NBER Study Analyzes Microstructure of AI Diffusion in US Firms
A new National Bureau of Economic Research working paper examines the diffusion of artificial intelligence across U.S. firms using data from the 2026 Census Bureau’s Business Trends and Outlook Survey. The study reveals that while 18% of firms adopted AI in at least one function between late 2025 and early 2026, usage is heavily concentrated in large enterprises within knowledge-intensive sectors like Finance and Professional Services. Adoption remains narrow in scope, with most firms limiting AI to three or fewer functions, primarily Sales, Strategy, and IT. Worker-level usage is distinct from firm-wide adoption, focusing on tasks such as writing and document analysis. Crucially, the research indicates that broader AI integration correlates with improved firm performance. However, while functional deployment and operational investment are linked to employment declines, direct worker-task use does not significantly reduce jobs when other factors are controlled. The findings suggest a complex diffusion pattern involving both top-down strategic implementation and bottom-up worker initiation, with task augmentation being the dominant use case rather than labor replacement.
Wire timeline
NBER Study Analyzes Microstructure of AI Diffusion in US Firms
A new National Bureau of Economic Research working paper examines the diffusion of artificial intelligence across U.S. firms using data from the 2026 Census Bureau’s Business Trends and Outlook Survey. The study reveals that while 18% of firms adopted AI in at least one function between late 2025 and early 2026, usage is heavily concentrated in large enterprises within knowledge-intensive sectors like Finance and Professional Services. Adoption remains narrow in scope, with most firms limiting AI to three or fewer functions, primarily Sales, Strategy, and IT. Worker-level usage is distinct from firm-wide adoption, focusing on tasks such as writing and document analysis. Crucially, the research indicates that broader AI integration correlates with improved firm performance. However, while functional deployment and operational investment are linked to employment declines, direct worker-task use does not significantly reduce jobs when other factors are controlled. The findings suggest a complex diffusion pattern involving both top-down strategic implementation and bottom-up worker initiation, with task augmentation being the dominant use case rather than labor replacement.
National Bureau of Economic Research Working Papers