· AI
Most AI Restructuring Isn't a Layoff. The Ones That Became One Share One Mistake
The restructuring stories that go wrong almost never fail because AI couldn't do the job. They fail because the company skipped a process the law already requires.
Ask most AI-using companies whether AI cost anyone their job and the honest answer, right now, is usually no. A survey of businesses in New York and northern New Jersey released by the Federal Reserve Bank of New York found that only 4 percent of AI-using service firms reported laying off workers because of the technology in the prior six months, and none of the manufacturers surveyed did. Just over a third of those service firms, and more than a fifth of manufacturers, reported retraining workers instead. The Fed's own framing: "Existing workers are much more likely to be retrained than replaced by AI." Treat that as a regional snapshot, not a national verdict: the Fed did not publish a sample size, and its researchers warned the pattern could shift as adoption matures.
My position: the AI restructurings that do blow up into costly layoffs almost never blow up because the technology failed. They blow up because the company treated a restructuring as an announcement instead of a process, and ordinary employment law, not any AI-specific statute, is what catches that. You do not need a new rule written for AI to get this right. You need to run the one you are probably already bound by.
The case for cutting first anyway
That case deserves a fair hearing. Accenture laid off more than 11,000 people between May and August 2025, and its CEO, Julie Sweet, was blunt about why retraining was not the plan: "reskilling, based on our experience, is not a viable path for the skills we need," she said, while telling investors the firm expected to grow headcount overall in fiscal 2026. That is not a company dodging the harder work of managing people. It is a company concluding the gap between today's job and tomorrow's need is too wide to train across, and acting on it instead of paying to pretend otherwise. Sometimes retrain-first is the wrong call for a given role, and saying so beats dressing a necessary cut in redeployment language nobody believes.
Where the actual failures come from
The clearest cautionary case is not a famous AI failure. It is Landcorp Farming, a New Zealand government-owned farming company trading as Pamu, ordered to pay an accounts payable team leader roughly $27,000 after introducing AI-based automation and proposing to eliminate her role. The Employment Relations Authority did not rule against automating the job. It ruled against how the company did it: it emailed her the decision while she was on sick leave, answered her feedback only after the decision was made, and told her to apply for roles her employment agreement required it to offer directly. As the Authority put it, consultation "involves the statement of a proposal not yet finally decided on." The AI was almost incidental. A manager who ran that same shortcut over any ordinary redundancy would have lost the same case.
The viral counterexample teaches the same lesson in reverse. The story that Ingka Group, one of IKEA's franchisees, retrained roughly 8,500 call center workers into remote interior design advisors after a chatbot called Billie absorbed routine questions is true in its basics, according to an analysis from The State of AI that traced the claim through Reuters' and Fortune's reporting. The viral framing inflates it, though: the pandemic drove the sales spike that created those design-advisor roles, Billie handled a share of contact volume rather than full jobs, and Ingka has cut hundreds of corporate roles since, just not in that unit. The real story is narrower than the headline: a two-year effort aimed at a specific, genuinely bridgeable skills gap, that happened to work. It is the Accenture judgment call made correctly in reverse, not proof that AI restructuring is painless.
What this means before you restructure anything
Put those three cases together and the pattern is not "AI replaces jobs" or "AI never replaces jobs." It is that the skills-gap question is answerable in advance, and the answer determines which path is honest. Accenture answered it one way for roles it judged unbridgeable. Ingka answered it the other way for a role it judged trainable on a defined timeline. Landcorp never answered it. It skipped straight to the outcome and improvised the process, the one move that turns a defensible business decision into a legal and reputational cost, with or without a machine involved.
That is also the case against reaching for a new AI-specific layoff law every time a restructuring goes badly. Landcorp did not need one. General consultation and redundancy law already required a real process, a timely response to feedback, and redeployment terms that matched the contract, and it punished the company for skipping all three. The fix for bad AI restructuring is the fix for any bad restructuring: answer the skills-gap question honestly, then run the process you are already bound to run.
This is the organizational-design work behind MojoAgents: figuring out what a fleet of agents can actually take over, versus what still needs a person, before anyone writes a headcount number down. Not sure which side of that line a role falls on? That is worth working out before you send anyone an email.
Sources
References used in this article. Links also appear alongside the relevant claims.
Let's make it happen.
You don't need a polished brief. A couple of lines about where you want your company to go is plenty, and we'll come back with what we'd tackle first.
We scope the work around your goals and timeline before quoting anything, so you know exactly what you're getting.