This paper documents how artificial intelligence has affected stocks and flows in the U.S. labor market. Combining CPS and JOLTS data with AI exposure and adoption measures from OpenAI and Lightcast, respectively, we construct labor force stocks, worker flows, and market tightness by AI exposure and adoption. Since the introduction of LLMs, workers with high AI exposure and adoption have experienced larger declines in job-finding and job-switching rates than other groups, while their within-job activity switching has increased noticeably relative to others. The positive effects of AI on nominal wage growth and hours worked attenuate markedly in the post-LLM period, suggesting that LLM-driven reallocation has operated through within-firm task reorganization and a reconfiguration of labor inputs, alongside weaker demand for workers most exposed to AI. To quantify LLM-driven reallocation pressure, we construct individual-level hirability and separability indices and showthat dispersion in job-finding prospects has risen markedly since 2023, largely driven by the AI factors. The natural rate of unemployment—recovered from the trend components of unemployment inflows and outflows by AI exposure and adoption, as well as from a theoretical model of structural unemployment—is estimated to have risen by about 0.1-0.2 percentage point since the introduction of LLMs, albeit with considerable uncertainty.

That is from a new NBER conference paper by Hie Joo Ahn and Nicholas A. Carollo.  Via Inclusive Productivity Network and Alex Imas.