Challenger treats AI as a fait accompli in its monthly cut totals. The NBER's 6,000 firm survey and the headcount data suggest the substitution has not actually arrived.
Companies are cutting workers to pay for an AI they have not yet built. That is the gap between Challenger, Gray & Christmas's March 2026 report, in which AI was the top-cited reason for the 15,341 jobs eliminated that month, and the NBER's global C-suite survey of nearly 6,000 executives, in which almost 90% reported that AI has had little to no impact on employment or productivity at their firms over the past three years. The two findings can both be true, and the labor market is what happens in between.
The numbers describe a hiring freeze, not a substitution. U.S. employers announced 60,620 job cuts in March, up 25% from February's 48,307, according to Challenger. AI was the stated cause for 25% of them. April was sharper: 88,387 cuts, with 21,490 of them attributed to AI, and Technology leading all sectors with 33,361. Year to date, the Technology sector has shed 52,050 jobs, 40% more than during the same period in 2025 and the highest YTD total Challenger has recorded since 2023. The most recent quarter, with 217,362 cuts, is the lowest Q1 since 2022, down 56% from a federal-heavy Q1 2025. The labor market is not in free fall. It is being trimmed with a stated reason that is not what most executives, in a separate survey, say they are actually seeing on the ground.
Andy Challenger, the firm's senior vice president, gave the cleanest version of the split on the way out of the March report: "Companies are shifting budgets toward AI investments at the expense of jobs. The actual replacing of roles can be seen in Technology companies, where AI can replace coding functions. Other industries are testing the limits of this new technology, and while it can't replace jobs completely, it is costing jobs." In April, he told CBS News the version that lands harder: "Regardless of whether individual jobs are being replaced by AI, the money for those roles is." That second sentence is the story. The budget line is being moved before the substitution has happened.
The replacement story also has an empirical test. Goldman Sachs estimates that AI is currently reducing U.S. employment by roughly 16,000 jobs per month. That is a real number, attributed to a real estimate, and it is smaller than the AI-cited monthly cut totals Challenger has been recording. The gap is not proof of overstatement. It is a sign that the rationale a company puts on a layoff press release is not the same as the headcount actually offset by deployed AI. The first is a budget declaration. The second is an operational fact, and operational facts lag.
The NBER survey is the other data point that complicates the official story. In a poll of about 6,000 C-suite executives across industries, almost 90% said AI has had little to no effect on their employment or productivity in the past three years. That is not a vindication of the technology. It is a measurement of how unevenly the substitution has been distributed. A small number of large companies in Technology have plausibly swapped coding capacity for model output. The rest of the economy is in the testing phase that Challenger named, and the cuts are arriving before the tests are done.
Recent graduates are paying the bill for the testing phase. The New York Fed's most recent data, summarized by Yale's Jeff Sonnenfeld and his co-authors, puts unemployment for recent college graduates at roughly 6%, double the rate of the broader workforce since 2022. Computer science graduates now report more difficulty finding jobs than humanities majors, an inversion that would have read as a typo a decade ago. The entry-level rung is where the "AI can replace coding functions" line meets labor markets, and the entry-level rung is the one being cut.
The forecasts from the people selling the transition are louder than the deployment data. Anthropic's Dario Amodei has predicted that up to half of entry-level white-collar jobs could disappear within five years. Verizon's Dan Schulman has warned of a 10% to 30% unemployment rise within two to five years. BCG has put 10% to 15% of existing jobs at risk of elimination by 2031. These are forward-looking claims from people with stakes in the outcome. The realized headcount, so far, is a fraction of those projections, and the productivity gains the cuts are supposed to fund are not yet visible in the survey responses from the executives doing the cutting.
The transition is being financed, in headcount, before the productivity gains have been delivered, and the cost of that sequencing is being borne by the workers and the buyers rather than the sellers. Companies are treating the AI substitution as a fait accompli in their layoff announcements while reporting, in separate surveys, that the substitution is still largely hypothetical inside their operations. The Challenger report and the NBER survey are not in conflict. They are measuring different stages of the same process. One is the announcement. The other is the result. The announcement is arriving first.
What to watch next is the gap between the two. The next quarterly Challenger report will show whether the April surge in Technology cuts continues, and whether the AI-cited share stays in the 25% range or climbs toward the share the CEO forecasts imply. The next NBER refresh will show whether the 90% "no impact" number moves. If the first falls and the second stays flat, the sequencing problem is closing. If the first stays high and the second stays flat, the gap is widening, and the workers being cut are funding a transition the buyers are not yet receiving.