By Justin Lahart
America's chief financial officers say that artificial intelligence will push some people out of their jobs: primarily workers in routine, clerical and administrative roles. Workers with highly skilled roles, such as architects and engineers, are more likely to keep their jobs, especially if they can use AI to their advantage.
A new study, based on a survey of about 750 chief financial officers, found that so far AI had essentially no employment effect in 2025 and that most expect AI will lead their companies to trim only a small number of their overall jobs this year.
It is still possible that workers with jobs that require more education and more training could eventually get hit, "but probably not in 2026," said John Graham, an economist at Duke University and one of the paper's authors. It was released this week as a working paper on the National Bureau of Economic Research website.
Graham has been surveying chief financial officers about their expectations for their companies and the overall economy for 30 years. CFOs are uniquely placed to understand the inner workings of their companies, Graham said, since it is their job to keep watch on how company resources are being deployed.
The survey, produced with economists from the Federal Reserve Banks of Atlanta and Richmond, was conducted in late 2025 and early 2026. It showed that, in aggregate, CFOs expected that AI would reduce their companies' head count this year by about 0.4%, compared with what it otherwise would have been.
The CFOs represent a range of industries, including finance, tech, manufacturing and professional services, and the survey is conducted quarterly. For this edition, in addition to regular questions about their outlook, the CFOs were asked an array of questions about AI.
The CFOs were twice as likely to say that AI could lead to job cuts as they were to say it would enhance work in office- and administrative-support areas such as bookkeeping, clerical work and customer service.
But for other, more advanced roles, they were more likely to say that AI would enhance work as opposed to eliminating it. This was especially true of some roles that required high levels of education.
That pattern echoes what economists call skills-biased technological change: the tendency of some new technologies to hollow out routine work while complementing jobs held by more highly educated workers.
When personal computers started arriving in offices in the 1980s, college-educated employees such as financial analysts, scientists and consultants were able to do more at work. But jobs that entailed doing more routine cognitive work such as typists and back-office bookkeepers -- roles that had once promised a solid path to the middle class -- were no longer so vital.
Those jobs didn't disappear, but the share of workers doing those kinds of office support roles shrank. More workers who lacked a college degree crowded into lower-paying roles that hadn't been displaced by the computer, such as leisure and hospitality work.
Whether AI will ultimately be skills-biased, hurt the highly educated more or broadly raise worker productivity is a topic of debate among economists.
One unsettling problem for workers: The people who do lose their jobs won't necessarily get the new jobs that AI creates. Atlanta Fed economist my plea, one of the study's authors, is optimistic that AI will eventually create new types of work. But she also said that many of the roles the CFOs point to AI reducing are "stepping stones" for moving into the middle class.
That could be especially hard on young people looking to land that first job.
Graham cautioned that even though the study signaled that AI would slightly weigh on overall employment, the survey only includes companies that are already established, as opposed to new ones.
That matters, because it is often new companies embracing new technologies and figuring out how to use them that propel job creation. The personal computer didn't just change what existing businesses did, for example, but gave rise to entirely new industries.
Indeed, the survey hinted at that dynamic. Larger companies -- those with 500 or more employees -- were more apt to say they were cutting routine workers, while keeping employment of "skilled technical" workers flat. In contrast, smaller companies said that they planned to keep employment of routine workers flat, and step up employment of more skilled technical workers.
That suggests that larger companies, which tend to grow more slowly and are focused on squeezing out efficiency, have stronger incentives to use AI to cut costs. On the other hand, said Graham, "small companies look at this and think, 'This gives us opportunities to expand.'"
Write to Justin Lahart at Justin.Lahart@wsj.com
(END) Dow Jones Newswires
Atlanta Fed economist Salomé Baslandze, one of the study's authors, is optimistic that AI will eventually create new types of work. In "America's Chief Financial Officers Say AI Is Coming for Admin Jobs," at 5:30 a.m. ET, the Atlanta Fed economist's name was missing from an earlier version of this article.
(END) Dow Jones Newswires
March 24, 2026 07:14 ET (11:14 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments