But behind closed doors, business leaders controlling company headcounts are actually telling a more subtle story. A working paper from the National Bureau of Economic Research found that, out of a survey of 750 chief financial officers from U.S. firms, less than half (44%) say they plan on some AI-related job cuts. When the co-authors calculated what that amounts to across the broader economy, they found just 0.4%, or about 502,000 roles out of about 125 million roles, are expected to be lost this year. Just about half of those job losses will come from the white-collar world.

That 9x increase from last year’s 55,000 AI-attributed layoffs is striking — but still a rounding error against the overall workforce.

“It’s not the doomsday job scenario that you might sometimes see in the headlines,” John Graham, co-author of the study and the director of the Duke CFO survey, conducted in partnership with the Federal Reserve Banks of Atlanta and Richmond, told Fortune.

Moreover, the study finds a wide gap between the perceived and actual productivity gains from AI, finding that perceptions of AI’s gains are larger than the reality. The researchers say this likely reflects a delay in realized revenue. This reported lag matches what economists have been saying about AI’s productivity gains.

Goldman Sachs senior economist Ronnie Walker noted earlier this month that amid AI investment zeal, “we still do not find a meaningful relationship between productivity and AI adoption at the economy-wide level.” It’s not just economists; workers have reported that AI is actually making them less, not more productive, placing greater strain on their workflows, with time spent across some job responsibilities increasing by up to 346%.

Lagging productivity gains and Solow’s paradox

Economists have a name for this gap — and it dates back to the dawn of the personal computer era.

In a comparison to the lagging technological innovation associated with the dawn of the internet era, the researchers invoke Solow’s paradox, also known as the productivity paradox, to contextualize the current divide between perceived and actual productivity from AI. Coined by Nobel Laureate Robert Solow in 1987, the paradox speaks to the observation that transformative technology—like computers, or in this instance, AI—can appear ubiquitous while remaining absent from economic data. 

“You can see the computer age everywhere but in the productivity statistics,” Solow said.

Graham said what executives are seeing with productivity today is actually more of a wish than a realized fact. Companies see the potential of AI without the financial results to match. “Companies have invested and they’re realizing all these kind of cool things that they’re either starting to do or they hope to do in the near future,” he said. “But it’s not really showing up yet in revenue”

The current state of AI-related layoffs

However, the study still represents a step in the direction of heightened job losses thanks to the technology, adding some credibility to what tech execs have painted publicly. Employers reported about 55,000 layoffs attributed to AI in 2025, according to research firm Challenger, Gray & Christmas. That’s just 4.5% of all job losses from last year. Still, if the study’s numbers are right, that would mean a 9x increase in AI-related layoffs this year.

There have already been several sizable AI-related layoffs reported by firms this year. Jack Dorsey’s Block cut about 40% of its workforce, or more than 4,000 employees, because of the technology. Australian-American financial services firm Atlassian cut 10%. And Meta is reportedly planning on cutting 20% of jobs, as CEO Mark Zuckerberg reportedly creates an AI agent clone of himself. On top of that, the job market is in a lull. U.S. employers posted 92,000 jobs losses last month, and the unemployment rate ticked up to 4.4%.

But the report also finds that AI adoption could actually lead to an increase in hiring among smaller firms, further contradicting the public claims from AI’s biggest leaders. Many smaller firms, or those with fewer than 500 employees, have only just begun to invest in AI as most incur the brunt of AI-related operating expenses, slowing down adoption rates. But as adoption increases, small firms noted they plan to increase hiring in technical roles. On top of that, larger firms are planning on holding technical roles constant.

“If anything, small companies are hiring a bit on the technical side which will offset [losses] a little bit,” Graham said.

Still, the study only paints a picture for the short-term, so it’s difficult to entirely rule out the dire predictions some tech leaders have made of the technology’s ability to soak up white-collar jobs.

“Who knows what’s going to happen in 2028?” Graham said. ”I’m not making a prediction that there will never be any jobs lost two, three and five years from now to AI.”