The Layoffs I Called a Cover Story Are Starting to Look Like the Real Thing
A month ago I called the “AI” in your layoff email an alibi. July’s numbers made me reopen the file — here’s what changed my mind, and what a two-track labor market means for you.
A month ago I told you the robots were an alibi — that "AI" in the layoff email was mostly corporate housekeeping in a costume. Then the July numbers landed on my desk, and I had to reopen the file. Here's what changed my mind, what I still stand by, and what a two-track labor market means for the job you're holding right now.
Let me do the thing writers are supposed to dread and eat my own words in public.
A month ago, in these pages, I called the layoffs a cover story — corporate housekeeping wearing a robot costume. My argument was tidy and, I thought, well-armored: when a company blames "AI" for cutting your job, it usually isn't the software that replaced you. It's the story about the software that gave a nervous executive permission to do the cost-cutting he already wanted to do. "The software hadn't replaced her," I wrote about a woman who'd lost her job to a chatbot that couldn't do it. "The story about the software had." I still think that sentence was true when I wrote it.
But I've spent a career telling readers that intellectual honesty means updating your priors when the evidence moves — and the evidence has moved. So this is not a victory lap. It's a correction, or at least a revision, filed in the same place I filed the original claim. I may have gotten part of it wrong. Let me show you exactly which part, and why the new data made me flinch.
What the July numbers actually say
Let's put the facts on the table before we argue about what they mean. I'll show my work, because "trust me, the vibe shifted" is not an argument, and because a few of these figures genuinely surprised me.
Start with the counters. Challenger, Gray & Christmas — the outplacement firm that has tracked American layoff announcements for decades — reports that AI has now been named in roughly 102,000 announced cuts this year, about a quarter of all job cuts, making it the leading stated reason for layoffs four months running. This isn't a one-month blip. Back in the spring, CBS was already reporting that AI drove more than one in four job cuts in April — 21,490 of 88,387 — and it hasn't relinquished the top spot since. John Challenger, who runs the firm, put it in words that stuck with me: it's making an impact, he said, "in a way that no technology has before."
Now, I'd normally treat a self-reported reason with a raised eyebrow — companies say "AI" for the same PR reasons I flagged last month. But here's the part that's harder to wave away, because it doesn't come from a press release. It comes from the payrolls themselves.
According to a Bloomberg analysis carried by Insurance Journal, the two sectors where AI adoption is furthest along — finance and information — are now losing about 28,000 jobs a month in 2026, even as the rest of the economy keeps hiring. Sit with the shape of that. It's not that the whole labor market is cratering. It's that a specific slice of it, the slice where the machines are best deployed, has flipped from growth to contraction while everything around it grows.
And then there's the study that made me put my coffee down. Stanford's Digital Economy Lab published a paper with the bleak-but-apt title "Canaries in the Coal Mine," and its finding is precise in a way the headlines usually aren't: early-career workers, ages 22 to 25, in the most AI-exposed occupations have seen a 16 percent relative decline in employment. The declines aren't spread evenly across every job a computer touches. They cluster where the software automates instead of assists — and they go quiet where it merely helps. That distinction — automate versus augment — is the whole ballgame, and I'll come back to it.
So here's the honest reckoning. A month ago I said the AI in the layoff email was mostly narrative. The narrative part is still real. But underneath it, in the finance and information sectors, in the entry-level rungs of AI-exposed jobs, something more than narrative is now showing up in the government's own numbers. The cover story and the real thing have started to converge.
"But the aggregate is fine" — the case I have to answer
Before I let anyone (including me) declare the robots victorious, I owe you the strongest version of the counterargument, because it's genuinely strong and I used to make it.
Zoom all the way out, and the aggregate numbers still show stability, not disruption. That's the read from the Yale Budget Lab, whose economist Martha Gimbel has said, bluntly, that no matter how you slice the data, "it just doesn't seem like there's major macroeconomic effects here." The unemployment rate has not spiked. The economy, in the aggregate, is still adding jobs. If a technology were truly eating the labor market whole, you'd expect to see it in the top-line figures — and you don't. Not yet.
This is the AI-washing thesis, and it isn't some fringe skeptic's cope. It's the argument I made last month, and it has powerful adherents — including, remarkably, the man selling the technology. Even Sam Altman admits some of it is "AI washing": the OpenAI chief has conceded that some companies are "blaming AI for layoffs that they would otherwise do." A cost cut is a cost cut. Attach a fashionable cause to it and the same pink slip reads as visionary restructuring instead of a bad quarter.
So which is it — cover story or real thing? Here is the uncomfortable answer I've landed on: both, at once, and that's precisely why it's confusing. Altman didn't stop at the concession. He finished the sentence: there's AI washing, "and then there's some real displacement by AI of different kinds of jobs." The aggregate can look calm while a specific sector quietly drains, the same way a lake looks placid while a current pulls hard beneath one particular dock. The stability camp is measuring the lake. Stanford and the payroll data are measuring the current. They're not contradicting each other. They're looking at different depths.
Meanwhile, in the same economy, two labor markets
Usually this is the part where I take you abroad — to how they do it in Europe, or what Singapore tried, or the lesson from some other jurisdiction that saw it first. Not today. Today the contrast isn't across an ocean. It's inside a single set of monthly numbers, and it's the most important thing I can show you.
Because here is the split, running through one economy at the same moment. The wider American labor market added more than 113,000 jobs a month this year through May. The finance and information sectors shed 28,000 a month. One country. One calendar. Two labor markets pointing in opposite directions — and the arrow that points down is the one aimed at the desks where AI works best.
The consultancies have a name for this. PwC calls it a two-track labour market, where the jobs AI touches split into winners and losers rather than simply vanishing. Their barometer found that jobs demanding specific AI skills are growing almost eight times faster than the overall market — 69 percent against 9 — and command a wage premium of up to 62 percent. The companies best able to use AI are growing headcount faster than the ones that can't. So it isn't a story of pure destruction. It's a story of divergence: if you're on the track that rides the machine, the machine is a raise. If you're on the track the machine replaces, it's a countdown.
And the concentration is the tell. HR Dive reports an 83% jump in tech-sector cuts year over year — 139,156 announced in the first half of 2026, against 76,214 a year earlier — in the very industry that builds the tools. When the sharpest pain lands first on the people closest to the technology, "coincidence" gets harder to say with a straight face.
Now imagine the next three years
Let me get speculative for a moment, because you hired me to look around corners, not just read the odometer.
Right now the divergence is legible mostly to people who read labor reports for fun (guilty). But extrapolate the two tracks forward and the picture gets vivid. Picture a mid-size insurance firm in 2029. Its underwriting floor, which employed sixty analysts in 2025, employs twelve — not because anyone was marched out in a dramatic purge, but because every time someone retired or left, the role simply wasn't refilled; the model absorbed the work. There was never a layoff to blame on AI. There was just a hiring req that quietly never got posted. The cuts you can count on a Challenger report are the visible edge of this. The invisible edge is the job that's no longer created — the entry-level seat that used to teach a twenty-three-year-old how the business actually worked, now dissolved into a prompt.
I warned about exactly this before the aggregate data caught up. Earlier I argued the machine hadn't taken the whole job, it took the first rung of the ladder — the routine apprenticeship work that used to turn a junior into a senior. Now watch the two forecasts collide. If Stanford is right that the damage concentrates where AI automates the young, and if PwC is right that the top track pays a 62 percent premium for AI skill, then we are engineering a labor market where the ladder loses its bottom rungs precisely as the top rungs get more lucrative. A great place to already be. A brutal place to be starting out.
That's the future I'd bet against with policy and prepare for without it.
What the smart people across the spectrum are saying
Here's where I try to keep myself honest by refusing to cherry-pick the doomers.
The optimists have a serious case, and it deserves your ear. BCG argues that most work will be reshaped, not replaced — its central projection is that 50 to 55 percent of US jobs will be meaningfully changed by AI within two to three years, but changed, not eliminated. That's the hopeful frame, and it's not wrong. It's just cold comfort to the roughly 12 percent of jobs BCG files under "substituted" — the ones where the machine takes over the core task and demand doesn't grow to soak up the displaced worker. Reshaping is a promotion for most and a pink slip for some, and averages have a way of hiding the some.
Then there's the stability camp — the Yale Budget Lab's insistence that the macro data is calm — which I take seriously enough to have built last month's essay on it. And the industry counters, from PwC's two-track optimism to Challenger's grim monthly tallies. What strikes me is that these are not really factions disagreeing about the facts. They're honest observers standing at different distances from the same event. Up close: real displacement, concentrated and cruel. Far away: an economy still standing. Both true. The mistake — my mistake, last month — is to let one distance cancel the other.
What does this mean for you?
Enough diagnosis. If you take nothing else from this, take a plan. Here's what I'd actually do, whichever track you're on:
Find out honestly whether your work is automated or augmented. This is the single most important distinction in the data. If AI does your core task for you, you're on the exposed track; if it does the task with you, you're likelier on the rising one. Be brutal in the assessment — the comfortable answer is the dangerous one.
Move toward the augment side on purpose. The 62 percent wage premium PwC found isn't for knowing that AI exists; it's for wielding it. Become the person who directs the tool, not the person whose output the tool reproduces. That's a deliberate skill, and you can start acquiring it this quarter.
If you manage juniors, defend the first rung. The apprenticeship tasks AI is eating are how the next generation learns your business. Automate them all away and you win this year's efficiency at the cost of never growing another senior. Keep some grunt work human on purpose.
Don't be reassured by the top-line unemployment rate. It's measuring the lake, not your dock. Watch your own sector's monthly numbers instead — finance and information are the warning, not the exception.
Treat "we're restructuring with AI" with calm skepticism — but not dismissal. Some of it is still cover story. Some of it is now the real thing. The mature move is to stop needing it to be only one.
The lesson, as I see it
So did I get it wrong? Partly. I was right that "AI" is often a narrative bolted onto ordinary cost-cutting — Altman himself confirms the washing is real. But I was too confident that the narrative was all there was. It wasn't. Beneath the story, in the sectors where the software genuinely automates rather than assists, real displacement is now visible in numbers no press office wrote. The cover story and the real thing didn't turn out to be rivals. They turned out to be roommates.
The lesson I'm taking isn't "AI is destroying work" — the aggregate won't let me say that, and I won't pretend otherwise. It's subtler and, I think, more useful: we are not living through one labor market anymore. We're living through two, and which one you wake up in depends less on how hard you work than on whether the tool at your elbow was built to help you or to replace you. That's not a fact to panic about. It's a fact to navigate — with clear eyes, an honest inventory of your own tasks, and a refusal to be soothed by an average that isn't about you.
I updated my thesis in public because that's the deal I make with you every time you open one of these. When the data moves, I move. The alternative — defending last month's certainty against this month's facts — is exactly the kind of institutional dishonesty I spend my life criticizing. I'd rather be caught changing my mind.
If this one made you rethink something you were sure about a month ago — good, welcome to the club, I wrote it from inside that feeling. Pass it to the person in your life who's either riding the fast track or watching the slow one, and let's argue about which track we're on together. That's the whole point of the HAIA Foundation: humans and AI figuring out how to flourish side by side, out loud.