Two years of AI-first layoffs are coming back around. Two in three companies that cut jobs for AI are already rehiring, and a third say the rehiring cost more than the layoffs ever saved. Here's what actually broke, and how to use AI without getting hit.
MJBy Micah Johnson · Biggest Goal7 min readFord · Klarna · IBM · Gartner
Late last month, Ford quietly rehired about 350 veteran engineers, many of them retirees, after admitting its AI quality tools couldn't do the job alone. It's the clearest sign yet of a pattern every leadership team should be watching: the AI layoffs of the last two years are boomeranging, and the companies that cut deepest are the ones now scrambling to hire people back. If you're under pressure to trade headcount for AI, this is the story to read first.
Ford isn't an outlier. It's the flight of a boomerang. And for anyone who isn't careful, that boomerang comes back around and clocks them right in the face.
The promise that sold the layoffs
For two years the pitch to every executive was irresistible: AI does the work faster, cheaper, around the clock, and it never quits. All true. But it came bundled with a quieter pitch: cut the headcount, book the savings, tell the board you're "AI-first."
Klarna said it out loud. Between 2022 and 2024, the fintech eliminated roughly 700 customer service roles and handed the work to an OpenAI-built assistant that, at its peak, handled two-thirds of all customer conversations. The efficiency numbers looked incredible. For a while, this looked like the future: fire the humans, install the software, watch the margin expand.
Then the bill came due.
The reversal
By mid-2025, Klarna was hiring humans back. Customer satisfaction had dropped, complaints climbed, and the AI produced generic, repetitive answers on exactly the interactions that mattered most: disputes, fraud claims, financial hardship. CEO Sebastian Siemiatkowski put it plainly: the company had "focused too much on efficiency and cost," and the result was "lower quality." Klarna now runs a hybrid model. AI on routine volume, humans on anything where a wrong answer is expensive.
Ford's version is the same lesson in a different industry. Its own executives admitted the company "mistakenly thought that by just introducing artificial intelligence and ingesting the design requirements we had, that would produce a high-quality product." It didn't. The rehired "gray beard" engineers now run mandatory weekly design reviews and retrain the very AI tools that were supposed to replace them.
This isn't two anecdotes. It's the shape of the whole curve.
55%
Of execs will regret replacing staff with AI within 18 months
2 in 3
Companies that made AI cuts are already rehiring
50%
Of firms that cut service jobs to AI will rehire by 2027
Forrester projects that 55% of executives who replaced employees with AI will regret it within 18 months. A February 2026 Careerminds survey of 600 HR professionals who ran layoffs found two in three companies that made AI-driven cuts are already rehiring, more than a third have brought back over half the roles they eliminated, and 52% started within six months. Gartner forecasts that by 2027, half of the companies that cut customer service jobs because of AI will be rehiring for the same work.
The math didn't survive contact with reality
The layoffs were justified with cost savings. In the Careerminds data, most of those savings evaporated. Only about a quarter of companies came out ahead. For everyone else, the rehiring either erased the savings or cost more than the cut ever saved.
31%
42%
27%
Cost more than saved
Rehiring outran the layoff savings entirely.
Roughly a wash
Savings and rehiring costs canceled out.
Net ahead
The cut actually paid off.
Outcome of AI-driven layoffs · Careerminds survey of 600 HR leaders, Feb 2026
What actually broke
The tempting read is "the AI wasn't good enough yet." That's mostly wrong. The models are extraordinary. What broke is the thing AI is genuinely, structurally bad at: judgment.
AI learns from data. But the knowledge that runs your business, why this customer gets an exception, which spec is technically fine but will fail in the field, how to read a tense negotiation, was never written down. Researchers call it tacit knowledge, and by definition it doesn't live in any database a model can train on. As one Forbes analysis put it, AI "isn't built for ambiguity, emotion, or edge cases," and the clearest failures come from businesses that removed human judgment too early.
That's the pattern in every reversal above. AI matched or beat humans on the routine, high-volume, well-documented middle of the work. It fell apart at the edges: the unusual case, the emotional customer, the design flaw no requirement document anticipated. And the edges are exactly where the expensive mistakes live.
People aren't perfect at judgment either. But a seasoned employee brings something the model can't: accountability, context they've never articulated, and the instinct to notice when something is off even when it technically passes. Ford didn't rehire its veterans to write more code. It rehired them to catch what the AI couldn't see.
AI raises the floor of your operation. It does not raise the ceiling. The ceiling is still set by human judgment. Cut the humans and you don't just lose hands, you lose the ceiling.
Same technology. Opposite strategy.
Contrast the boomerang companies with IBM, which is probably the smartest version of this story. IBM automated 94% of routine HR tasks with an AI agent and did cut a few hundred HR roles. But total employment at IBM went up, not down. CEO Arvind Krishna's explanation: automating the routine "gives you more investment to put into other areas," so they redeployed the savings into programmers and salespeople doing higher-judgment work.
✕ Subtract
Cut the humans, keep the AI
Treat AI as a replacement. Quality drops on the cases that matter, complaints climb, and the panic rehiring erases the savings.
Klarna · Ford
✓ Redeploy
Automate the routine, move people up
Treat AI as a tool. Let it own the repetitive middle and redeploy people into higher-value work. Total headcount rose.
IBM · 94% of routine HR tasks automated
Same technology, opposite strategy, opposite result.
The companies that stumbled subtracted people. The companies pulling ahead treat AI as a tool and redeploy people toward the work only humans can do. Same technology, opposite result.
Before you cut anything
The expertise that makes AI work is the first thing a layoff deletes.
Ford's AI failed because the experts who could have trained it had already been asked to leave. The move that actually works is capturing that judgment and handing your people the tool, not the pink slip. Our free Cowork Masterclass walks you through setting AI up as an amplifier for your team, step by step, from your first context folder to a working setup they'll actually use.
Free · Self-paced · 26 short lessons
The teams that wield AI are pulling away
This isn't a "go slow on AI" piece. The evidence that AI-augmented teams win is just as strong as the evidence against replacement. The landmark study here, by Erik Brynjolfsson of Stanford and colleagues, tracked 5,179 customer support agents who used an AI assistant.
+14%Average productivity lift across all agents
+34%Lift for the newest, least-experienced workers
5,179Support agents in the field study
Read that carefully, because it's the opposite of the layoff logic. The AI worked by capturing what the best agents did and teaching it to everyone else. Its biggest gift isn't headcount you can cut. It's a floor you can raise: your newest people perform like veterans faster, your whole team levels up, and the humans stay to handle what the AI can't. A team that knows how to wield AI beats a team without it, and it also beats a company that fired its team and kept only the AI.
What this means for you
Four moves separate the winners from the boomerang crowd:
1
Capture the tacit knowledge before you automate
Get the expertise out of people's heads before you cut. The reason Ford's AI failed is that the experts who could have trained it were already gone. That judgment is your asset, not the software.
2
Keep humans on the high-cost-of-error decisions
Let AI own the routine, high-volume middle. Put people on the disputes, the exceptions, and the judgment calls where being wrong is expensive. That's the hybrid model Klarna rebuilt the hard way.
3
Augment, don't amputate
Before you cut a role, ask whether AI replaces the person or just the routine part of their job. Usually it's the latter, which means the smart move is redeploying that person to higher-value work, IBM-style.
4
Measure outcomes, not activity
The layoffs were justified with cost savings that mostly evaporated. Track quality, customer satisfaction, and rework, not how many tickets the bot closed or how much headcount you removed.
The lesson of the boomerang isn't "AI doesn't work." It's that AI is a power tool, and power tools make skilled people dramatically more productive and unskilled operations dramatically more dangerous. The companies rehiring in a panic bought the tool and fired the operators. The ones winning kept the operators and handed them the tool. That gap, between owning AI and knowing how to use it, is exactly the one we exist to close.
Yes. A February 2026 Careerminds survey of 600 HR leaders found two in three companies that made AI-driven cuts are already rehiring, and 52% started within six months. Ford rehired about 350 veteran engineers, Klarna rebuilt its human support team, and Gartner expects half of the firms that cut service jobs for AI to rehire by 2027.
Why do AI layoffs backfire?
AI matches or beats people on routine, well-documented work, but it falls apart at the edges: the exception, the dispute, the design flaw no spec anticipated. Those edges need judgment and tacit knowledge that was never written down, so companies that cut their experts lost the very thing that made the AI usable in the first place.
What is the right way to use AI instead of replacing staff?
Augment, don't amputate. Automate the routine, high-volume middle and redeploy people to higher-judgment work. IBM automated 94% of routine HR tasks and its total headcount still rose, because it moved people into higher-value roles instead of eliminating them. Keep humans on the decisions where being wrong is expensive.
Does this mean we should slow down on AI?
No. A study of 5,179 support agents found AI raised productivity 14% on average and 34% for the newest, least-experienced workers. The teams pulling ahead aren't cutting people, they're handing them the tool. The free Cowork Masterclass shows non-technical teams how to set that up.
Free Masterclass · $0
Own the tool. Keep the operators.
If you're under pressure to cut headcount for AI, or you already did and the cracks are showing, that's the conversation we have every week. The free Cowork Masterclass is the fastest way to give your team the thing the winners have: AI set up as an amplifier, not a replacement.
