Goldman Sachs Warns AI Job Displacement Could Leave Workers Behind

Losing a job to technology may carry a cost that lasts well beyond the first period of unemployment. That is the warning in Goldman Sachs research now drawing fresh attention as employers adopt artificial intelligence. In a 3 November report on a newer Goldman Sachs note, Axios said workers assessed as most likely to have been displaced by AI in 2025 took about a month longer to find another job than other laid-off workers. That is a concerning signal, but it is not a definitive count of jobs eliminated by AI.

The distinction matters. Researchers cannot inspect every redundancy decision and reliably mark it as caused by a chatbot or an automation system. According to Axios, the newer analysis estimates which cases were most likely AI-related using the worker’s occupation, adoption of AI in the industry and the circumstances of the job loss. It examined more than 15,000 workers who lost jobs between 2016 and 2025. The results point to a potential pattern, not a controlled experiment proving that AI alone produced every measured difference.

The risk is more than a slower job search

Axios reported that the workers classified as most likely to have been displaced by AI were roughly 10 percentage points more likely to remain unemployed a year after losing their jobs. Among those from AI-exposed jobs who found new work in 2025, about three-quarters moved into occupations with similar or greater exposure to AI. Finding another role may therefore not remove the risk of a second disruption.

This latest note builds on Goldman Sachs’ September analysis of earlier waves of automation. That earlier work found that people displaced by technological change historically took roughly one month longer to find employment than people leaving stable or growing occupations. It also found greater real earnings losses on re-employment. The September findings concern automation broadly, including changes long before today’s generative AI boom. Presenting them as a direct measurement of AI layoffs would go too far.

There is a plausible reason why the damage can persist. A worker may lose not only a job but also the market value of the skills used in it. Goldman Sachs says some displaced workers move into more routine, lower-complexity occupations. A new job then becomes a partial recovery rather than a return to the same career trajectory. The bank’s researchers also found that workers who undertook at least four weeks of retraining within three years of displacement tended to experience better long-term wage growth and less repeat unemployment than peers who did not.

What the evidence does not say

The findings do not establish an economy-wide AI jobs collapse. A Stanford Digital Economy Lab study, revised in August, found no evidence of widespread displacement across the economy. It did find weaker employment among younger workers in AI-exposed occupations, driven mainly by reduced hiring rather than a surge in dismissals. The authors described those results as descriptive early indicators, not causal estimates.

That mixed picture makes careful reporting important. A company can reduce staff while also investing in AI, but that does not prove every position was replaced by a model. TechBooky’s reporting on Barclays’ planned job cuts made that distinction. Equally, the absence of a dramatic unemployment spike would not mean affected workers are doing well. Lost income, longer searches and repeated displacement can be serious even when national employment figures appear stable.

For employers and policymakers, the useful question is how to measure the transition honestly and support people whose skills are losing value. That could mean tracking the quality and pay of new jobs, not merely whether former staff find any work. It could also mean retraining before roles disappear, when workers still have income and time to move. AI may create new work as it changes old work, but those opportunities will not necessarily reach the same people at the same speed.