Companies Are Rehiring the Workers They Replaced With AI. Here's What They Got Wrong
By Lexi Banks · · AI Strategy
In 2026, companies from Ford down are rehiring staff they cut for AI. Here is why the "replace people with AI" bet backfired, and how to deploy AI without it.
Key takeaways
- A measurable "AI boomerang" is underway in 2026: roughly a third of managers say they rehired for a role they cut for AI, and most regret the original decision.
- AI reliably covers the repetitive ~60% of a workflow and struggles with the ~40% that needs judgment, escalation, and ownership.
- The biggest hidden cost was institutional knowledge, not labor, as Ford's engineering reversal shows.
- Not every "AI layoff" was about AI; some was cover for cuts companies wanted anyway, which is its own warning.
- The lesson is not that AI does not work. It is that replace-then-bolt-on fails, while augment-in-place works.
A Ford VP just admitted the AI plan backfired
In June 2026, a Ford vice president said out loud what a lot of executives are thinking privately. Charles Poon, who leads vehicle hardware engineering, admitted the company moved too fast on AI and paid for it in quality.
"Mistakenly, we thought that by just introducing artificial intelligence and adjusting the design requirements that we had, that would produce a high-quality product," he said.
Ford ended up hiring, training, or rehiring more than 350 experienced engineers to rebuild the expertise it had thinned out. Those people were not brought back to fight the AI. They were brought back to mentor younger staff and make the AI systems actually work.
Ford is not an outlier. It is a visible example of a pattern that now has a name: the AI boomerang.
What is the "AI boomerang"?
It is the trend of companies rehiring for the exact roles they cut after assuming AI could replace them. The 2026 data is hard to wave away.
| Finding | Source |
|---|---|
| ~32% of hiring managers say they eliminated a role for AI or automation, then rehired for it | Robert Half |
| 55% of employers regret laying off workers for AI reasons | Forrester |
| ~50% of companies that cut service or operations roles for AI will restaff them by 2027 | Gartner |
This is not just sentiment. It is companies reversing their own decisions and paying twice to do it.
The precedent showed up early. Klarna said in 2024 that an AI agent could do the work of 700 support reps, then by 2025 walked it back and started rehiring for the complex, sensitive cases. 2026 is when that one-off became a category.
Why did replacing employees with AI backfire?
Because AI covered the routine work and missed the part that actually needed a person. The pattern people keep describing is a rough 60/40 split.
Automation handles the high-volume, repetitive 60% of a workflow well. It stumbles on the remaining 40%: the edge cases, the judgment calls, the escalations, and the moments a customer needs to feel heard.
When you remove the humans entirely, that 40% does not vanish. It resurfaces as quality problems, frustrated customers, and rework that quietly eats the savings.
Ford's version of the failure was institutional knowledge. The AI could follow the design rules it was given, but it did not carry the hard-won context that experienced engineers use to catch problems before they ship.
What did the cuts actually cost?
More than payroll. The bill showed up in places that never made the original spreadsheet.
- Lost institutional knowledge: the undocumented judgment that senior people carry in their heads.
- Quality and customer trust: the 40% AI could not cover degraded the experience.
- Rework and oversight: managers spent time cleaning up AI output instead of doing their own jobs.
- Paying twice: rehiring, often into hybrid roles that demand data and oversight skills, sometimes at higher pay than the jobs that were cut.
That last point is the quiet punchline. The "savings" became a more expensive workforce a year later.
Were these even real "AI layoffs"?
Not always, and this is the part most takes miss. Some of the 2026 cuts blamed on AI were not really about AI at all.
Paul Osterman, an MIT Sloan professor and author of Disposable Workers, argues that a lot of this is "AI-washing": AI as a convenient cover story for layoffs a company wanted to make anyway. "They've been saying that for 20 years," he told Fortune, pointing to cuts at firms like Wix, Block, Snap, Atlassian, and Cisco.
Why does this matter to you as an operator? Because it separates two very different failures. One is cutting people you needed and blaming the robot. The other is deploying AI badly and discovering the gap in production.
Both end in the same place: the boomerang. And both are avoidable if you are honest about which one you are actually doing.
What is the real lesson for operations leaders?
The lesson is not that AI does not work. That conclusion is as lazy as the layoffs were. The lesson is that the replace-then-bolt-on model is the wrong design.
When AI is treated as a headcount swap, it gets bolted on top of the org as a standalone bot, disconnected from the systems and people who hold the context. The routine work speeds up, and everything that needs a human falls through the cracks.
The companies that are not rehiring are the ones that used AI to make their existing teams faster, not to delete them.
How do you deploy AI without triggering the boomerang?
Start by augmenting the team you have instead of trying to replace it. A few principles separate the deployments that stick from the ones that snap back.
- Automate the repetitive 60%, keep humans on the 40%. Route exceptions, judgment, and relationships to people on purpose.
- Build AI into existing workflows, not on top of them. It should live inside the tools and processes your team already uses, not as a separate system that loses context.
- Keep institutional knowledge in the loop. Use AI to capture and extend what your best people know, not to push them out the door.
- Measure quality and customer outcomes, not just headcount saved. If CSAT or error rates move the wrong way, the savings are not real.
- Treat roles as AI-augmented, not deleted. The durable jobs belong to people who direct and check AI, not to people AI replaced.
How does Kalyxi approach this?
Kalyxi's model is the opposite of replace-and-hope. Our whole positioning is AI built into your existing operations, not on top of them.
In practice that means we automate the repetitive work inside the systems your team already runs, and we keep your people on the decisions that carry risk. You get the efficiency that made AI attractive in the first place, without gutting the expertise that makes the work good.
The boomerang is expensive because it treats people and AI as substitutes. The companies pulling ahead in 2026 treat them as a team. That is the work we do.
If you are weighing where AI fits in your operations, the better question is not who you can replace. It is what your people should stop doing by hand. We can help you answer it.