AI job losses: Look up which workers are most vulnerable

No one has a perfect road map to the future, but researchers at GovAI, which studies technology policy, and the Brookings Institution, a Washington think tank, used a novel approach to estimate which workers may be most and least able to adapt to AI.

They concluded that many people most at risk if AI transforms work are also the best placed to find new jobs. You can use the search box and interactive chart above to explore which occupations may have bright prospects and which may not.

But history shows that economists and researchers have been terrible at predicting the effects of new technologies on work and workers, so take forecasts like this one seriously but not literally. Even researchers cranking out studies of AI in workplaces caution that they’re making useful but fallible best guesses.

“All the important questions about AI’s effects on the labor market are still unanswered,” Jed Kolko, a senior fellow at the Peterson Institute for International Economics, recently concluded. Economists at Anthropic, the AI start-up behind the Claude chatbot, stressed the need for “humility” in their analysis of AI seeping into occupations. (Humility is uncommon in Silicon Valley.)

Sifting through reports about AI-related job displacement will make you dizzy with the apparent contradictions.

One influential Stanford University analysis last year found it’s probable that AI is bleeding jobs from young people in occupations such as software development and customer service, where AI adoption has been fastest. Different research, from the Economic Innovation Group think tank, essentially concluded the opposite, that young workers in those occupations were faring better than their peers in less AI-exposed fields such as fitness training and roofing.

The Federal Reserve Bank of Dallas has said it’s unlikely AI is coming for jobs in the next decade, while prominent CEOs keep predicting that AI will put millions of people out of work soon.

A flood of sometimes conflicting analyses shows the yawning gap between what little is known about how AI is changing work and everyone’s understandable hunger for certainty. The divide lets Americans, business leaders and policymakers cherry-pick their preferred narratives. If you’re afraid of being cast aside for AI, there’s informed and uninformed evidence to fuel your nightmares. There’s plenty of support, too, if your think your job is safe.

Two points of general agreement stand out: There’s no measurable evidence so far that AI is putting Americans as a whole out of work, economists say. And while the victims of past workplace automation were mostly factory and trade workers, it’s white collar jobs that are first in line for AI shake-ups today.

That’s where the recent research, led by Sam Manning, a senior research fellow at GovAI, and his colleague Tomás Aguirre, took a novel approach.

They started with an industry-standard measure of more than 350 occupations’ AI “exposure.” That estimates how many job-related tasks a worker can plausibly do more efficiently with AI, such as a teacher grading homework.

Those estimates find that skills used in computer programming, marketing, financial analysis and customer service have high overlap with AI capabilities, and therefore in theory workers might be more easily replaced by machines. (There are spicy debates in economics over measuring AI “exposure” and whether high AI exposure helps or hurts workers.)

The researchers went one step further and also sought to quantify how easily people within occupations could shift to other good-paying work if AI killed their jobs. The researchers figured that people with more education and varied work experience could more easily switch occupations, and that it would help to be wealthier, under 55 years old and live in cities where jobs are plentiful.

While web designers and secretaries both scored high in the research for exposure to AI, they diverged in their estimated ability to adapt. Secretaries were among the 6.1 million largely clerical and administrative workers considered both highly exposed to AI and with the lowest estimated adaptability, Manning and Aguirre found. (Manning is also affiliated with the technology policy research group the Foundation for American Innovation.)

The findings suggest that the majority of workers whose jobs may be transformed by or lost to AI can bounce back. But a smaller share of workers may have a harder time finding new jobs.

Women make up about 86 percent of those most vulnerable workers, the researchers said, suggesting the negative effects of automation won’t be borne equally across society.

Mark Muro, a senior fellow at Brookings who assessed the policy relevance of the research, said the most vulnerable workers “may be out of sight and out of mind” to policymakers and the American public. The researchers cautioned that it’s hard to accurately assess the likelihood of people finding other jobs.

Allison Elias, a professor at the University of Virginia business school, said that previous technology shifts show why people in female-dominated clerical occupations might be on the losing end of the AI revolution.

In her historical research, secretaries and other administrative staff often hoped that new technologies would free them to do higher-level work and help advance their careers. Instead, she said, workers were often expected to do more work for the same or lower pay. They continued to report low levels of job satisfaction.

“These people are really vulnerable because they won’t have a lot of decisions over how AI is used, and their exit opportunities are going to be pretty low,” said Elias, who was not involved in the GovAI and Brookings analysis.

Economists say it’s nearly impossible to forecast AI’s effect on the labor market from the current capabilities of the technology or the business sectors it’s seeping into first. And they point to the track record of past technology revolutions, such as electricity and smartphones, that eliminated some types of jobs but also created new work and economic growth few foresaw.

The predictions mostly didn’t pan out from a prominent study more than a decade ago that estimated nearly half of jobs could be destroyed by computer automation. Forecasts were off base that ATMs would wipe out bank tellers, that earlier forms of AI would decimate radiologists and that player pianos would kill the jobs of pianists. Few people imagined that smartphones would usher in new jobs in social media marketing and influencing. And you’re probably not experiencing the 15-hour workweek that economist John Maynard Keynes forecasted in 1930.

“We do not have a good track record of predicting how technological change will play out in the labor market,” said Martha Gimbel, executive director of the Budget Lab at Yale University. It would have been hard to predict that the invention of electricity would lead to the new occupation of elevator operators, and that a subsequent innovation — “buttons,” she said — would wipe out those jobs.

Another extinct occupation, telephone switchboard operators, offers reasons for both hope and pessimism about AI’s effects. It was once one of the most common jobs for American women, but jobs were wiped out as telephones modernized starting in the early 20th century, according to a research paper published in 2024 by James Feigenbaum and Daniel Gross.

Switchboard operators who lost their jobs were far more likely than their peers to never find other work or to take lower-paying jobs, the research found. But within years, new opportunities opened for young women as secretarial and restaurant work boomed. “I read that as somewhat hopeful,” Feigenbaum, a Boston University economic historian, said in an interview.

Feigenbaum doesn’t buy the argument that AI will be much different for American workers than prior technology revolutions. The invention of electricity, the internal combustion engine and the internet were massively transformative technologies, he said, and “that didn’t eliminate all jobs.”

Taylor Telford contributed to this report.