AI Utilization and Economic Performance

Abstract

Do state-industry cells that report more intensive artificial intelligence (AI) use in 2025–2026 also exhibit stronger economic performance over 2016–2024? We combine a state-by-industry measure of worker-reported frequent AI use from the Gallup Workforce Panel with an industry-level measure of employer-reported AI use from the U.S. Census Bureau’s Business Trends and Outlook Survey (BTOS), pooling observations from Q2:2025 through Q1:2026. We link these measures to annual employment, earnings, and real-output data from the Longitudinal Employer-Household Dynamics program and the U.S. Bureau of Economic Analysis and estimate retrospective dynamic specifications with state×industry and industry×year fixed effects. State-industry cells with higher worker-reported AI use exhibit stronger post-2020 real-output paths and positive, though imprecisely estimated, employment differences. The industry-only BTOS specification does not recover the same output pattern. These estimates are descriptive rather than causal and underscore the value of granular measures of realized AI use.