St. Louis Fed Finds No AI Productivity Bump in 490,000 Earnings Calls

3 articles · Updated · Fortune · Jul 31

A St. Louis Fed study of nearly 490,000 earnings-call transcripts from 5,198 U.S. public firms found AI has yet to produce a measurable aggregate productivity gain, matching three years of official data.
About 15% of productivity commentary was tied to AI by end-2025, but roughly 95% of those remarks described expected future gains rather than benefits already realized.
Serdar Ozkan said AI may still be boosting output in ways statistics miss: when content becomes far cheaper and more abundant, its market value falls, offsetting measured productivity gains.
The paper also points to slower-moving bottlenecks and historical diffusion lags—often 20 to 30 years for general-purpose technologies—suggesting AI's economy-wide payoff may take decades to appear.
Researchers said firms discussing AI positively are increasingly backing that optimism with higher R&D, capital spending and investment, even as the aggregate payoff remains unclear.