI replaces labor, and augmentation, in which AI makes workers more productive and can increase demand for their services. [3]
The 15 million figure is therefore a forward-looking cumulative estimate, not a tally of jobs already lost. Briggs said that even if 9% of workers were displaced, the annual unemployment increase could remain below one percentage point if the transition occurred gradually. [1]
Goldman’s 2025 research also found that adoption remained limited: in one U.S. survey, 9.3% of companies said they had used generative AI in production during the previous two weeks. The firm reported no significant statistical relationship between AI exposure and broad labor-market measures at that stage. [4]
Economists question the speed of adoption and reabsorption #
MIT economist Neil Thompson said AI capability is only one step in the automation process. Employers also need access to the right data, reliable systems and a cost-effective way to integrate them into production. Because most jobs contain many tasks, he expects partial and uneven automation rather than entire occupations disappearing at once. [1]
Thompson said the effect on workers will depend on which tasks are automated. Removing routine work may leave a worker with more valuable responsibilities; automating the most expert part of a job could instead reduce wages or change who can perform it. He said it remains too early to know whether new tasks and productivity gains will outweigh disappearing work. [1]
Daron Acemoglu, also of MIT, offered a more cautious near-term view. He said AI could produce limited net job losses over the next five years—possibly below 2% to 4%—because reliable, easy-to-use applications have not yet spread broadly among large employers. He identified routine cognitive work, including customer service and back-office tasks, as especially vulnerable and warned that inequality could increase if displaced workers move into lower-paid jobs. [1]
Other early evidence also cautions against treating exposure as proof of causation. Anthropic researchers found no systematic increase in unemployment among workers in highly exposed occupations since late 2022, although they found suggestive evidence that hiring of younger workers had slowed in exposed fields. [5]
Companies mentioned #
Further sources #
[[1] Goldman Sachs Research’s July 2, 2026 Top of Mind report and accompanying trans… ↗](https://www.goldmansachs.com/insights/goldman-sachs-exchanges/how-will-ai-impact-the-labor-market)
[[2] Goldman Sachs’ global workforce analysis covers more than 800 occupations, desc… ↗](https://www.goldmansachs.com/insights/articles/how-will-ai-affect-the-global-workforce)
[[3] Goldman Sachs’ April 24, 2026 analysis estimates a roughly 16,000-job monthly d… ↗](https://www.goldmansachs.com/insights/articles/the-jobs-ai-is-likely-to-boost-and-those-it-may-disrupt)
[[4] Goldman Sachs reports the 9.3% production-use survey figure and says it found n… ↗](https://www.goldmansachs.com/insights/articles/how-will-ai-affect-the-global-workforce)
[[5] Anthropic’s March 5, 2026 labor-market study reports no systematic unemployment… ↗](https://www.anthropic.com/research/labor-market-impacts)
The stories that matter, in one email. Free — unsubscribe anytime.