The AI Layoff Undo Button

Remember eighteen months ago, when “AI efficiency” started showing up in layoff memos like a magic word? Welp, those companies were going to be quietly buying those exact people back, at a markup, before the year was out. And now they are.

Robert Half told CNBC that 32% of U.S. hiring managers who cut a role primarily because of AI have already rehired for the same or a similar position. Careerminds (shout out to my homies!) put it even more starkly: more than two-thirds of companies that made AI-driven cuts rehired at least some of the roles, and more than a third replaced more than half the positions they let go.

Orgvue found that 55% of the leaders who made people redundant because of AI later admitted the decision was wrong. That’s a whole bunch of executives standing in the return line with the receipt, hoping the store takes it back.

The company laid people off for AI, then rehired for the same job six months later at a premium. That’s not efficiency, that’s an expensive lesson.

The Rehiring Wave

See, the pattern goes: company announces AI will do the job, staff gets cut, six to twelve months pass, the AI handles maybe 60% of the work beautifully and face-plants on the other 40%, and the company hires humans back to do the part that was always going to need a human. Oh and also the folks you kept, whose workload doubled and learning curve steepened dramatically? They don’t like you no more.

IBM ran this exact loop with its HR function (paging Alanis Morisette!) — the AI handled 94% of routine requests and choked on the 6% that involved actual ethical judgment. Ford automated quality control, watched its veteran engineering judgment walk out the door, and rehired, promoted, or newly hired 350 experienced engineers, and whaddya know? They then topped J.D. Power’s Initial Quality Study for the first time since 2010. The lesson wasn’t exactly subtle. Their own VP said it straight: AI is only as good as the information you train it on, and the people who hold that information were the ones getting fired.

The Cost of Reversing a Layoff

God, this is so stupid. Anyone could see this coming! And here’s the part that should make any CFO put down the efficiency memo. The rehire is not free, and it is frequently more expensive than the original role. That $55,000 job you cut? It’s coming back at $75,000 or more, because now it requires someone who can do the work AND babysit the machine that was supposed to replace them.

You paid for the layoff.
You paid the recruiting and onboarding costs to backfill.
You’re paying a premium salary for a scarcer, AI-literate version of the same person.
And you lost the institutional memory; the months of client context and internal history that no onboarding doc captures, SBI.

Careerminds found the whole exercise cost 75% of these organizations more than it saved. Fewer than three in ten actually came out ahead.

Where We’ve Landed on This

We’ve been saying a version of this for a while, and not because it’s a comfortable thing to sell. When we wrote about leading with a human-first approach while half the agency world was going full-automation, the argument was never that AI is evil or bad. We use it prolifically, we advise on it, we’ve built production agents on it. What we’re trying desperately to get across (in the middle of never AI and full on automatons) is that AI is a phenomenal accelerator AND a catastrophic full replacement, and everyone seems to be confusing or conflating those two things! Now they’re paying for it, and despite my stock portfolio getting smashed to bits, I’m kind of loving it?

So what does an experienced operator actually do with this, beyond feeling smug?

What This Means for Operators

Reframe Headcount as Capability

First, reframe the entire conversation from headcount to capability. The failed layoffs share one root cause: leaders treated jobs as bundles of tasks, when jobs are actually bundles of judgment wrapped around tasks. I have prattled on about this ad nauseum with the Human Capability Model and hate to break it to you but I AM STILL RIGHT) AI eats tasks (and not in the cool way the kids are saying it.)

But y’all, it does not, CANNOT, eat judgment, context, or the relationship the customer actually wants. When you plan a cut, the question isn’t “can AI do 60% of this role?” It’s “what happens to the 40% that’s the reason the role exists.” If you can’t answer that crisply, you’re not cutting costs. Also, with all the self-help and business books that have come out just during my career, how did we forget the WHY of it all? Isn’t there literally a book called Start With Why?

AI eats tasks. It cannot eat judgment. Confusing the two is how you end up rehiring your own layoffs at a premium.

Follow the Money

Second, follow the money, honey, because it’s telling you where the work went. It didn’t up and disappear. It moved from production to supervision. Congrats, we made more managers. The median comp for the AI-wrangler roles companies are now scrambling to fill runs around $245,000 at the top firms. The budget didn’t get smaller. It got redistributed toward governance, oversight, and the humans who keep the machine from confidently lying to your customers. In B2B environments like this, that’s a hiring signal, not a hiring freeze. The reqs are opening under new titles.

Stop Drinking Your Own Kool-Aid

Third, stop drinking your own koolaid, specially the poisoned kind. IBM’s CHRO said the quiet part into a microphone: if you stop hiring entry-level because AI does the grunt work, there’s no pipeline, and the well simply dries up (which is a whole ‘nother article because, shocker, we never learn our lesson the first time do we?). The grunt work was the training ground where juniors became the seniors with experience that have suddenly become quite difficult to hire. Cut it at the source, and you’re not saving money, you’ll just have to fix it later. We connected these same dots when we wrote about a labor market that looks fine in the GDP numbers and is quietly collapsing underneath them the damage doesn’t show up in the top-line metric until it’s too late.

The Companies That Got It Right

Some companies figured out where the line sits between automation and augmentation, and staffed deliberately on both sides of it. IKEA automated half its call volume and kept all 8,500 workers, retraining them into design consultants — now their fastest-growing revenue line. Same technology. Opposite outcome. The difference was someone in the room understood what their people were for.

You won’t need an undo button if you never hit delete on the thing that was working.

This is a sensitive topic for a lot of people living through it, and I am sorry. The layoffs are real and the human cost is real. If that’s you or someone on your team, the point of this piece is that the “AI made you redundant” story was wrong more often than it was right. It was the jerk bosses.