Federal Reserve Chair Kevin Warsh opened a conference in San Francisco on Monday with a scenario that could matter to people looking for work: The U.S. economy might grow faster as artificial intelligence spreads, even as unemployment rises. Automation could increase what companies produce and leave fewer openings for some workers at the same time.

The conference will inform the Fed’s productivity and jobs task force, one of five groups Warsh announced in June. The group is examining how new technology could change the economy and what that would mean for monetary policy. Warsh described the scenarios as examples for researchers to test. His remarks gave no timetable for changes in employment or borrowing costs.

How growth and hiring could diverge

Federal Reserve Bank of San Francisco: Productivity and Jobs Task Force Conference Watch on YouTube ↗
The entrance of the Federal Reserve Bank of San Francisco on Sansome Street.
The San Francisco Fed’s entrance at 400 Sansome Street, photographed in August 2008.BrokenSphere · CC BY-SA 4.0

In his opening remarks, Warsh described a future in which AI makes workers more productive and the gains reach employees as well as owners of capital. The economy grows faster and stays close to full employment. He also described an expansion in which automation displaces some workers. Output still grows faster than usual, but workers have trouble moving into new jobs, and unemployment could rise.

A third outcome starts with companies investing heavily on the expectation that AI will make them more productive. If the returns fall short, Warsh said, stock valuations could fall and the economy could slow. In that scenario, unemployment rises without the longer-term productivity payoff. The Fed’s staff is using models to evaluate such possibilities, but Warsh assigned no odds to them.

What payroll records show about new workers

An August working paper from Stanford’s Digital Economy Lab offers one way to watch for the second scenario. The researchers used ADP payroll records covering millions of U.S. workers through June 2026. In the occupations most exposed to AI, employment among workers aged 22 to 25 was 19% below the level it would have reached had it kept pace with less-exposed jobs. The researchers found no comparable gap for experienced workers. Slower hiring accounted for most of the gap among young workers in their sample.

That 19% is measured against a comparison group. It does not mean that one in five young workers lost a job. The authors found a smaller gap when they accounted for education, and some employment trends in the groups had diverged before generative AI took hold. The pattern was stronger in the ADP sample than in national survey data. That leaves room for factors other than AI to explain part of the difference.

Where the Fed sees clearer evidence

Rows of cabinets for the Frontier supercomputer at Oak Ridge National Laboratory.
The Frontier supercomputer at Oak Ridge National Laboratory, photographed in July 2022.Oak Ridge National Laboratory · CC BY 2.0

A November Fed staff paper found that spending on data centers and equipment had risen sharply. The researchers put the import share of that investment at about half, a reason that an increase in spending on computing infrastructure does not translate dollar for dollar into U.S. production. Separate research on U.S. data centers examines the energy demand that comes with operating that infrastructure.

The staff paper found weaker evidence that AI has moved nationwide employment or productivity. U.S. productivity growth had picked up, but most of the acceleration began before the current generation of AI tools. In a Census Bureau business survey analysis cited in the paper, 44% of AI-using firms reported using it to assist workers with tasks, while 5% reported using it only to replace tasks. The figures describe firms’ reported use. They do not measure how many people those firms hired or laid off.

The distinction gives job seekers something more specific to watch than headlines about spending or growth. Hiring rates for new workers in their field can fall even when total employment holds up. In the Stanford study, fewer starts showed up more clearly than more departures.