The rehiring of employees who were let go because of artificial intelligence has become a reality that most people didn’t expect to see this soon.
And the stat that sums up this situation is pretty straightforward: nearly 1 in 3 hiring managers in the United States admitted they had cut someone claiming AI would handle the job — and then had to reverse course.
This isn’t an isolated phenomenon.
Companies like Ford, Commonwealth Bank of Australia, and IBM have already gone through this firsthand, and each of them learned the hard — and expensive — way that there’s a massive difference between automating tasks and replacing entire roles.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
What these cases have in common is easy to grasp:
- Automation worked great until the moment it didn’t
- The problems left behind were exactly the most complex ones
- And that’s when the lack of skilled people started hitting the bottom line
No fear-mongering here, and no romanticizing human labor either.
What we have is a real, documented pattern that’s worth understanding before making any decisions about layoffs and work quality in the age of AI. 🤖
What the numbers say about layoffs and AI
According to data cited by CNBC, the staffing firm Robert Half found that nearly 1 in 3 hiring managers in the United States had eliminated a position citing artificial intelligence as the reason, only to later bring someone back for that same role or an equivalent one. The reason wasn’t emotional regret. It was purely practical: deliverables stopped meeting expectations, the workload didn’t drop as predicted, and internal processes started losing consistency. The problem wasn’t the AI itself, but how the decision to cut people was made without a clear assessment of which roles could genuinely be replaced and which ones would simply benefit from technology as a support tool.
A report from consulting firm Orgvue, also mentioned by CNBC, painted this picture even more bluntly. Among business leaders who eliminated positions as part of an AI implementation — a group representing 39% of those surveyed — more than half, specifically 55%, later admitted the cuts had been a mistake. That number is telling because it doesn’t come from outside critics. It comes from the very decision-makers who pulled the trigger. When more than half of the people who bet on replacement openly acknowledge they got it wrong, it’s clear we’re not talking about a few outliers. This is a pattern repeating itself across different industries and companies of all sizes.
Another thing that stands out is that this cycle of layoffs and rehiring gained momentum right as AI-related cuts were climbing. According to outplacement firm Challenger, Gray & Christmas, artificial intelligence was the top stated reason for layoffs across all sectors in the United States for three consecutive months, and in May, AI-related cuts reached 38,579 — roughly 40% of all announced layoffs that month, the highest monthly total since the category started being tracked in 2023. So, at the very same time companies were cutting based on automation promises, many were already realizing they’d need to hire people back. 📊
Real-world cases that show where automation hit its limits
Ford is one of the clearest examples of when aggressive automation needs a course correction. Over the past three years, the automaker hired 350 veteran engineers to fix vehicle quality issues that automated systems simply couldn’t catch. Charles Poon, Ford’s vice president of hardware engineering, admitted the company had operated under the flawed assumption that feeding AI the design requirements would be enough to produce a high-quality product — no oversight from experienced professionals needed. The engineers who came back revamped the AI tools and took the lead on problem-solving sessions. CEO Jim Farley said this effort has already generated hundreds of millions of dollars in savings, thanks to lower warranty and recall expenses.
The Commonwealth Bank of Australia went through a different situation, but with the same root cause. Last year, the bank cut more than 40 customer service positions and replaced them with an AI-powered voice assistant, expecting to reduce costs while keeping service running smoothly. The opposite happened: the technology couldn’t handle the workload and actually drove call volumes up instead of bringing them down. Facing that reality, CBA reversed the cuts. The bank acknowledged in an official statement that it hadn’t adequately considered all the relevant business factors when it announced the layoffs. Work quality only returned to previous levels after the balance between humans and technology was restored.
IBM, meanwhile, went through a pretty telling chapter with an AI system deployed to take over human resources tasks. The tool managed to resolve about 94% of incoming requests, which at first glance looks like an excellent number. The problem was squarely in the remaining 6%, which involved situations requiring ethical judgment and sensitive decisions — exactly where full automation runs into its limits. In response, IBM announced plans to triple entry-level hires across all its U.S. business units in 2026. Nickle LaMoreaux, the company’s chief human resources officer, summed up the thinking at an event in New York: if the company stops investing in entry-level hiring, in three to five years there won’t be a talent pipeline left — the well just dries up. 🏥
Why work quality is the most honest measure of this equation
When companies evaluate artificial intelligence adoption with a sole focus on cost reduction, they tend to measure success by what they stopped spending, not by what they’re actually delivering. That blind spot is exactly where the problem starts to take shape. Work quality is the metric that takes the longest to show up in financial reports, but it’s the most honest indicator of what’s really happening inside operations. When a customer starts noticing that responses feel more generic, that errors are piling up, or that deadlines keep slipping, it doesn’t hit the balance sheet right away — but it shows up in retention rates, brand reputation, and eventually, revenue. The companies that got caught in the cycle of layoffs followed by rehiring are largely the ones that took too long to realize the right thermometer wasn’t short-term financials.
Automation works really well when the work it takes over is repetitive, predictable, and has low contextual variation. Processing structured data, sorting standardized documents, answering frequently asked questions with well-defined scripts — these are areas where artificial intelligence models deliver real and consistent gains. The problem shows up when the cost-cutting logic extends beyond those boundaries and starts reaching roles that look repetitive on the surface but in practice require constant adaptation. A writer producing technical reports, an analyst interpreting customer behavior, or an engineer solving quality issues on a production line aren’t doing cookie-cutter work — they’re exercising contextual judgment all the time, and that’s exactly the kind of judgment that still escapes current models when the bar is set high.
It’s also worth pointing out that some companies were criticized for using artificial intelligence as a public explanation for layoffs that were actually driven by other factors — a practice that some management experts have dubbed AI-washing. That makes the landscape even more complicated, because not every cut announced as an AI consequence was truly caused by it. What the cases at Ford, Commonwealth Bank, and IBM ultimately show is that technology and people, when properly positioned, deliver more together than either one can alone. The companies that learned this lesson the hard way are now structuring their processes differently: automation steps in where it makes sense, and people stay where judgment still can’t be replicated. 💡
