Fed & ratesCentral banks & interest rates
Fed's Jefferson says AI is raising demand and prices before productivity, keeping inflation at the center of rate policy
Federal Reserve Vice Chair Philip Jefferson argued that AI's measured productivity payoff remains limited, while AI-driven stock gains, household spending and data-center chip demand are adding to price pressures. With core PCE inflation at 3.1%, inflation rather than AI optimism will guide rates.
From the Federal Reserve · Speeches article of November 14, 2026 and cited sources: In his November 13 speech, Jefferson said AI currently works more like a demand shock than a supply boost. [1]
SOURCES[1] Federal Reserve: Jefferson, Nov 13, 2026 speech federalreserve.gov
AI-assisted, not human-edited. Methodology Corrections Who owns this site
- What happened
- In his November 13 speech, Jefferson said AI currently works more like a demand shock than a supply boost. [1]
- Original source
- Federal Reserve · Speeches article, company source
- Brief by
- AT Pulse research desk AI-assisted; publisher and contact on the desk page
- When
- November 14, 2026 article published
- Covered
- Nov 15 edition
- Evidence
- 9 sources across 7 sites
- 1Federal Reserve: Jefferson, Nov 13, 2026 speechfederalreserve.gov
- 2Federal Reserve: Jefferson, Nov 7, 2025 speechfederalreserve.gov
- 3NBER w34836: Firm Data on AInber.org
- 4Census CES-WP-26-25: Microstructure of AI Diffusioncensus.gov
- 5BLS: Productivity and Costs, Q3 2026bls.gov
- 6TokenPost coveragetokenpost.com
Reach: not measured yet

In his November 13 speech, Jefferson said AI currently works more like a demand shock than a supply boost. Optimism about AI has lifted stock prices, and the wealth effect is supporting household spending. The data-center buildout is also driving up memory chip prices. With rates on hold, he expects these cost pressures to ease as planned DRAM capacity comes online. [1]
Task-level gains are large, but firm-wide gains are small. Customer-support agents were 14% more productive on average and 34% more for novices. A survey of nearly 9,000 executives, however, puts AI's contribution at about 0.2% a year in labor-productivity growth over three years. Most respondents reported no effect, though executives expect about 0.6% a year ahead. [7] [3] [9] [1]
Adoption is still narrow and hard to measure. About 25% of firms use AI when each firm counts equally, versus 95% when firms are weighted by employment. Among firms that use AI, 57% use it in three or fewer business functions, led by sales and marketing. Production and operations are the least integrated so far. [1] [4]
Aggregate productivity has strengthened, but the link to AI is unproven. Third-quarter nonfarm productivity rose at a 3.6% annualized rate, and labor's share of output fell to 53.0%, the lowest since 1947. Jefferson noted that faster productivity growth since 2020 began before chatbots appeared, and he did not address the falling labor share. [5] [1]
aggregate productivity is up, but studies attribute only small firm-wide gains to AI so far.
Read the full assessment
Implication: rate decisions are likely to track inflation, including pressure from AI-driven chip costs and spending, rather than expected productivity gains.
Executive brief
Federal Reserve Vice Chair Philip Jefferson said on Nov 13, 2026 that AI is raising prices before it raises productivity. Expectations about AI have lifted stock prices, and the resulting wealth is lifting household spending. The data-center buildout is also driving steep price increases for memory chips. Yet his reading of the evidence is that AI's contribution to measured productivity is still "limited." Core inflation (PCE, excluding food and energy) was 3.1% in September, and the Fed is holding rates steady. His conclusion is that inflation, not AI optimism, will set policy for now. Before AI can raise the economy's capacity, it is behaving like a demand shock.
What changed and event timeline
Earlier AI speech in Frankfurt
Jefferson pointed to fast worker adoption and task-level gains, such as 14% for customer-support agents. He warned that telling cyclical change from structural change would be hard for policymakers.
"Firm Data on AI" released
An NBER paper surveyed nearly 9,000 executives. It estimates AI raised annual labor-productivity growth by 0.2% over three years, and executives expect 0.6% a year ahead. Revised October 2026.
Census finds narrow AI use
Among firms that use AI, 57% use it in three or fewer business functions. Sales and marketing (52%) is the most common.
Q3 productivity jumps
Nonfarm business productivity rose at a 3.6% annualized rate (2.1% year over year). Labor's share of output fell to 53.0%, the lowest since the series began in 1947.
UC San Diego speech
Jefferson revisited Solow's 1987 productivity paradox. He judged AI's productivity effect limited so far but potentially large, and named inflation as the key policy signal.
More detail
Capabilities and access
This is a policy speech, not a product launch, and no specific model is assessed. A footnote says these technologies are "sometimes referred to as super intelligence or SI." Census data show about 25% of firms use it, ranging from about 40% in information and professional services to about 10% in mining and hospitality (speech).
Read the full section
This is a policy speech, not a product launch, and no specific model is assessed. Jefferson refers to generative AI, chatbots and coding assistants in general terms. A footnote says these technologies are "sometimes referred to as super intelligence or SI." On adoption, he cites 40–50% of working adults using generative AI at work, though few use it daily. Census data show about 25% of firms use it, ranging from about 40% in information and professional services to about 10% in mining and hospitality (speech).
Technical analysis for researchers and developers
- Census counts each firm equally and finds about 25% adoption. Results depend heavily on the weighting choice.
- Proving cause is hard. Labor productivity grew 2.2% a year in 2020–25, up from 1.5% in the previous cycle.
- Why task gains don't add up yet. He cites narrow deployment and offsetting costs, for example AI-written code that needs more debugging.
Read the full section
- Measurement matters. Census counts each firm equally and finds about 25% adoption. The executive survey weights firms by employment and finds 95% (speech fn. 13). Results depend heavily on the weighting choice.
- Proving cause is hard. Labor productivity grew 2.2% a year in 2020–25, up from 1.5% in the previous cycle. But the rise started before chatbots, and Jefferson credits other factors such as new business formation.
- Why task gains don't add up yet. He cites narrow deployment and offsetting costs, for example AI-written code that needs more debugging.
- Imports. Much IT equipment is imported, so it adds less directly to GDP. It still raises productivity by giving workers more capital (fn. 18).
Claims and evidence
- Support agents were 14% more productive on average and 34% more for novices (NBER w31161).
- Jefferson puts the effect at about 0.5% more sales per employee over three years.
- He also cites NBER w35684 (Babina et al.) finding AI adoption linked to productivity growth. These are correlations, not causal estimates.
Read the full section
- Big gains on specific tasks. Support agents were 14% more productive on average and 34% more for novices (NBER w31161). The published journal version reports 15% more chats resolved per hour (EurekAlert).
- Small gains for whole firms. Jefferson puts the effect at about 0.5% more sales per employee over three years. That is consistent with the survey's 0.2% a year, and 89% of executives report no effect at all (NBER; Atlanta Fed).
- Positive firm-level associations. He also cites NBER w35684 (Babina et al.) finding AI adoption linked to productivity growth. These are correlations, not causal estimates.
- AI demand is pushing up inflation. This is Jefferson's own judgment. No separate estimate of its size is cited.
Context and prior work
- The 1990s IT boom. Total factor productivity (output growth beyond added inputs) rose later, as firms reorganized around computers.
- Greenspan argued in 1998 that IT was lifting productivity. The link was only confirmed afterwards (Oliner and Sichel, 2000).
- Change from 2025. The Frankfurt speech stressed AI's promise. This one puts more weight on inflation from AI-driven demand (2025 speech).
Read the full section
- The 1990s IT boom. IT investment added about 0.75 percentage point to annual productivity growth in the late 1990s (Oliner, Sichel and Stiroh, cited in the speech). Total factor productivity (output growth beyond added inputs) rose later, as firms reorganized around computers.
- Proof came late. Greenspan argued in 1998 that IT was lifting productivity. The link was only confirmed afterwards (Oliner and Sichel, 2000).
- Change from 2025. The Frankfurt speech stressed AI's promise. This one puts more weight on inflation from AI-driven demand (2025 speech).
Limitations, safety and contested findings
- The firm-level figures come from executives describing their own firms, and samples differ across studies.
- Jefferson's main downside risk is AI displacing workers faster than they can retrain. Monetary policy cannot fix the structural part of any resulting unemployment.
- Faster productivity growth would probably raise it. But the historical link between productivity and real interest rates is noisy.
Read the full section
- Survey evidence. The firm-level figures come from executives describing their own firms, and samples differ across studies.
- Unemployment risk. Jefferson's main downside risk is AI displacing workers faster than they can retrain. Monetary policy cannot fix the structural part of any resulting unemployment.
- **Neutral interest rate (r*).** Faster productivity growth would probably raise it. But the historical link between productivity and real interest rates is noisy.
- Wages vs. output. The record-low labor share (BLS) suggests productivity gains are not reaching workers proportionally. Jefferson did not address this.
- No independent check. The reviewed sources include only two secondary write-ups of the speech, neither with independent analysis.
Business and practitioner implications
- Policy: Rates are on hold with headline PCE inflation at 3.6%.
- Cost of AI inputs: Memory and logic chip prices remain under pressure in the near term.
- Where the gains are: Gains will come from moving beyond sales and marketing into production and operations, the functions least integrated so far (Census).
Read the full section
- Policy: Rates are on hold with headline PCE inflation at 3.6%. Jefferson expects AI-related cost pressures to fade as chip supply grows, citing planned DRAM capacity expansions (fn. 19).
- Cost of AI inputs: Memory and logic chip prices remain under pressure in the near term.
- Where the gains are: Gains will come from moving beyond sales and marketing into production and operations, the functions least integrated so far (Census).
- Competition: Jefferson warns that firms slow to adopt could lose ground, and some workers, including in parts of software development, could be displaced.