Google, Meta, IBM, Ford and others are quietly rehiring workers they replaced with AI. What failed? What did it cost?
Between 2024 and 2025, the enterprises were narrating a clean story. Fewer people. Lower costs. AI handling the volume. Boards were satisfied. Investors applauded. Some of the world’s most recognisable companies like Google, Meta, IBM, Ford, and Salesforce cut deep into their workforce under the banner of AI-driven efficiency. Meta alone cut around 8,000 employees in May 2026,10 percent of the workforce, following earlier rounds that had already removed more than 21,000 jobs. Amazon eliminated 30,000 corporate roles across two rounds.
The logic behind these job cuts was simple: AI could do the work, so why pay people to do it?
However, something eventually changed the course of this trend, leading to its reversals.
According to multiple public reports, companies such as IBM, Google and Meta have added undisclosed numbers of workers in redefined roles to help steer their generative AI services.
Job listings for marketers, copywriters, content moderators and human resources administrators have shown partial revival as employers restaff previously eliminated roles. According to talent firm Robert Half, 32 percent of US hiring managers who eliminated a role primarily due to AI have already rehired for the same or a similar position. Workforce firm Careerminds found that in one of every three cases, the restaffing cost more than the original layoffs saved.
So what actually went wrong?
The anatomy of the mistake
IBM’s experience is the clearest illustration of the core problem. The company’s AskHR system handled approximately 94 percent of routine HR requests without issue. The remaining 6 percent including cases that called for ethical judgment, still needed a person. Performance disputes, medical accommodations, whistleblower concerns, complex terminations; everything was beyond the jurisdiction of an AI.
At IBM’s scale, operating across 170 countries, that 6 percent represents tens of thousands of annual cases, precisely the ones carrying the highest legal and human risk. The AI was not able to handle them; the people who had been handling them were gone. IBM has since announced plans to triple US entry-level hiring in 2026.
“If we don’t continue to invest in entry-level hires, what happens in three to five years? There’s no pipeline. The well simply dries up,” said IBM’s Chief Human Resources Officer Nickle LaMoreaux.
The pattern repeated across the industry. Google and Meta discovered that content moderation, a function both had trimmed aggressively for could not be fully automated. AI systems flag obvious violations at scale. They cannot navigate cultural context, local regulatory nuance, or the policy edge cases that determine where a line sits in a given jurisdiction. Both tech companies have been quietly restaffing in areas like content moderation, digital marketing, and specialised human resources roles, realising that automated AI systems require immense human oversight to prevent catastrophic platform errors.
Ford’s failure was the most operational tangible example. Ford spent the past three years hiring 350 veteran engineers to address vehicle quality problems that automated systems had failed to catch.
Charles Poon, Ford’s Vice President of Vehicle Hardware Engineering, said the company had wrongly assumed that feeding AI its design requirements would produce a high-quality product without experienced human oversight. The engineers it had paid to exit had to be rehired — at market rates that had moved in their favour.
The real cost was never measured
Every organisation that reversed course made the same initial error of putting their focus on measuring what was easy to measure and ignored what was not.
With reduced headcounts, salary lines and cost per transaction, the numbers looked like ROI. What they failed to capture was the cost of errors at scale, degraded service quality, regulatory exposure from AI handling decisions it was not built for, and the institutional knowledge that walked out the door with every departing employee.
In some cases, organisations found that smaller teams struggled to maintain existing levels of quality, oversight, or customer responsiveness once those roles were eliminated. The AI did not know the company’s culture. It did not carry the client history. It could not read the unwritten rules that experienced employees navigate without thinking.
The ultimate lesson
Some organisations moved quickly from ‘AI can assist this work’ to ‘AI can replace this work,’ and they are now recalibrating. That recalibration is expensive. It is also clarifying. Returning workers are now commanding salary premiums of 20 to 35 percent above pre-layoff levels, because the market has repriced the skills these organisations demonstrated they cannot function without. One in three employers who restaffed spent more on rehiring than they saved from the layoffs.
Gartner now projects that by 2027, half of all companies that cited AI for workforce cuts will rehire for similar roles — often under new titles that explicitly acknowledge what these organisations learned the hard way.
AI replacing humans and AI working alongside humans are not the same strategy. The companies that confused the two have spent two years and considerable investment finding out why.