In this article
Federal Reserve Governor Michael Barr urged the United States to prepare for possible disruptions in its labor market as businesses adopt artificial intelligence. Speaking on November 12, 2026, he pointed to early signs that AI may substitute for some workers, particularly those who are younger or have less experience. He also cited evidence of productivity gains and increased demand in parts of the economy. Whether those gains reach a wide share of workers, in his view, depends partly on how quickly companies adopt AI and how much competition survives as the market grows. Barr’s speech
What the early evidence shows
Barr said relatively few workers had been displaced by AI so far and that adoption remained limited. The evidence he cited includes workers getting more done with the tools, rising demand in some sectors and the emergence of new occupations. Signs of substitution in other parts of the economy can coexist with those gains. A study of workers who already use a tool, for example, answers a different question from a study of who is getting hired.
Reuters highlighted Barr’s warning about the possibility of serious short-term disruption. Its severity would depend partly on whether the technology can take over work now performed by people and how quickly employers deploy it. Barr’s remarks express his own assessment, as the speech notes. They give no forecast for the unemployment rate and announce no decision on interest rates. Reuters via MarketScreener
Adoption speed changes the time workers have to adjust
A job often includes work that responds differently to automation. In customer support, routing a familiar request may follow a predictable pattern. Handling an unusual complaint can require a professional to understand what happened before and decide what remedy fits. A company could delegate part of the routing while giving its staff more time to handle difficult cases. It might also change the staffing it needs. The result depends on how management organizes the work.
If companies introduce new tools gradually, workers may have time to train or move between roles as demand changes. If many employers automate similar tasks at once, hiring in some jobs could fall before opportunities expand elsewhere. A person who finds another job may still need training or earn a different wage. Those costs can show up during a transition even if the economy eventually produces more per hour.
Entry-level work is also a way to learn
A new software engineer may learn how a product works by writing tests for common failures. An operations analyst may learn to recognize a suspicious or incomplete document by reviewing many ordinary ones. If AI takes over some of that first pass, employers may hire fewer people for those tasks or teach new staff to handle them differently. The outcomes depend on whether firms invest in other ways for workers to build experience.
Hiring data by occupation and experience can help track this risk. Fewer openings in one function might reflect technology, weaker demand or an earlier decision to change staffing. The Brazilian digital lender Nubank said it slowed hiring after gaining capacity. Its account illustrates one company’s decision and concerns a different labor market from the one Barr discussed.
Who receives the gains from higher output
Barr cited a U.S. labor share of national income of 52.8% in the second quarter of 2026 for the series he discussed. The share measures how much income goes to labor on that definition. Assigning certain kinds of income to labor or capital affects the measurement, and the long decline in the series predates the current wave of AI. The figure provides context for the distribution question in the speech.
An improvement in productivity creates room for higher wages or profits, and competition can also pass some benefits to customers through lower prices. How gains are divided depends in part on workers’ bargaining power and firms’ ability to compete. Income and wealth are also different measures. A household that pays its bills from wages may have limited ownership of the businesses whose shares rise in value, even if it holds some stocks through a retirement plan.
Competition could spread the gains or concentrate them
One possibility Barr described is that competition makes AI affordable and widely available. Small businesses could gain access to tools and skills that were previously beyond their budgets. If many firms can adopt them, improvements in services and productivity could spread to consumers and companies across several industries.
Another possibility is that access to computing power, data and the resources needed to improve models gives a small set of providers a durable advantage. Larger businesses may be able to spend more on adoption and win customers from firms with less access. If returns collect in a small group of companies and their owners, labor could receive a smaller share of income. One test of competition is how easily customers can switch providers. Another is whether small firms can afford tools comparable to those their larger competitors use. The arrival of new suppliers would also help show whether existing firms' advantages can be overcome.
How to judge whether the transition is working
Barr pointed to job creation, training and efforts to connect workers to new jobs as ways to manage disruption. A training program can enroll many participants and still leave them without a job that uses what they learned. Employment and earnings after training give a better picture of whether the program helps people move into work they can sustain.
Barr’s economic figures concern the United States. Applying his questions to another country would require local evidence on adoption, job categories and access to training. Tracking hires, departures and earnings by occupation would help show whether changes in everyday tasks are developing into lasting shifts in who gets work and who benefits from it.
