U.S. Employers Added 172,000 Jobs in May as Hiring Rebounds from 2025's Near-Freeze, but AI Tools Are Burning Out the Workers Who Survived It

Since this outlet's June 18 coverage of jobless claims — where continuing claims hit a three-month high even as initial filings fell — the broader labor-market picture has come into sharper focus through two new reports from The Atlantic.

The headline number: U.S. employers added 172,000 jobs in May, according to The Atlantic, spread across leisure and hospitality, local government, construction, manufacturing, and health care. That's a meaningful step up from the year's monthly average of 114,000 and a dramatic reversal from 2025, when employers added just 10,000 jobs per month on average.

What Actually Happened in 2025

The 2025 hiring drought had two plausible causes, and they're worth separating.

First, immigration enforcement. The Trump administration ramped up deportations aggressively after January 2025. The Congressional Budget Office estimated net migration last year at 410,000 people — roughly one-fifth of pre-Trump projections. The Brookings Institution believes the actual number could be even lower. Fewer arrivals means fewer new labor-force entrants, which helps explain why unemployment stayed relatively contained at 4.3 percent despite weak hiring. The labor pool simply shrank.

Second, tariff paralysis. Employers facing volatile, unpredictable trade policy responded the same way most rational actors do under uncertainty: they waited. Diane Swonk, chief economist at Goldman Sachs, told The Atlantic plainly: "There was no game in town other than health care in 2025."

Now some of that paralysis has lifted. Unemployment has stayed under 5 percent for roughly five years, and the May job numbers show diversification beyond health care. That's a real improvement.

What the Recovery Is NOT

The strongest cautionary argument against reading this as a boom deserves a fair hearing: 114,000 jobs per month is moderate, not strong. The post-COVID snapback of 2021 through early 2023 was far more aggressive. Continuing jobless claims, as covered here yesterday, hit a three-month high, suggesting some workers who lose jobs are taking longer to find new ones. A thaw is not a boom, and framing it as one would be inaccurate.

That concern holds up. The Atlantic's own framing is careful on this point, calling it a "cautious transition" rather than a recovery story. That's the honest read.

The AI Exhaustion Problem Nobody Is Counting

A separate Atlantic report raises a question that doesn't show up in any jobs data: what is AI actually doing to the people still employed?

Steve Yegge, a programmer and tech blogger, described to The Atlantic what happened when he shifted to managing AI coding agents instead of writing code himself. Productivity went up. So did exhaustion. He began experiencing midday "nap attacks" lasting 90 minutes. "I've fallen asleep slower at the anesthesiologist," Yegge wrote on his blog.

This isn't an isolated anecdote. Boston Consulting Group surveyed roughly 1,500 workers across major American companies and found widespread "mental fatigue from excessive use or oversight of AI tools beyond one's cognitive capacity." Workers described a "buzzing" and "fog"-like feeling, along with headaches, slower decision-making, and attention problems. One engineering manager compared it to having "a dozen browser tabs open in my head, all fighting for attention."

18 percent of developers in the BCG survey reported AI-induced exhaustion. Rates were even higher in HR and marketing, per the same survey — roles where AI has also taken over significant workflow.

The core problem: AI agents still require continuous human oversight. They ask follow-up questions constantly, need detailed instructions, and can cause serious errors if left unsupervised. Run several simultaneously, and there is no downtime. Workers aren't being freed up; they're being turned into supervisors of systems that never rest.

Two Stories, One Labor Market

These two reports describe the same labor market from different vantage points, and neither cancels the other out.

The macroeconomic picture — more jobs, lower unemployment, diversified sector growth — is genuinely better than it was twelve months ago. That's real. The 2025 freeze had specific causes (immigration contraction, tariff uncertainty) that appear to have partially resolved.

But the microeconomic experience for workers actually inside the AI-integrated economy is more complicated. Jobs that remain after automation aren't necessarily easier or less demanding. They may be harder in ways that don't show up in payroll statistics.

The BCG data also raises a productivity accounting question that no government agency is currently tracking: if AI tools are boosting output while simultaneously degrading cognitive function, what is the net gain, and who bears the cost?

That question doesn't have a sourced answer yet, which is precisely why it's the one worth watching as this hiring recovery continues.