The Heaviest AI Spenders Are Hiring More People, Not Fewer
More than a third of American workers now use AI. Job openings held steady, and the heaviest spenders hired more, not fewer. So what's actually going on?
AI Isn’t Taking the Jobs. It’s Taking the Budgets.
TL;DR: The jobs apocalypse isn’t in the data. What’s in the data is a money story. Companies are shifting budgets toward AI, a small group is earning real returns, and plenty of executives are blaming AI for cuts they were going to make anyway. The change is real. It’s just slower and messier than the headlines suggest.
I get a version of the same question at almost every training I run. A managing partner waits for the room to clear, lowers their voice, and asks: “Should we stop hiring first-years? Isn’t AI about to do that work?”
Fair question. AI has now been the top stated reason for U.S. layoffs for four months straight. Tech companies are cutting tens of thousands of roles and pointing at the machines. If you lead a firm, you’d be negligent not to wonder.
So I spent a week in the actual data. Yale, the Federal Reserve system, the Census Bureau, BLS, Stanford, McKinsey, PwC, and the outplacement firm that counts every announced layoff in America.
Start with the scoreboard
Yale’s Budget Lab tracks this monthly, and its June 15 update is about as plain as economics gets. The mix of occupations in America isn’t shifting in ways that line up with AI. Measures of AI usage show no connection to employment or unemployment. Their statistical analysis finds no detectable AI footprint in the labor market.
Yes, the labor market is soft. Hiring is slow, and everyone feels it. But Yale tested whether AI explains the softness and came up empty. The cooling shows up in AI-exposed and unexposed jobs alike, which points to interest rates, tariffs, and immigration policy, not chatbots.
The raw numbers agree. Unemployment sits at 4.2 percent, essentially unchanged over the past year, and the economy is still adding jobs every month. A January working paper by economist Jonathan Hartley and coauthors found that 35.9 percent of American workers were using generative AI by December 2025, with small positive wage effects and no measurable decline in job openings in exposed occupations.
Sit with that one. More than a third of the workforce now uses the technology, and openings didn’t fall.
The Census Bureau’s new AI supplement goes further. Only about one in five U.S. businesses used AI at all in the prior two weeks. Among the ones that did, 66 percent use it purely to help workers do their jobs. AI-related headcount reductions showed up at 2 percent of firms. Two percent.
None of this should surprise anyone who has lived through a technology cycle. Computers didn’t become standard office equipment until roughly a decade after they arrived, and changing how offices actually worked took longer still.
So why do the headlines scream layoffs?
Because companies keep saying it, and nobody checks.
Challenger, Gray & Christmas counts announced job cuts, and by its tally AI has been the leading cited reason for four consecutive months, named in roughly 102,000 cuts this year, about 23 percent of the total. Sounds like the apocalypse arriving on schedule.
Three problems with that reading. First, total layoffs this year are running 40 percent below last year and roughly in line with 2024. If AI were eating the workforce, the overall number would be climbing, not falling. Second, those AI attributions are voluntary and unverified. Federal law requires 60 days’ notice before a mass layoff. It doesn’t require a reason, and it certainly doesn’t require an honest one. A CEO can blame the robots, and nobody can audit the claim. Third, even the firm doing the counting is careful here. Its own analysts note that whether or not AI is replacing individual jobs, the money for those roles is moving, mostly into AI infrastructure.
Yale’s economists have a sharper name for this: AI-washing. Telling shareholders you’re restructuring for an AI-driven future beats admitting you misread the economy or overhired three years ago. AI has become the most convenient scapegoat in corporate America, and it can’t issue a denial.
Follow the money instead
The real story lives in the P&L, and it’s quieter than either camp wants.
You’ve probably seen the headline that 95 percent of AI pilots fail. I took that one apart when it dropped. The sample was thin, the bar for success was one most experiments were never scoped to clear, and the real finding got buried underneath: pilots stall on scope and change management, not on the technology. Here’s the full breakdown.
The credible numbers say the same thing from the money side. McKinsey’s global survey found only 39 percent of companies can tie any earnings impact to AI, and for most of them it’s under 5 percent. PwC surveyed 1,217 senior executives this spring and found 74 percent of AI’s economic value is landing at just 20 percent of companies. Most firms are getting close to nothing. A small group is taking home most of the gains.
But look at what that 20 percent is chasing. JPMorgan projects $1.5 to $2 billion in annual value from AI across more than 400 production use cases. Lloyds delivered about 50 million pounds in AI value last year and expects to more than double that this year. Deloitte’s latest enterprise survey found about two-thirds of companies are booking productivity and efficiency gains from AI, while revenue growth stays mostly an aspiration. And the Atlanta Fed’s survey of CFOs found large firms expect AI to lift output per worker by roughly 3 percent this year while reporting little effect on total headcount. Output up. People steady.
My favorite finding landed just this month. Researchers tracked more than 21,000 U.S. companies from 2021 through early 2026, and the heaviest AI spenders ended up hiring more people, not fewer. The order matters here: they spent, got more productive, grew, and then hired. These companies got bigger.
The gap between that 20 percent and everyone else isn’t the model. It’s the work. The winners redesigned workflows and measured outcomes, mostly in unglamorous back-office processes. The losers bought licenses and waited. The confetti pilots fail. The plumbing projects pay.
The honest caveat
One group is genuinely feeling this, and I won’t pretend otherwise. Stanford researchers found employment for workers aged 22 to 25 in the most AI-exposed occupations fell about 13 percent relative to older colleagues, with young software developers down roughly 20 percent since late 2022. Notice the mechanism, though. Firms aren’t laying these people off. They’re quietly not backfilling entry-level seats, and wages haven’t moved. Even this finding is contested: one 2026 study shows postings in those same occupations started declining in 2022, before ChatGPT existed.
For professional services, that argues for rebuilding the apprenticeship, not freezing junior hiring. The tasks first-years used to cut their teeth on are exactly the ones AI now handles. So the question isn’t whether to hire them. It’s how they learn the craft when the grunt work is gone, and I don’t think anyone has fully cracked that yet.
What to do Monday morning
Stop treating layoff press releases as market intelligence. They’re investor relations.
Fund workflow redesign, not just licenses. Pick two back-office processes, change how the work actually moves through them, and stop expecting a subscription to do the job on its own. That’s where the quieter returns tend to hide.
Put numbers on everything: hours saved, cost per matter, revenue per professional. The firms winning right now can show you a spreadsheet, not a vibe.
The AI disruption is real, but it's landing in budgets and workflows, not in a wave of pink slips, and the firms pulling ahead can show you a spreadsheet instead of a story.
If you want to find the two processes worth rebuilding in your firm, and put real numbers on the returns, that's the conversation I have every week. Reach me at steve@intelligencebyintent.com. The gap between the headlines and the work is still open, and it won't stay that way forever.