Empirical Evidence on Generative AI's Uneven Economic Impact

Summary

This article examines the growing empirical literature on generative AI's economic impact, noting that randomized trials and adoption surveys now cover customer support, software development, legal analysis, writing, education, and firm-level outcomes. The key finding is that task-level gains from AI are large but heterogeneous—AI improves output quality but unevenly. The article highlights that adoption is uneven, complementary investments matter significantly, and the workers most exposed to AI's substitution potential are different from those benefiting from its augmentation. The central challenge is understanding the gap between task-level results and economy-wide outcomes.

Source

Twitter / XEmpirical Evidence on Generative AI's Uneven Economic Impactaei.org