Key takeaways
- AI changes the work before it changes the headcount.
- Redesign the work, prove it in production, then decide on roles.
- The cost of a rehire is a planning input, not a surprise.
- If nobody owns checking the AI and handling exceptions, the role isn't gone. It's just unstaffed.
Before you cut a role for AI, price the rehire.
It sounds backwards. The whole point of the cut is the saving. But a growing number of companies are discovering that the saving was a loan, and the repayment comes due when they quietly hire the role back.
The rehire is already happening
Careerminds surveyed 600 HR leaders whose organizations had made layoffs in the previous 12 months. Among the organizations with AI-led layoffs, 35.6% had already rehired more than half of the roles they cut. And 30.9%, nearly a third, found that rehiring cost more than the layoffs saved.
Gartner sees the same pattern coming at scale. In October 2026 it predicted that by 2029, 30% of employees terminated and replaced by AI will be rehired, often at a higher cost, unless organizations overhaul how they plan their workforce.
Those aren't arguments against AI. They're arguments against the order most companies are doing things in.
Why the cut comes back
A job is a bundle. Over years, a role collects the core task everyone can see plus a long tail of work nobody wrote down: checking, correcting, catching the odd case, knowing who to call when something breaks.
When a leader cuts the role because AI can do the visible task, the invisible work doesn't disappear. It lands on someone else, piles up, or simply stops getting done until a customer notices. Then the role comes back, usually hired in a hurry, from a tighter market, at a higher salary, and without the context the original person carried.
That's the rehire math. It isn't a surprise cost. It's a predictable one that wasn't in the business case.
Three checks before any cut
Before a headcount decision tied to AI, I'd ask three questions.
- Who checks the AI's work? AI output needs review, especially early on. If the answer is "the person we're cutting," the plan has a hole in it.
- Who handles the cases it can't? Every process has exceptions. Someone has to own them, with the authority and knowledge to resolve them.
- What does the rehire cost if you're wrong? Put a real number on it: recruiting, ramp-up time, the premium you'll pay, and the customer impact while the seat is empty. Then compare that against the saving.
If you can't answer all three with names and numbers, you're not ready to cut. You're ready to redesign.
Sequencing, not sentiment
My Cyborg CEO take is simple. AI changes the work first. So redesign the work, prove it in production, then decide the headcount.
That order matters. Redesigning the work shows you what the role actually contained. Proving it in production shows you whether the AI holds up on real volume, real exceptions and real customers, not a demo. Only then do you know what the team should look like.
Sometimes that leads to fewer people. Often it leads to the same people doing different, higher-value work. Either way, it's a decision made on evidence instead of a headline.
This isn't about protecting every job, and it isn't about slowing AI down. It's about not paying twice for the same work.
What leaders should take from this
- AI changes the work before it changes the headcount.
- Redesign the work, prove it in production, then decide on roles.
- The cost of a rehire is a planning input, not a surprise.
- If nobody owns checking the AI and handling exceptions, the role isn't gone. It's just unstaffed.
Watch the short video on YouTube for the quick version: https://www.youtube.com/watch?v=j3LrjbWdfH0

