Research update · 12 October 2026
6.1 million workers face a harder fallback if AI displaces them
A 2026 U.S. working paper estimates that 6.1 million workers, or 4.2% of the workforce it covers, are in occupations with both high AI task exposure and low average capacity to recover from a job transition. This is a map of conditional vulnerability, not a count of layoffs or a prediction that those jobs will disappear.
What the study measured
Manning and Aguirre cover 356 occupations representing 95.9% of U.S. employment. They combine an AI task-exposure score with an occupation-level index of net liquid wealth, skills transferable to growing jobs, local labor-market density and worker age. The authors' Brookings account says the high-exposure, low-capacity group is concentrated in clerical and administrative roles. The Brookings article explains the same study; it is not a second replication.
Among 37.1 million workers in the top exposure quartile, 26.5 million are in occupations with above-median estimated adaptive capacity. That comparison does not mean a transition would be easy for any of them. Occupation averages can hide a worker with little savings or few local openings.
What it cannot answer
The study does not observe AI-related layoffs, wages after displacement or whether a specific worker found comparable work. It cannot show that 6.1 million people will lose jobs. Its exposure measure describes potential task change; the capacity index is a proxy for adjustment if job loss occurs. New jobs, local demand and employer choices could change the outcome.
The Copilot hiring studies remain a separate question about observed hiring, with different denominators. The useful next study would link actual employer AI adoption and separations to workers' later jobs and pay, then test whether the proposed capacity factors explain who recovers.