Ford just rehired 350 engineers.
The reason? The AI they'd deployed for quality control couldn't match what human engineers caught. After months of automation, the company reversed course — and brought back the people it had let go.
Ford isn't alone. A growing number of companies that cut workers in favor of AI are quietly walking it back. CNBC reported in early July that employers who laid off staff citing artificial intelligence are "already starting to regret it." Gartner now predicts that half of AI-driven job cuts will be reversed by 2027.
This isn't a story about AI failing. It's a story about expectations colliding with reality — on performance, and on price.
What AI Actually Does Well — and What It Doesn't
The pattern is consistent across industries. AI handles narrow, repeatable tasks beautifully. It can scan thousands of documents, flag anomalies in manufacturing data, or route customer inquiries faster than any human team.
What it can't do: exercise judgment when the situation doesn't match the training data. Recognize when a "defect" is actually a new design variation worth investigating. Handle the customer who's angry for a reason the chatbot wasn't programmed to understand.
Ford's quality engineers weren't just checking boxes. They were applying years of experience to spot problems that didn't fit a pattern. The AI caught the obvious stuff. It missed the subtle stuff. And in manufacturing, the subtle stuff is what causes recalls.
Klarna learned a similar lesson. The fintech company made headlines as an early AI adopter, proudly announcing it had replaced hundreds of customer service roles with automation. Then customers pushed back. They wanted to talk to a person. Klarna reversed course and started hiring again.
The Math Is Changing, Too
The performance gap is one problem. The cost is becoming another.
Fortune reported in May that internal Microsoft documents exposed an uncomfortable reality: running AI at scale is more expensive than paying human employees. An Nvidia executive put it bluntly — AI costs "far more" than the people it's supposed to replace.
The economics are simple. AI isn't a one-time purchase. Every query, every analysis, every customer interaction burns compute — and the bills compound. Meanwhile, the humans you laid off took their institutional knowledge with them. The engineer who knew which supplier's parts drifted out of spec in hot weather. The customer service rep who remembered the Johnson account always had billing issues in Q3. That knowledge doesn't transfer to a model, and replacing it costs more than most companies budget for.
What Smart Companies Are Doing Instead
The businesses getting this right aren't asking "what can we automate?" They're asking "where does AI actually help, and where does it create new problems?"
That's a different conversation. It requires understanding both the technology and the business context — what the tools can actually do, where they break, and what the real cost of a mistake looks like. It also requires running the numbers honestly: not just the license fee, but the compute cost, the integration work, the productivity dip during rollout, and the institutional knowledge you lose when people walk out the door.
Some tasks are perfect for automation. Others need human judgment, and the penalty for getting that wrong is measured in rehiring costs, damaged reputation, and lost time you can't get back.
The companies calling back their former employees learned that lesson the expensive way. The smarter move is to figure out the boundary before you cross it. If you're weighing automation against human expertise, talk to someone who understands both sides of that equation.