Skip to content

Policy

Fed’s Warsh outlines three AI scenarios for growth and jobs

Edited and reviewed by AI Ping Editorial Team

3-line summary

  1. Federal Reserve Chair Kevin Warsh outlined three ways AI could affect productivity and jobs.
  2. His scenarios cover broadly shared gains, growth with higher unemployment, and a slowdown if investment returns disappoint.
  3. Fed staff have started examining possible outcomes. The November 16 event was a research conference for the Productivity and Jobs Task Force.
PracticalOffice workers
A lined notebook and pencil rest on a rail near a row of servers
AI-generated conceptual image: servers beside a notebook represent the study of AI investment’s economic effects · AI Ping

Federal Reserve Chair Kevin Warsh described three ways wider adoption of artificial intelligence could play out for the U.S. economy on November 16. Productivity gains could reach workers and owners of capital broadly. Growth could also coincide with job losses from automation. A third scenario has stocks and economic activity falling if returns on AI investment fall short of expectations.

Opening a Fed Productivity and Jobs Task Force conference in San Francisco, Warsh said staff at the Board had started using economic models to examine possible outcomes. He presented the three cases as examples for planning over the coming years. The conference focused on questions such as how quickly the technology spreads and when its effects might appear in productivity and employment data.

Growth and employment can move in different directions

In the first case, AI makes work more productive, and the gains reach workers as well as capital owners. Real gross domestic product, which measures output after accounting for price changes, continues to grow above its long-run average. Employment stays close to full employment as AI complements workers’ tasks.

Real growth also beats its long-term average in the second case. Here automation replaces some jobs, and workers have difficulty moving into new ones. Unemployment can rise above levels consistent with full employment even as the economy produces more. Warsh identified that combination as one possibility for the staff’s analysis.

Warsh also raised questions about how widely AI is deployed, whether it improves the efficiency of production overall, and how long any gains last. The San Francisco Fed’s agenda set aside separate sessions for the speed of adoption, which economic measures could register changes and what happens when businesses go from pilots to regular use.

One question Warsh posed concerns total factor productivity, a broader measure than output per hour worked. The Bureau of Labor Statistics calculates its private-sector measure by comparing real output with an index of labor and capital inputs. If a business raises production while hours stay flat but it has added equipment, the measure takes the extra capital into account.

A robotic gripper over a tray of parts, with a work glove and timer on the bench
AI-generated conceptual image: machine parts, a gripper and a timer depict how automation can change productivity and work · AI Ping

If returns on AI investment disappoint

Warsh’s third case starts after substantial investment and stock-market gains. If expected AI returns weaken, equity values can fall and economic activity can slow, with a larger rise in unemployment. His example then allows for a recovery without the productivity improvement companies had expected.

At a separate event the same day, Fed Governor Michael Barr discussed AI investment financing in the November Financial Stability Report. The arrangements he highlighted included long-term data center leases, power contracts and asset-backed financing. Barr said intertwined financing arrangements can make it difficult to identify where losses might land when investor sentiment shifts.

AI was the most frequently cited risk in a market survey included in the same report. The New York Fed gathered responses from 23 contacts in September and October, including people at broker-dealers, banks, funds and advisory firms. Some flagged uncertain investment returns and financing arrangements that could amplify losses if asset values fall. The box in the report summarizes those contacts’ views.

Copper-colored disks and a leaning domino sit on green felt
AI-generated conceptual image: a leaning domino beside metal disks illustrates how lower expectations for AI investments can affect asset prices and the wider economy · AI Ping

The task force’s research feeds into policy discussions

The Fed announced leaders and objectives for five task forces in July, including the group on productivity and jobs. That group evaluates how widely and quickly general-purpose technologies like AI spread and how they affect the economy and employment. The Fed says the groups work independently with staff support and provide findings to the Federal Open Market Committee.

The task force is co-led by venture investor Marc Andreessen, economist Charles Jones and Microsoft’s Asha Sharma. San Francisco Fed President Mary Daly moderated a session on business adoption on November 16. The agenda listed an economist from the jobs site Indeed and executives from robotics and packaging-equipment companies as panelists.

The Fed’s monetary policy responsibilities include maximum employment and stable prices. The work Warsh outlined examines what AI adoption could mean for those responsibilities as its effects on production and labor demand become clearer. The November 16 program consisted of research discussions and panels and carried no interest-rate decision.

Sources