When CFOs talk to shareholders, they emphasize AI’s productivity gains. When they talk to researchers, the story is more direct: they’re planning to cut jobs.
A new working paper from the National Bureau of Economic Research surveyed 750 US chief financial officers about their AI plans. Less than half—44%—said they’re planning AI-related job cuts in 2026. But across the broader economy, that translates to roughly 502,000 positions expected to disappear this year.
That’s nine times higher than the 55,000 AI-attributed layoffs in 2025.
The Numbers in Context
The headline sounds alarming, but the math tells a different story. Half a million jobs is just 0.4% of America’s roughly 125 million private sector positions. For comparison, the economy adds or loses that many jobs in a typical month from normal churn.
So far in 2026, artificial intelligence was cited in 12,304 US job cuts announced between January and February—about 8% of the total layoff figure during that period. The tech sector has eliminated nearly 60,000 jobs total since January, with 171 separate events hitting 59,121 workers.
Who’s Actually Cutting
The most aggressive cuts are coming from tech and finance:
Amazon leads with roughly 16,000 cuts this year, despite posting record 2025 revenue of $716.9 billion.
Block (Jack Dorsey’s company) cut about 40% of its workforce—more than 4,000 employees—explicitly citing AI as the reason. Dorsey predicted most companies will make similar cuts within the next year.
Meta reportedly directed senior executives to plan workforce reductions of roughly 20%, translating to about 15,000 positions.
Atlassian cut 10% of staff, then announced plans to hire approximately 800 new roles focused on AI engineering, machine learning operations, and AI safety.
The Productivity Paradox
The most striking finding isn’t about layoffs—it’s about what companies are getting from AI.
The researchers found a significant gap between perceived and actual productivity gains. Executives are seeing potential more than results. Companies have invested heavily and are seeing “cool things they’re either starting to do or hope to do in the near future,” but the financial returns aren’t materializing yet.
This tracks with other recent surveys showing that while 72% of enterprises have at least one AI workload in production, only 66% report actual gains in productivity and efficiency.
Who Wins, Who Loses
For workers in roles that involve routine information processing—data entry, basic analysis, customer service scripts, document review—the outlook is getting worse. These are exactly the tasks that current AI handles well.
For AI specialists, the picture is the opposite. Atlassian’s move is typical: cut traditional roles, then hire aggressively for AI engineering, prompt engineering, MLOps, and AI safety.
For executives, there’s a growing credibility problem. They’re telling Wall Street that AI will transform their businesses while telling researchers they expect modest productivity gains. Someone is getting oversold.
The 9x increase in AI layoffs sounds catastrophic until you realize it represents less than half a percent of the workforce. The question is whether that number stays small—or whether 2026 is just the beginning.