Paper Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI

This study examines the early labor market effects of AI chatbot adoption by linking large-scale employer and worker surveys to administrative labor market records in Denmark. The evidence reveals a striking disconnect between the rapid spread of AI tools and their measurable impact on conventional economic outcomes. On the adoption side, the picture is one of swift and widespread uptake: the majority of employers in occupations exposed to AI chatbots had launched related initiatives within two years of ChatGPT's release, workers broadly reported productivity gains, and AI-related tasks had become commonplace across workplaces. Despite this, the study finds no detectable effect on earnings or recorded working hours at either the worker or workplace level. Using a difference-in-differences research design, the authors estimate precise null effects, ruling out impacts larger than two percent over the observation window. In other words, rapid technological adoption has not yet translated into shifts in the aggregate wage or hours distributions. What has changed, however, is the internal structure of work. Employers appear to be absorbing AI primarily through task reorganization — introducing new responsibilities in areas such as content generation, AI oversight, and system integration. Workers who adopt AI chatbots also show a tendency to transition into higher-paying occupations where these tools are more relevant, though this group remains too small to meaningfully shift average earnings. The findings suggest that technological change reshapes the nature of work well before its effects become visible in standard labor market indicators such as earnings and hours.

Get the Paper