How Generative AI Changes Labor Demand: Evidence from 41 Countries

Summary

A Stanford Digital Economy Lab working paper examines how firms’ adoption of generative AI changes labor demand across 41 countries. The researchers develop a model that separates firm-wide productivity effects from worker-type-specific changes in task content. They apply it to 1.25 billion job postings and 154 million employment records, inferring AI adoption from job advertisements that involve generative AI use. An instrumented event study finds that foreign affiliates of adopting firms reduce the junior share of their workforce relative to comparable affiliates. The change appears to be driven mainly by growth in senior employment rather than a decline in junior employment, with suggestive evidence of modest overall employment growth. Senior workers move toward occupations exposed to AI, while the estimates suggest that junior workers shift away from those occupations. The model-based decomposition also provides suggestive evidence of modest productivity gains after AI adoption.