Seventy-nine percent of Americans believe AI will reduce the number of US jobs over the next decade, according to the Bentley University-Gallup survey fielded in May and released in late July, up from 73 percent a year earlier. US employers reinforced the fear with hard numbers: Challenger, Gray & Christmas counted AI as a stated reason in roughly 101,000 job-cut announcements in the first half of 2026, making it the single most-cited cause of layoffs for four consecutive months. Yet the broad, economy-wide displacement that AI executives warned of has not shown up.
The gap between expectation and measurement is now the most honest thing to say about AI and work. Even the people selling the technology are walking back their forecasts. OpenAI chief executive Sam Altman has said he "thought there would have been more impact" on white-collar and entry-level jobs by this point, and that the jobs picture "is likely to be very different than we thought." That is a notable retreat from Anthropic chief executive Dario Amodei's prediction that AI could eliminate nearly half of all entry-level white-collar jobs in fields such as tech, finance, law and consulting.
Labour economists are more cautious still. Christopher Stanton, who studies AI in the workplace at Harvard Business School, says it is "too early to tell" whether the technology will cause net job losses, noting that about 35 percent of white-collar tasks overlap with what AI can do, but that automating some tasks within a job often reshapes the role rather than erasing it. He points to radiologists, whom many expected AI to replace a decade ago, and who are busier than ever as the cost of imaging has fallen. The World Economic Forum's Future of Jobs report projects 92 million jobs displaced globally by 2030 but 170 million created, a net gain.
What the layoff figures do and do not show is the crux of the dispute. Challenger's tally relies on public announcements and employer disclosures, so it captures what companies choose to attribute to AI, not what automation independently caused. Analysts at the Society for Human Resource Management note that AI-attributed cuts, though real and rising, remain a minority of total layoffs, and it is unclear how much reflects genuine automation rather than cost-cutting and restructuring relabelled under an AI banner during a soft hiring market.
Where early effects are visible, they are uneven rather than universal. Entry-level analyst and junior software roles show the highest exposure, while more senior work built on judgment and stakeholder management is harder to automate. Researchers warn the burden may fall unequally by gender: the Kenan Institute found eight in ten women in the US workforce are in occupations highly exposed to AI-related cuts, against six in ten men, because AI's current strengths overlap most with administrative and support work where women are overrepresented.
The result is a labour market defined by anxiety running well ahead of evidence. Public fear is climbing, the layoff tallies give it a number to point to, and the executives who stoked it are now hedging. Three questions remain open: how much of the AI-attributed layoff figure is genuine automation, whether demand for generalist roles will fall or simply shift as the tools mature, and why public fear so sharply exceeds the net job gains that institutional forecasters still expect.