A study by economists Hee-Joo Ahn and Nicholas Carollo, presented at a conference of the U.S. National Bureau of Economic Research (NBER), showed that the proliferation of large language models (LLMs) has affected the U.S. labor market, leading to changes not only in hiring processes but also in the redistribution of responsibilities within companies.
According to the researchers’ estimates, the natural rate of unemployment in the country may have risen by 0.1–0.2 percentage points following the introduction of LLMs, though this figure remains uncertain. Economists, including Tyler Cowen, were intrigued by how data from the CPS and JOLTS surveys align with AI estimates from OpenAI and data from Lightcast.
It was found that workers most affected by AI began looking for new jobs less frequently after 2023; instead, they were more likely to switch roles within the same company. This suggests that workforce reorganization has become the primary consequence of the adoption of large language models.
The study’s authors also noted that the positive impact of artificial intelligence on wage growth and the number of hours worked has become less pronounced. This may indicate a weakening of demand for workers whose jobs are most susceptible to automation.
To assess the scale of these changes, economists developed hiring and firing indices, which showed an increase in disparities in employment prospects after 2023. A significant portion of this relationship, they say, is linked to the expanding use of AI, although estimates of structural unemployment remain uncertain.
The number of people using AI at work continues to grow, and most of them say the technology helps them work more productively; however, some workers express concern about the likelihood of layoffs.
