Behind closed doors, America's top financial executives are painting a sobering picture of AI's impact on the workforce. A new working paper from the National Bureau of Economic Research, based on the Duke CFO Survey conducted in partnership with the Federal Reserve Banks of Atlanta and Richmond, reveals that nearly half of surveyed CFOs are planning AI-related job cuts this year.
The numbers tell a striking story: approximately 502,000 roles across the U.S. economy are expected to be eliminated due to AI in 2026, a ninefold increase from the roughly 55,000 layoffs attributed to AI in 2025.
The Data Behind the Headlines
The survey polled 750 U.S. chief financial officers and found that 44% plan some form of AI-related workforce reduction. Extrapolated across the broader economy, that translates to about 0.4% of the nation's approximately 125 million jobs. Half of the projected losses are expected to hit white-collar positions.
John Graham, co-author of the study and director of the Duke CFO Survey, offered perspective on the findings. "It's not the doomsday job scenario that you might sometimes see in the headlines," he told Fortune.
A Gap Between Rhetoric and Reality
The study reveals a notable disconnect between the dire predictions of prominent AI leaders and what companies are actually planning. Microsoft AI chief Mustafa Suleyman has predicted that AI will cause office jobs to "crumble" within 18 months, while Anthropic CEO Dario Amodei has suggested entry-level positions could be cut in half over a similar timeframe.
Yet the CFO data suggests a more measured reality. While headline-grabbing layoffs at companies like Block (4,000 employees, roughly 40% of its workforce), Atlassian (10%), and Meta (reportedly planning 20%) dominate the news cycle, the aggregate picture is far less dramatic.
The Productivity Paradox Returns
Perhaps the most counterintuitive finding is that AI may actually be increasing workloads in some cases. The study found workers report AI increases workloads by up to 346% for certain tasks — echoing Nobel laureate Robert Solow's famous 1987 observation that "you can see the computer age everywhere but in the productivity statistics."
Smaller firms appear to be bucking the trend entirely, planning to increase technical hiring rather than cut positions. Larger enterprises, meanwhile, are more likely to hold headcount constant — replacing departing workers with AI tools rather than conducting mass layoffs.
What It Means for Workers
The 502,000 figure, while significant, represents a fraction of normal annual job churn in the U.S. economy. The real shift may be more gradual than sudden: a steady reshaping of roles and responsibilities rather than the overnight disruption that dominates public discourse. For workers in white-collar positions, the message is clear — upskilling in AI tools is no longer optional, but the wholesale elimination of knowledge work remains, for now, more prediction than reality.
