The Story Everyone Bought

For two years the playbook was simple. Announce an AI programme. Cut headcount. Book the savings. The market applauded, boards signed off, and every earnings call had the same slide: fewer people, more output.

That story is now unravelling in public.

What Actually Happened

The rehiring has started. IBM's AI handled roughly 94% of routine HR requests, and almost none of the judgment calls. It's now tripling entry-level hiring. Ford is bringing back experienced engineers to fix quality problems its automated systems couldn't. Commonwealth Bank of Australia cut customer service roles for a voice bot, then offered people their jobs back within months.

This isn't a handful of anecdotes. Forrester reports around 55% of employers now regret AI-related layoffs. A Robert Half survey found over 30% of hiring managers who cut roles for AI have since reinstated them.

The pattern is consistent. AI absorbed the routine share of each job and choked on the rest. The rest turned out to be where the value was.

Markets Have Noticed

The stock market has flipped on this faster than the operators have.

Goldman Sachs research found companies announcing layoffs now show higher debt, higher interest expense and lower profit growth than industry peers. Automation-linked layoffs now knock roughly 2% off a stock on announcement.

A headcount cut used to read as discipline. It now reads as distress. The omen has inverted, and most management teams haven't caught up.

Why It Keeps Failing

Not because the technology is useless. Because of the sequence.

Firms automated work they had never structured. No documented workflows. No SOPs. No named owners. Nobody could say precisely what the eliminated roles actually did all day. So nobody could specify what the AI needed to do.

You cannot automate a process nobody wrote down. AI amplifies structured operations. Pointed at chaos, it just accelerates the chaos.

What This Means for a Deal Team

In diligence: treat "AI-driven headcount reduction" in a management plan the way you'd treat customer concentration. A risk to test, not a synergy to bank. Ask one question: show me the documented process the AI will run. If it doesn't exist, the savings number is a guess.

In the portfolio: fix the sequence. Structure before automation. Map the workflow, document it, assign an owner. Then automate. It's cheaper, and it's the only version that holds.

In your own firm: don't cut operational capability on hype. The judgment, the chasing, the context: that's the 40% the machines couldn't do. It's also the part your deals run on.

The Question Worth Asking

Look at the last plan that crossed your desk with an AI efficiency line in it.

Was it built on a documented process, or on a story the public markets have already stopped buying?

The market has stopped buying this story. Don't be the one still paying for it.