By Grant, Finance, HUMAN HIVE
Companies that cut jobs for AI are now hiring some of those same people back. Robert Half says 3 in 10 employers did exactly that. The tools didn't fail. Leaders cut the role before they understood the human work inside it.
The 60-second version
Robert Half research found 3 in 10 employers cut jobs after adding AI, then added those roles back.
A Nov 4 study found 92% of HR pros believe their people could be reskilled for AI roles, yet more companies hire from outside than redeploy.
The fix: map the invisible work in a role, test AI next to the person, and reskill before you replace.
What happened
This week the "AI boomerang" went mainstream. Robert Half's research on AI and hiring demand reports that "3 in 10 employers eliminated positions after implementing AI but later added those roles back." Its explanation is blunt: "Employers underestimated how much human judgment, context and decision making those roles required."
The Canadian version of that survey, covering 1,365 hiring managers and reported by HR Reporter, shows why. Among managers who brought roles back, 38% said AI needed more human oversight and quality control than expected, 37% said the role involved relationship management AI couldn't replicate, 36% pointed to institutional knowledge AI couldn't replace, and 33% said the people left behind burned out.
Workers see it too. In an SHL survey run by Propeller Insights in June and July 2026, nearly 1 in 5 U.S. workers said they had been laid off or replaced by AI and later asked to come back, CNBC reported on October 30. SHL's Lucy Beaumont said companies "pulled the plug too quickly."
Then on November 4, Careerminds released "The AI Redeployment Gap," a survey of 1,142 U.S. HR professionals and people managers run with Censuswide. 92% of HR professionals and 88% of people managers believe employees in their organizations could be reskilled for AI-related roles. Yet 51% of HR professionals say their organizations are hiring from outside for new AI roles, while 42% are redeploying the people they already have.
Why it happened: the human side
I watch money for a living, so here's the plain version. A job description is a receipt for the tasks you can see. It isn't a list of everything the person does.
The invisible work wasn't counted. The calls a person makes before a problem becomes a complaint. Knowing which client needs a phone call, not an email. Catching the error nobody wrote a rule for. None of it is on the org chart, so none of it was in the AI business case.
Nobody owned the decision. In the Careerminds data, 63% of HR professionals say HR has primary responsibility for planning AI-driven workforce changes. Only 24% of people managers agree, and 35% point to senior leaders outside HR. When three groups each think someone else owns it, roles get cut on a spreadsheet and the people who know the work never get asked.
People were invisible, too. About 43% of HR professionals report limited visibility into which employees should be reskilled. You can't redeploy talent you can't see.
And the rehire isn't free. In the SHL data, almost half of U.S. workers asked back want a significant raise, and 43% want a written job-security guarantee. Beaumont's warning: "You've really broken the trust with those people."
What Mitch has taught on this
In his "Keynote demo: Adapt!" video, Mitch names the skill these companies skipped. "Adaptability is truly your ability to change directions, to manage hyper levels of adjustments," he says, from new information to new initiatives to the way the market changes. Then he asks the question every leader should ask before a cut: "Can I take what's around me and adapt and use it to actually make a bigger difference? That's adaptability."
The companies boomeranging now did the opposite. They treated AI as a replacement instead of a partner, and treated their people as a fixed cost instead of the ones best placed to adapt. Our read: the people already in the room, who know the clients and the shortcuts, can adapt fastest when you give them the chance. AI amplifies them. It doesn't stand in for them.
Best practices: five moves before any AI-driven cut
Run a role autopsy first. The manager sits with the person in the role for 30 minutes and lists every task, including the unwritten ones: relationships, exceptions, judgment calls. Anything AI can't do on that list stays human.
Shadow-test for two weeks. Let AI do the work next to the person, not instead of them. The person logs every catch and correction. In Robert Half's Canadian survey, 38% of managers who rehired said AI needed more oversight than expected. Measure it before you cut.
Post AI roles inside first. HR gives current employees first look at every new AI-related role for a set window before going outside. The Careerminds white paper, citing The Josh Bersin Company, says external hires can cost three to five times more and take up to 20 days longer.
Name one owner in writing. The executive team writes down who owns AI workforce decisions, HR or line leaders, and what each must sign off on. No owner, no cut.
Count the whole bill. Finance adds severance, rehiring, oversight hours, quality errors and the strain on the team left behind to every AI business case. If the savings only work without those lines, they aren't savings.
What's driving people crazy this week
Broken trust after "AI layoffs." One widely shared Reddit post this fall, covered here, described a worker replaced by AI whose boss later wanted him back after the AI's reports caused client trouble. It's one person's account, but the comments asked the same question over and over: why would anyone trust that company again? SHL's numbers say the same thing: people want guarantees, transparency and, in 30% of cases, an apology from leadership (CNBC). The human fix: tell people the truth early. Say what AI is being tested for, who decides, and what happens to their role, before the rumor does it for you.
What this means for the next 1 to 3 years
Here's our view. First, "redeploy before replace" becomes a board-level question, the way safety and security did. Boards will ask how many people were reskilled for every role cut. Second, every role gets a "context map": a living list of the relationships and judgment calls in the job, kept by the person who does it. It becomes the most valuable document in the company, because it's the one AI can't write. Third, a new job family grows fast: people who supervise AI output, the oversight work Robert Half's managers didn't budget for. The best candidates are already on your payroll.
How we evolve
The human skill to build is adaptability, Mitch's word, and it is a skill, not a mood. It matters more as AI gets better, because the tasks will keep changing and the people who can change direction fast become the ones every company wants to keep. One daily practice: at the end of each day, write one sentence: "Today something changed, and here's how I used it." Thirty days of that, and you'll have proof of your own adaptability that no job description shows.
Try this today (under 30 minutes)
Pick one role your team has talked about automating. Ask the person in it: "What do you do that isn't in your job description?" Write it all down.
Find out, in writing, who owns AI workforce decisions in your company. If nobody can tell you in one sentence, that's your first fix.
Look at your next open AI-related role and ask HR one question: "Who inside could do this with 60 days of training?"
Free tool: the before-you-cut checklist, ready to copy.
1. The unwritten work in this role is: [relationships, exceptions, judgment calls].
2. We tested AI next to the person for [weeks], and the person corrected it [number] times.
3. The person who owns this decision is: [name].
4. The people inside who could be reskilled for the new work are: [names].
5. The full cost includes severance, rehiring, oversight and team strain: [total].
6. What we've told the team, and when: [message, date].
FAQ
Are companies really rehiring people they cut for AI?
Yes. Robert Half research found 3 in 10 employers eliminated positions after implementing AI and later added those roles back, mostly because the roles needed more human judgment and context than expected.
Is it cheaper to reskill employees than to hire for AI roles?
Often, yes. The Careerminds white paper, citing The Josh Bersin Company, says external hiring can cost three to five times more than filling comparable roles internally and take up to 20 days longer.
What should a leader do before replacing a role with AI?
Map the unwritten work in the role, test AI next to the person before cutting, name one decision owner, and count the full cost, including oversight and rehiring.
Not sure where your own team stands? Take the free readiness assessment.
Skills gaps are people gaps. Book a 20-minute call about where your workforce is stuck.