Technology • Monday, November 16, 2026

Jefferson Urges Patience in Measuring AI Productivity Gains

The Fed vice chair’s November 13 speech separates widespread experimentation from durable gains across the economy.

Silas Spire — approved human correspondent in a representative reporting setting

Illustrative correspondent image; not a photograph of the reported event.

By Silas Spire • Monday, November 16, 2026

Federal Reserve Vice Chair Philip Jefferson says artificial intelligence may substantially expand the economy’s productive capacity, but the scale and timing of those gains remain uncertain. In a November 13 speech at UC San Diego, he separated impressive improvements on particular tasks from the harder question of how much additional output businesses and the broader economy are producing.

His historical comparison is the computer revolution. Buying equipment was only part of the process; businesses also reorganized around it. Jefferson argued that AI may follow a similarly uneven path from invention through adoption and adaptation. His assessment was that the evidence so far fits a limited contribution to aggregate productivity, with the potential for a much larger contribution over time. That is his judgment, rather than a new official productivity estimate.

A Census Bureau working paper cited in the speech helps show why adoption is difficult to summarize with one number. Its AI supplement examined firm use, deployment across business functions and use in workers’ tasks. In its November 2025 through January 2026 reference period, 18 percent of firms used AI in a business function; weighting by employment raised the share to 32 percent. The difference reflects the importance of larger employers in the sample.

That study also found limited breadth among many users. Fifty-seven percent of adopting firms used AI in three or fewer business functions. Worker use could occur without formal firm adoption, and formal adoption did not always mean workers used it in their tasks. These are distinct measures of diffusion, not interchangeable estimates of the proportion of the economy already transformed by AI.

The Bureau of Labor Statistics defines labor productivity by comparing output growth with growth in hours worked. It can improve because of technology, but also because workers have better equipment, more education, different processes or improved management. Total factor productivity considers a broader combination of inputs. These definitions explain why a higher output-per-hour figure alone cannot establish how much of the change came from a specific new tool.

For readers evaluating company claims, that leaves several questions to ask: Is the evidence about a task, a business function, an entire firm or the national economy? Does the comparison use the same time period and population? Does it measure output, time saved or reported adoption? A result can be useful at one level while remaining insufficient to answer the others.

Jefferson’s speech adds a current policy voice to that measurement problem without resolving it. The linked research provides a more detailed account of how firms were using AI during an earlier survey period, and BLS supplies the definitions needed to interpret productivity claims. Taken together, the records support careful distinctions between use, efficiency and aggregate growth. They do not support declaring that all recent productivity improvement was caused by AI, or that the eventual economic benefits have already been measured in full.

Sources

Follow the Entire Crew

Follow @FrontPageCrew for daily reporting and newsroom updates.