CFOs Admit AI Layoffs Will Be 9x Higher This Year, But Won't Say Where

Duke University's latest CFO Survey, conducted with the Federal Reserve Banks of Atlanta and Richmond, contains a number that should alarm anyone who earns a salary. CFOs now expect AI-attributed layoffs to reach roughly 495,000 in 2026. That's a 9x increase from the 55,000 cuts blamed on AI in 2025.

Those same CFOs, when asked about the measurable impact of AI on their businesses, reported seeing no meaningful gains in productivity, decision-making speed, customer satisfaction, or cost savings.

John Graham, a professor of finance at Duke's Fuqua School of Business and co-author of the study, told reporters: "It's not really hitting the top line yet in full force. There is some level of delay in here for sure."

Companies are eliminating half a million workers based on the promise of technology that isn't delivering returns. Duke's survey makes this explicit: AI investment in 2026 is "not expected to have much effect on the number of employees or produce measurable cost savings." Yet layoffs attributed to AI are accelerating.

The Block That Broke the Pattern

Jack Dorsey announced 4,000 job cuts at Block in early March, 40% of the company's global workforce. Dorsey said the reason was "the growing capability of AI tools to perform a wider range of tasks." It remains the largest single AI-attributed layoff event in tech history.

Block reported $24.7 billion in revenue for 2025, up 16% year-on-year, with operating income growing 34%. The company wasn't failing, and the layoffs weren't about survival.

JobsPikr, a firm that tracks hiring velocity through job posting data, said Block and Amazon both surged hiring in H2 2025 before posting volumes collapsed in Q1 2026. "If AI productivity were the driver, the contraction would have been gradual," the firm wrote in a recent analysis. "What the data shows instead is a strategic pause, a freeze following a decision made at the boardroom level."

Amazon accounted for 52% of all tech layoffs in Q1 2026. The company that spent billions automating its warehouses is now automating its corporate workforce. At least, that's the story being told to analysts. The productivity data tells a different story.

Wish-Casting as Strategy

Graham's phrase, "more of a wish than a realized fact," is the quiet part said out loud. CFOs are not seeing the productivity gains they expected, but they're hoping those gains will materialize. In the meantime, they're restructuring as if those gains already exist.

Duke's survey reveals that CFOs expect AI to "increase worker productivity, enhance decision speed and accuracy, improve customer satisfaction, and enable employees to refocus their time on higher-value tasks." All of it phrased in future tense, not present, not past.

Wall Street is rewarding this behavior. Stock prices rise on layoff announcements. Analysts praise "efficiency gains" before efficiency is measured. Companies are firing people at scale based on what AI might do, not what it's currently delivering.

Firms are dressing up layoffs as AI efficiency gains when the data shows they're not replacing workers with AI at scale. Telling investors you're cutting jobs for AI sounds better than admitting you overhired.

— Oxford Economics, March 2026

Graham told the Wall Street Journal: "It is possible that CFOs are just optimistic about all the potential." It is also possible that CFOs are under pressure to demonstrate AI adoption to investors regardless of whether it makes business sense. Duke's survey doesn't distinguish between the two.

The Readiness Gap No One Mentions

Companies are eliminating workers who hold institutional knowledge while simultaneously lacking the trained workforce needed to direct, validate, and quality-control the AI systems replacing them. JobsPikr, a hiring velocity tracker, calls this the "readiness paradox."

Customer service roles have declined 24-28% in posting volumes over 18 months. Data entry and bookkeeping roles are down 34% and 10% respectively. Media companies have cut copywriting roles 34% year-on-year. These are not jobs being automated successfully; they're jobs being eliminated in anticipation of automation that hasn't arrived yet.

Those same companies are not hiring the AI governance specialists, implementation managers, and quality control roles needed to manage the systems at scale. The result isn't efficiency. It's an expertise vacuum.

Workers Face a Calculation Problem

If you're in your twenties or thirties and working in tech, customer service, content, or analytics, the CFO survey is delivering an uncomfortable message. Your job security depends less on your performance and more on whether your CEO believes AI will someday replace you.

Junior developers, entry-level analysts, and customer support roles are facing the steepest pressure. Not AI has proven it can do the work. Executives have decided it eventually will. Senior roles with domain expertise remain more secure for now. AI systems still require experienced workers to architect, validate, and fix them when they fail.

The survey data shows 15% of companies plan to reduce headcount in 2026, 26% plan to keep it flat, and 59% expect to increase it. The median company expects employment growth of 1.7%, roughly the same as prior years. Those aggregate numbers obscure what's happening at the role level. Companies are cutting junior positions and back-office roles while hiring for AI implementation and engineering.

Graham said that "for some industries, AI is going to be about replacing the call center. For another, it's about a conveyor belt in a factory. For another, it's about having fewer analysts." The commonality: the people being replaced are rarely the ones making the decisions about AI adoption.

The Accountability Problem

Duke's survey was conducted between November 2025 and early 2026, polling 548 CFOs. Researchers released the findings in mid-March. Since then, at least 60,000 additional tech jobs have been cut, with AI cited as a factor in roughly 20% of those reductions.

Duke's survey doesn't capture what happens when the productivity gains fail to materialize. Will companies quietly rehire in 2027? Will they admit the cuts were premature? Or will they double down, blaming external factors like tariffs, interest rates, and regulatory uncertainty for results that were never tied to AI performance in the first place?

Graham's observation that this is "not the doomsday job scenario that you might sometimes see in the headlines" is technically correct. Five hundred thousand jobs represents 0.3% of the U.S. workforce. It's a rounding error at the macro level.

It's not a rounding error if you're one of the 495,000. And it's not a sound business strategy if the technology you're betting on hasn't proven its value.