Don't Be That Guy: The AI Layoff Playbook Is Already Backfiring

You know the type. His board keeps asking about the company's "AI strategy." Two CEOs at his last industry dinner mentioned, a little too casually, that they're running leaner now thanks to AI. His feed is wall-to-wall with peers announcing 30% gains and headcount they no longer need. The message he's absorbing isn't from any one of them—it's the ambient pressure that everyone else has figured this out, and if he doesn't move now he'll be the one left behind. So he does the thing that makes the pressure stop: a company-wide email about "doing more with less"—hiring freeze effective immediately, and everyone's expected to use the new tools or explain why not.

He isn't responding to evidence. He's responding to the fear of being the last one not doing it. And he's surprised when adoption stalls, the promised savings never show up in the numbers, and his best people start answering recruiter emails.

This is an unforced error. You're turning a capability into a threat, and you're calling your shot before you've taken it—on the strength of what your competitors say they're doing. Don't be that guy.

Two years ago, that was a warning. Now it's a documented pattern—enough companies have run the experiment that we can look at what actually happened.

The receipts

Start with Klarna. The fintech stopped hiring in 2023, cut its workforce by 22%, and replaced its customer service team with AI agents it proudly claimed could do the work of "700 full-time agents." It bragged about saving $10 million on marketing. Its CEO said in late 2024 that AI "can already do all of the jobs that we, as humans, do."

By mid-2025 the company was mounting a recruitment drive to hire people back. The CEO's explanation is the whole story in one sentence: "cost unfortunately seems to have been a too predominant evaluation factor when organizing this, what you end up having is lower quality." Klarna isn't a freak case. In one survey, 55% of employers who replaced workers with AI said they regret the decision.

Notice what went wrong. Klarna didn't lack the technology—it had OpenAI on speed dial. It optimized for cost, hit its cost target, and discovered too late that cost was standing in for things it hadn't measured: judgment, context, the customer's willingness to stay. That's the same lesson from a different angle than I make in The Efficiency Trap—chase the legible number and you can win it while losing the thing that mattered.

The cynical version is failing too

There's a less earnest way to be that guy, and it's more common: don't actually replace anyone, just say AI is the reason for cuts you were going to make anyway. Cassie Kozyrkov calls this AI-redundancy-washing—dressing up ordinary cost-cutting as visionary innovation. Telling shareholders you over-hired during the pandemic tanks the stock; announcing you're "pivoting to AI" sounds like the future. Fully 59% of hiring managers admitted to using AI as a cover story for layoffs.

Here's the part that should get every executive's attention: the market has caught on. A CNBC analysis found that of 23 S&P 500 companies that tied layoffs to AI, 13 were trading down after the announcement, with the decliners off about 25% on average. Investors have started reading "we replaced them with AI" not as strength but as an admission that you don't have a real story. The scapegoat stopped working.

The tell

The 2026 version of the mistake is subtler: mandating AI use and measuring it. Accenture now tells senior staff that "regular adoption" of its AI tools is required to move into leadership, and has begun tracking weekly logins as "a visible input to talent discussions."¹ Microsoft's developer chief told managers that using AI is "no longer optional" and belongs in performance reviews.² Shopify makes employees prove AI can't do a job before they're allowed to hire for it.³ Duolingo went furthest—an "AI-first" memo to stop hiring for anything AI could handle—then quietly walked it back in 2026 as it kept hiring people anyway.⁴ This looks like leadership. It's the opposite.

If a tool genuinely made people more productive, you wouldn't have to force them to touch it—they'd hide it from you to keep the advantage, which is a different problem I've written about in Leading AI Adoption. Mandating usage and tracking it means you're measuring inputs because the outputs aren't showing up. It's a confession dressed as a policy.

Why it backfires

The mechanism is not complicated. Announce that you intend to capture productivity gains before anyone has them, and you teach your most capable people that revealing a gain gets it taxed—so they stop revealing them. You pay the full cultural cost of extraction and collect none of the extraction, because individual gains don't automatically become organizational ones anyway; they leak out in the seams (see The New Productivity Paradox). Worst of both.

And the headcount you're cutting is worth more than it looks from the top. Up close, it's judgment, exception-handling, customer trust, and the expertise to notice when the system has quietly wandered off course. Complex technology needs more oversight, not less—usually from the very people being shown the door. Gartner projects that by 2027, half of enterprises without a people-centered AI plan will lose top talent. Good luck rebuilding trust with the people you just called redundant.

What good actually looks like

Imagine your operations team asks for a new hire right as you're rolling out AI. The bad answer: "We're not hiring—we bought AI, figure it out." The good answer: "We're rolling this out with real training and support. Let's see where your team is in a few months. If you still need someone, we'll talk." One creates fear; the other creates room to build capability while managing expectations honestly.

That's the whole posture. Invest in tools and training like you would for any serious technology. Say plainly that this is meant to enhance people's work, not thin their ranks—and then behave that way, so they believe you. Don't announce the savings before they exist. Let your normal performance systems reward the people who get good at this; you don't need a special framework.

None of this is anti-AI or anti-efficiency. If you've actually put a system into production—tested, monitored, humans in the loop—and it reliably does the work, then say so; that's honest. If you're deliberately moving spend toward AI infrastructure and redesigning workflows so your remaining people are more effective, that's a real strategy, and it's the one worth having. But that's a small minority. Most companies are somewhere between a sandbox, a pilot, and the revenge of the CFO—and reaching for the layoff announcement anyway.

Don't be that guy

The firms trying hardest to grab the AI payoff up front are the ones least likely to see it. They spook their best people, they degrade the work, and—now that the market has learned to see through it—they don't even get the stock bump for their trouble.

The harder you try to grab it, the more certainly you'll never see it. Provide good tools, real training, and honest messaging, then let your people surprise you. Just don't be that guy.

Footnotes

Accenture told senior staff that regular use of its AI tools would factor into promotions to leadership, and began tracking weekly logins. CNBC, Feb. 19, 2026: https://www.cnbc.com/2026/02/19/accenture-ai-orders-senior-staff-lose-out-promotions.html ↩
Microsoft's developer-division president said AI use is "no longer optional" and instructed managers to weigh it in performance reviews. Entrepreneur: https://www.entrepreneur.com/business-news/microsoft-staff-told-to-use-ai-more-at-work-report/493955 ↩
Shopify CEO Tobi Lütke's memo made effective AI use a baseline expectation and required teams to prove AI couldn't do a job before hiring for it. The Washington Post: https://www.washingtonpost.com/business/2025/06/03/ai-workplace-duolingo-shopify-employees/ ↩
Duolingo's 2025 "AI-first" memo (see The Washington Post, above) was walked back in 2026 as full-time hiring continued to grow. Metaintro: https://www.metaintro.com/blog/duolingo-ceo-walks-back-ai-first-memo-hiring-grows-2026 ↩