55% of Employers Regret Their AI Layoffs. Here Is the Job They Should Have Created Instead.

SuperStaff was founded in 2009 and is now in its 18th year of business. I joined SuperStaff as CEO in 2019, and from that seat I have watched every flavor of workforce panic: offshoring waves, recession cuts, the great resignation, return-to-office standoffs. I have never watched anything quite like 2026, when American employers announced six figures' worth of job cuts while saying the word “AI” out loud, and then, before the severance checks cleared, started quietly hiring people back. I run a company that trains and supervises AI workforces, so you might expect me to cheer the layoffs as validation. I am here to tell you the opposite. The layoffs were done backwards, the data now proves it, and the companies that got it right never had to choose between the agent and the human in the first place.

How big is the AI layoff wave, really?

Direct answer: AI went from a footnote to the leading stated reason for US job cuts in under a year. Challenger, Gray & Christmas, the firm that has tracked announced layoffs for decades, attributed 54,836 US job cuts to AI in all of 2025, roughly 4.5% of the year's 1,206,374 announced cuts. Then 2026 happened. In April alone, employers attributed 21,490 planned cuts to AI, 26% of the month's total, making AI the top cited reason for the second straight month. AI led the monthly reasons five months in a row, March through July, and by the end of July it had been cited in 112,713 cuts for the year, more than double all of 2025. The streak broke in August, but through September AI had been cited in 120,136 announced cuts and remained the leading reason for the year.

The names got big too, starting in late 2025. Salesforce's CEO said AI agents let him cut 4,000 customer support roles. Amazon announced 14,000 corporate cuts while citing AI. And plenty of smaller companies you have never heard of followed the leaders, because nothing emboldens a mid-sized business like watching a giant do the scary thing first.

Two honest caveats before the sermon. First, Challenger counts what employers say, and “AI” has become a fashionable thing to say when the real reason is softer demand. Gartner's own analyst, Kathy Ross, put it this way: “Most recent workforce reductions were influenced by broader economic conditions rather than automation alone.” Second, total announced US job cuts in 2026 were actually down 39% year over year through September. The story is not that the sky is falling. The story is what happened next to the companies that swung the axe in AI's name.

What is the AI boomerang, and how fast did the regret arrive?

Direct answer: within months, by every measure anyone has published. The research firms do not agree on much, but they agree on this:

• Forrester, in its Predictions 2026 report, put the share of employers who regret their AI layoffs at 55%, and expects roughly half of the layoffs blamed on AI to be undone, with much of the work going to lower-wage or offshore workers. Orgvue, which sells workforce planning software, reported a regret figure that matched almost exactly: among surveyed leaders whose companies let people go over AI, 55% said they had made the wrong decisions.
• Gartner, in research from its customer service practice, predicts that one in two companies that pinned headcount reductions on AI will be rehiring for comparable work under new job titles by 2027. When Gartner polled 321 service and support leaders in October 2025, just one in five said AI had actually led them to cut agents, and its researcher Emily Potosky put the reason plainly: “AI simply isn't mature enough to fully replace the expertise, empathy, and judgment that human agents provide.”
• Careerminds surveyed 600 HR leaders in February 2026 and found that about two thirds of the companies behind AI layoffs had already started rehiring, more than a third had brought back over half of the jobs they eliminated, and 52% said the rehiring happened within six months.
• Robert Half found that roughly one in three US hiring managers who scrapped a position over AI later filled it again, with finance the most boomerang-prone department at 44%, followed by HR and tech.

Six months. That is not a strategic workforce transformation. That is a round trip. Somebody was always going to name this, and “AI boomerang” stuck, with the returning staff cast as boomerang employees, which is a cute name for “the people who knew how to do the job all along.”

What did firing people for AI actually cost?

Direct answer: for three out of four companies, nothing good. In the Careerminds data, 30.9% of organizations paid more to restaff than the automation ever saved, 42.4% roughly broke even, and only about a quarter came out ahead. Orgvue's accounting, which covers layoffs in general and not only the AI kind, is even uglier: once you add up severance, lost output and the cost of hiring replacements, the average layoff costs about USD 1.27 for each USD 1.00 it saves. Imagine pitching that to your board as a project: “We will pay a dollar twenty-seven to save a dollar, destroy morale, and make our customers talk to a bot that is not ready.” Approved unanimously, apparently, at over a hundred thousand desks this year.

The case studies put faces on the math:

• In July 2025, Commonwealth Bank of Australia told 45 customer service staff their jobs would go to an AI voice bot. According to the Finance Sector Union, call volumes rose and team leaders ended up answering phones, and by August 21 the bank had backed down, admitted the roles were wrongly declared redundant, and apologized. This from a bank that booked a record profit the same year. The union called it a win. I call it a free masterclass, paid for by someone else. (It made our list of 12 AI launches that backfired in public for the same reason.)
• Klarna's 2024 boast was that its AI assistant could handle what 700 agents did. Then its CEO admitted the cost-first push produced lower quality and started hiring humans back, saying customers need to know “there will be always a human if you want.”
• IBM is the quiet counterexample. In a Wall Street Journal interview, CEO Arvind Krishna said IBM's AskHR assistant now takes care of roughly 94% of the routine HR workload, that it displaced only “a couple hundred” people, and that IBM's overall headcount grew. Automating tasks while growing the workforce: it turns out that is allowed.

And Forrester predicts that much of the returning work will go to lower-wage workers, offshore or at lower salary. Let me say the quiet part at full volume: when a company fires you for a robot, discovers the robot cannot do your job, and then offers you the job back for less money, that is not artificial intelligence. That is an old-fashioned wage squeeze wearing a very expensive costume. I run an outsourcing firm in the Philippines; nobody on this earth is better positioned to profit from that move than I am, and I am telling you it is a short-sighted way to treat the people who hold your institutional knowledge.

Why did the layoff-first math fail?

Direct answer: because companies cut the exact people their AI needed, and skipped the supervision layer that makes AI output usable. You do not have to take my word for it; the hiring managers confessed in a survey. Among Robert Half's hiring managers who brought AI-cut roles back, 40% said the roles needed institutional knowledge AI could not replace, and 38% said the work needed more human quality control than they had expected. Smaller-than-expected productivity gains made the list too.

Read those findings against everything we know about why AI projects fail, which I graded source by source in my fact-check of the famous failure statistics, and against why so many agentic AI projects get canceled. They are the same list. The agent was never trained on the company's actual workflows, because the person who knew the workflows was packing a box. Nobody owned quality control, because the budget assumed the whole point was removing the humans who do quality control. In my graduate AI studies the literature calls this a socio-technical integration failure. On the street it is simpler: you fired the teacher, then wondered why the student failed.

There is a second error underneath, and it is the one executives keep making with a straight face: confusing tasks with jobs. IBM handed most of its routine HR work to software and, by its CEO's account, cut only a couple hundred roles, because a job is a bundle, and the AI takes the chores inside it, not the person who answers for the result. The companies that fired whole humans because software absorbed a slice of their task list are the ones now rehiring “staff to perform similar functions, but under different job titles,” in Gartner's gorgeous phrasing. They did not eliminate the role. They eliminated the training data, lost the judgment, kept the liability, and bought the role back at a markup with a new name on the door.

Here is the one-liner to tape to the conference room wall: they did not automate the job, they automated the excuse.

What job should they have created instead?

Direct answer: the AI supervisor, which is what those “different job titles” turn out to be once the boomerang lands. The entire cycle of fire, fail, regret, rehire exists because companies ran the sequence backwards. The correct sequence creates a role before it removes one. Across SuperStaff's history of building technical teams, now in its 18th year of business, the pattern for every new technology has been the same: the winners move their best operator up a level, to run the new tool, rather than out the door.

What the role looks like, concretely:

• Owns the workflow map. Before any agent is deployed, someone who has done the job documents how the work actually flows, including the exceptions that never made it into the manual. That someone is usually the person companies laid off first.
• Trains the agent. Writes the agent's job description, feeds it the real examples and edge cases, and sets the escalation rules. A bot trained by the internet answers like the internet. A bot trained by your best bookkeeper answers like your best bookkeeper.
• Supervises to risk. Reviews outputs daily where mistakes are expensive, samples where they are cheap, and owns the exceptions. This is the quality-control line item that 38% of Robert Half's rehiring managers said they underestimated.
• Measures against a human baseline. The agent earns autonomy the way a new hire earns it, by hitting the standard under supervision, not by being cheaper on a spreadsheet.

Price it out and the economics embarrass the layoff plan. One trained supervisor overseeing agents across several workflows can cost a fraction of the staff those workflows used to need, without the severance, the rehiring fees, the morale crater or the USD 1.27-per-dollar surprise. And the growth data points the same way: the US Chamber found 47% of small businesses that use AI say it is already creating jobs for them, and AI users were at least 7 points likelier than non-users to say their sales, profits and headcount grew. The companies winning with AI are not the ones with the bravest layoffs. They are the ones that changed what their people do.

What should you do before your own AI staffing decision?

Direct answer: automate tasks, redeploy people, and do not cut anyone until a supervised agent beats their baseline. The owner's checklist, free of charge:

1. List tasks, not jobs. Score each task for volume, rules, documentation and cost of error. Automate the scorable ones. The judgment and the exceptions stay human.
2. Promote before you terminate. Your most experienced person in the workflow becomes the agent's trainer and supervisor. If the agent truly absorbs the work of three people, you need fewer doers and one better-paid checker, and attrition usually gets you there without a single layoff headline.
3. Demand the baseline. No agent replaces a human until it performs within a few points of that human's accuracy on your own data, under supervision, for a sustained period. If a vendor will not measure this, that is your answer.
4. Budget the supervision. If the business case only works with zero human review, you do not have a business case. You have a liability with a go-live date.
5. Keep the receipts. Measure what the automation actually saves against what the transition actually costs. Nearly three quarters of the companies in the Careerminds data broke even or lost money on the round trip.

FAQ: the AI layoff boomerang, answered straight

Do companies regret AI layoffs?

Yes, at remarkable rates. Forrester's Predictions 2026 research puts the share of employers regretting their AI layoffs at 55%, and Orgvue reported a nearly identical figure.

Are companies rehiring workers they laid off for AI?

Yes. Gartner expects one in two companies that pinned cuts on AI to be hiring for comparable work under new job titles by 2027. A Careerminds survey of 600 HR leaders found about two thirds of the companies behind AI layoffs had already started rehiring, and 52% said they rehired within six months.

Did AI layoffs actually save money?

Usually not. In the Careerminds data, about 31% of companies paid more to restaff than automation saved and 42% broke even. Separately, Orgvue estimates that layoffs in general cost about USD 1.27 for each dollar they save, once you add severance, lost output and the cost of hiring replacements.

Why do AI layoffs backfire?

Hiring managers who brought AI-cut roles back gave two leading reasons in Robert Half's research: AI could not replace institutional knowledge (40%), and the work needed more human quality control than expected (38%). The people cut first were usually the ones whose knowledge the AI needed to be trained, and whose review it needed to be trusted.

How many US job cuts are attributed to AI?

Challenger, Gray & Christmas attributed 54,836 US job cuts to AI in 2025. In 2026 AI was the leading stated reason five months running, March through July, and through September it had been cited in 120,136 announced cuts, more than double the full 2025 total. These counts reflect what employers state, which can be influenced by fashion as much as by automation.

The part where I tell you what we do

The collection plate, briefly. At Agentic Trainers, powered by SuperStaff, we train AI agents on your actual workflows and supply the trained humans who supervise them, which is precisely the role this entire article argues you should create before anyone gets a severance packet. If you are weighing an AI staffing decision, a free 30-minute workflow audit call will show you which tasks are automatable and which ones will boomerang. Do the math before the layoff. The companies in the Careerminds data did it after, and nearly three in four of them did not come out ahead.

Disclaimer: This article is commentary and opinion based on public sources available as of November 3, 2026. It is not legal, HR or financial advice. Forrester's and Gartner's rehiring figures are predictions; the Careerminds, Orgvue and Robert Half figures are self-reported surveys; Challenger's counts reflect the reasons employers state, not verified causes. Agentic Trainers sells AI agent training and human supervision services. If you believe anything here is inaccurate, contact us at hello@agentictrainers.com and we will review it promptly.

Written by Matt Narciso with AI-assisted research and drafting. Every fact and source was reviewed by a human editor.

Sources

1. AI layoffs to backfire: Half quietly rehired at lower pay, The Register, on Forrester's Predictions 2026 (October 29, 2025)
2. Gartner Predicts Half of Companies That Cut Customer Service Staff Due to AI Will Rehire by 2027, Gartner (February 2, 2026)
3. AI-led layoffs: what HR leaders wish they knew before making job cuts, Careerminds (survey fielded February 2026)
4. 55% of businesses admit wrong decisions in making employees redundant when bringing AI into the workforce, Orgvue (April 29, 2025)
5. The fire-to-hire cycle, Orgvue (October 2025, PDF)
6. May 2026 Labor Market Update: What employers and job seekers should know, Robert Half (June 16, 2026)
7. The AI boomerang effect: more data suggests employers are reversing AI layoffs, TechSpot (June 9, 2026)
8. 55% of Companies That Slashed Jobs for AI Now Regret It After Restaffing Costs Soar, International Business Times UK (September 10, 2026)
9. 2025 Year-End Challenger Report, Challenger, Gray & Christmas (January 8, 2026)
10. Challenger Report: April Job Cuts Rise 38% from March; YTD Cuts Down 50%, Challenger, Gray & Christmas (May 7, 2026)
11. Challenger Report: Layoffs Fall, Hiring Picks Up; AI Leads For Fifth Straight Month, Challenger, Gray & Christmas (August 6, 2026)
12. Job Cuts Fall in September; Hiring Plans Up 3% Over 2025, Challenger, Gray & Christmas (October 1, 2026)
13. CBA reverses AI-driven job cuts, admits 'error', Information Age (ACS) (August 21, 2025)
14. Klarna dials back its AI customer service strategy: now it's hiring humans again, Maginative (May 8, 2025)
15. Empowering Small Business: The Impact of Technology on U.S. Small Business, U.S. Chamber of Commerce (October 2, 2026)