The AI Jobs Apocalypse Just Failed Its First Contact With the Evidence

Imagine a chief executive standing before his investors after a bruising quarter. He overhired during the pandemic, bet on products that failed, and now he must cut 8,000 jobs. He has two scripts available. The first script admits error: we misjudged demand, we misallocated capital, and workers will pay for our mistakes. The second script sounds like the future: artificial intelligence has transformed our operations, and we are restructuring for the age of automation. Which script would you choose? The second one, obviously. It converts a confession into a prophecy. And that choice, repeated across dozens of boardrooms, is the raw material from which the media has manufactured an AI jobs apocalypse that the actual labor market stubbornly refuses to deliver.

Begin with the evidence, because the apocalypse thesis is, at bottom, an empirical claim. It predicts fewer jobs, rising unemployment, and collapsing demand for human labor. None of that has happened. As of the June 2026 Employment Situation report, nonfarm payrolls grew by 57,000 jobs, the unemployment rate stood at 4.2%, and average hourly earnings were up 3.5% over the year. Professional and business services, precisely the white-collar category the doomsayers said would be gutted first, added 36,000 jobs in June and 172,000 jobs since October 2025. The May 2026 JOLTS data showed 7.6 million job openings, 5.2 million hires, and a layoff and discharge rate of 1.1%, a figure that describes ordinary churn, not a machine-driven purge. A skeptical reader might ask whether the damage is simply hiding beneath the aggregates, concentrated in AI-exposed occupations. Researchers have looked. The Yale Budget Lab examined the labor market 33 months after ChatGPT’s release and found no sign that AI exposure, automation, or augmentation was related to changes in employment or unemployment, and no discernible broader labor-market disruption at all. Brookings, hardly a MAGA outfit, published a paper titled “New Data Show No AI Jobs Apocalypse, for Now” and reported that if AI were automating jobs at scale we would expect fewer workers in high-risk occupations, but the data showed the opposite: stability, not disruption.

Why would a technology this powerful fail to destroy jobs? The question assumes the wrong model of how technology and labor interact. The correct model is simple, and conservatives have been explaining it for decades. When a worker gains a tool that lets him produce more per hour, his labor becomes more valuable, not less. Higher productivity lowers the firm’s cost per unit of output. Lower costs let the firm cut prices, improve service, enter new markets, and pursue business that previously sat below the profitability line. Growing firms hire. Mario Loyola of the Heritage Foundation compresses the whole chain into one sentence: technology increases labor productivity, which increases investment and wages, which increases demand, which increases employment. Oren Cass of the Manhattan Institute, a man nobody mistakes for a Silicon Valley cheerleader, made the same point during the last automation panic: technology is not the culprit behind job loss, and robots can be workers’ best friends.

History keeps vindicating this model. Consider the bank teller. Cash dispensing was among the most automatable tasks in retail banking, and between 1995 and 2010 the number of ATMs in America quadrupled from roughly 100,000 to 400,000. Teller employment should have collapsed. Instead it rose, from about 500,000 in 1980 to 550,000 in 2010. The machine lowered the cost of operating a branch, urban branches grew by 43%, and more branches needed more people for lending, service, and relationship banking. Or consider the 1990s, when computers were supposed to hollow out the office. James Bessen found that occupations using computers grew faster, not slower, and the decade delivered one of the strongest labor markets in modern history with unemployment near 4%. MIT’s David Autor, one of the country’s most respected labor economists, states the general principle: technology automates, it complements, and it creates new expertise and new work. Goldman Sachs, an institution once associated with alarming AI projections, now concedes that predictions of technology reducing the need for human labor have a long history but a poor track record, notes that 60% of current US workers hold occupations that did not exist in 1940, and finds no significant correlation between AI exposure and layoffs, unemployment, hours, or earnings growth.

The AI-specific evidence fits the historical pattern precisely. The landmark study “Generative AI at Work,” published in the Quarterly Journal of Economics, followed 5,172 customer support agents and found that access to AI raised productivity by 15%, with the largest gains going to the least experienced and lowest-skilled workers. Agents with two months of tenure performed like agents with six months of tenure. Customer treatment improved and attrition fell. Pause on what that means. When AI shortens the ramp-up period, hiring a new worker becomes cheaper and less risky, which makes firms more willing to hire, not less. The St. Louis Fed estimates that workers using generative AI save 2.2 hours per week and are 33% more productive during AI-assisted hours. The Philadelphia Fed surveyed businesses and found that roughly 70% of AI-using firms reported no impact on their need for workers, 47% expected retraining, and zero manufacturing firms anticipated AI layoffs. MIT Sloan and Brookings both find that AI-adopting firms grow: 6% higher employment growth and 9.5% higher sales growth over five years, along with more trademarks and more product patents. The tool makes the worker more valuable, the worker makes the firm more competitive, and the firm builds the next layer of jobs around both.

So where does the apocalypse narrative come from? Follow the incentives, and three culprits emerge. The first is the press, which committed to the “AI kills jobs” storyline in late 2022 and now filters every corporate announcement through that script. The second is the executive class. Challenger, Gray & Christmas reported that AI was cited in 15,341 announced job cuts in March 2026 and 21,490 in April, and headlines treated those citations as causation. They are nothing of the kind. A layoff announcement is a press release, not a forensic audit. MIT professor Paul Osterman said it plainly: AI is a perfect excuse to justify big layoffs, because it makes it seem as if it is not our decision, our fault, it is the technology. Barron’s reported that analysts looking at firms citing AI, including Snap, Block, Atlassian, and Dow, pointed instead to pandemic overhiring and the convenience of the AI explanation. Tellingly, announced job cuts fell to 45,849 in June 2026, the lowest level since December 2025 and down 53% from May. Genuine technological waves do not ebb with the news cycle. Managerial fashions do.

The third culprit is the most important, because it explains why the panic persists even as the AI executives themselves walk it back. Sam Altman told a Harvard Business School audience that in every technological revolution people predict the end of jobs, and it never happens, and by 2026 he was telling reporters he did not expect a jobs apocalypse at all. Yet the frontier labs continue to benefit from fear, because fear creates demand for regulation, and regulation is a moat. OpenAI’s written Senate testimony declared that regulation of AI is essential and urged licensing or registration requirements for models above capability thresholds. Anthropic’s policy framework proposes government authority to block deployments, public testing regimes, civil penalties, and rules keyed to compute, revenue, and R&D thresholds. Ask yourself who can afford that world. The Competitive Enterprise Institute answers directly: large firms can absorb regulatory costs that smaller competitors cannot, which means regulation raises barriers to entry and reduces competition. Cato and Mercatus issue the same warning. The labs that warn loudest about AI taking your job are, not coincidentally, the labs proposing rules that would guarantee they face fewer challengers. Senator Ted Cruz identified the stakes at a Commerce Committee hearing: the way to beat China in the AI race is to outrace them in innovation, not to saddle American developers with European-style regulations. Europe chose the compliance path and produced no frontier labs worth naming. China will not pause out of sympathy for our anxieties.

Which brings us to the moral heart of the matter. The apocalypse narrative is not merely wrong. It is a demand, dressed in compassion, that ordinary Americans accept a slower, more regulated, less competitive economy so that a small circle of incumbents, regulators, and journalists can control a technology they did not build and do not use on a loading dock or in a call center. The worker is the moral center of this story, and the worker is doing fine, better than fine, he is picking up a tool that makes his hours more valuable, his training faster, and his employer more likely to grow. Every previous generation of American workers was told the machines were coming for them, and every previous generation ended up richer, more productive, and employed in occupations their grandparents could not have imagined. The evidence is in, the openings number in the millions, the productivity gains are measurable, and the doomsday clock keeps getting quietly reset by the very people who wound it. Do not let them frighten you into a smaller future. The apocalypse thesis met the data, and the data won.

If you enjoy my work, please subscribe https://x.com/amuse/creator-subscriptions/subscribe

---

Alexander Muse is a Fellow at the John Milton Freedom Foundation and publishes daily political analysis at amuseonx.com. Primary sources cited in this piece are linked inline; campaign finance figures are drawn from FEC filings, polling data from publicly released crosstabs, and legal claims from filed pleadings. Corrections are posted to the original URL with a dated changelog. Readers who identify errors are invited to contact the author directly. Data in sponsored partnership with Polymarket.