Must AI Efficiency Lead to Layoffs? Figure Out the People and Money First

When AI lands well, the replacement rate in roles with mature workflows runs about 60%. Customer service that used to need a hundred people can get by with just over thirty after AI. The sixty-plus people freed up mostly get laid off. Given the choice, I’d rather work on revenue growth — but if a company can’t survive, costs have to come down.

Many leaders assume AI adoption stalls because of technology or incomplete data. They buy tools, run training programs, and still can’t get traction. What’s really stuck is that employees won’t cooperate. How a sales rep talks to clients and closes deals is exactly what makes them valuable at the company. Hand that over, and they worry they’ll stop mattering.

In April 2026, a US enterprise AI software company and a workplace research firm surveyed 2,400 workers across the U.S., the U.K., and Europe, and about 30% admitted to actively undermining their company’s AI initiatives — among those in their twenties, over 40%. In August 2025, a UK tech consultancy surveyed 4,000 people across four countries: one in three was hiding what they knew, with nearly 40% refusing to teach coworkers their most valued skills.

Their fear is justified. In a study published in 2023 and in a top economics journal in 2025, economists from Stanford and MIT looked at over 5,000 customer service agents and found that after an AI assistant was introduced, novices improved by over 30%, while veterans and top performers barely improved at all. AI had learned the experts’ methods and was teaching them to newcomers. Under the company’s existing bonus structure, top performers’ income could actually decline. They contributed their expertise and watched others’ numbers rise while theirs stayed flat — sharing meant losing.

Some leaders think they can hire cheap newcomers plus AI and replace expensive veterans. This is extremely shortsighted. Veterans’ understanding of customers and the business is exactly what you need to explore new directions. Replace them all with newcomers and AI, and you can only make the existing business cheaper — grinding away in an ever-narrower lane. Leaders need to get the money right first: when AI takes what experts know and teaches it to newcomers, the newcomers’ improved performance should include a share for the experts.

Layoffs shouldn’t be rushed. In February 2026, a US outplacement firm surveyed 600 HR leaders; about a third said critical skills and institutional knowledge walked out the door after layoffs. More than half admitted they never considered whether those people could have been reassigned. Moving freed-up people into revenue-growth roles is worth more than cutting them outright.

A Wharton researcher noted in January 2026 that part of the time AI frees up should go back to employees — for learning new things and trying new initiatives. That’s how they come to see AI as helping them, not replacing them.

From the employee’s side, fighting it gets you nowhere. The wheels of history roll over it sooner or later — a carriage driver is better off learning to drive a car early, and taking what he knows about the roads with him. Your knowledge of clients and the business is still valuable in the AI era. BCG surveyed 1,000 executives globally in October 2024 and found that among companies achieving real results with AI, algorithms accounted for just 10%, technology and data for 20%, and the remaining 70% was spent on people and processes. In enterprise AI, what matters isn’t how advanced your technology is — it’s whether you can figure out the people and the money.

EP0102 · 2026-09-20 · 03:45. The picture uses a digital human, and the voice uses my cloned voice. This is an English translation of the approved Chinese script used to create the narration.

When AI actually lands well, from a cost-cutting perspective, it means cutting people. In roles where the workflow is already mature, the replacement rate is about 60%. Customer service that used to need a hundred people — after AI, just over thirty can handle it. And those sixty-plus people freed up, is there really new work for them? Not necessarily. Most get laid off.

That’s why I don’t love taking on cost-cutting projects. Given the choice, I’d rather work on revenue growth — expanding the business, exploring new directions. Because only when the pie gets bigger do the people contributing have something to gain, and a reason to contribute, right? But if a company can’t survive, then costs have to come down. There’s no way around it.

A lot of leaders think AI adoption stalls because of technology or incomplete data. They buy tools, run training sessions, and still can’t get it moving. Think about it — what’s really stuck? Employees won’t cooperate. When outside help comes in to build AI for the boss and asks a sales rep for their chat logs, they often wait a long time and get nothing. That rep knows you’re working for the boss. How they talk to clients, how they close deals — that’s exactly what makes them valuable at this company. Hand it over, and they worry they won’t matter anymore, right?

This isn’t an isolated thing. In April 2026, a U.S. enterprise AI software company partnered with a workplace research firm and surveyed 2,400 workers across the U.S., the U.K., and Europe. About 30% admitted they were actively undermining their company’s AI initiatives. Among those in their twenties, it was over 40%. In August 2025, a British tech consultancy surveyed 4,000 people across four countries. One in three was hiding what they knew, and nearly 40% refused to teach coworkers their most valued skills.

And their fear is justified. Economists from Stanford and MIT published a study in 2023 that appeared in a top economics journal in 2025. They studied over 5,000 customer service agents at one company. After an AI assistant was introduced, novices and weaker performers improved by over 30%. Veterans and top performers barely improved at all. The reason: AI had learned the experts’ methods and was teaching them to newcomers. Worse still, under the company’s existing bonus structure, top performers’ income could actually decline. They contributed their expertise, only to watch others’ numbers go up while theirs stayed flat. That’s not selfishness — they genuinely lose out by sharing.

So some leaders think: since AI can bring new hires up to speed quickly, why not hire cheap newcomers plus AI and replace the expensive veterans? This is extremely shortsighted — it kills the company’s potential for innovation. What veterans carry — their understanding of customers and the business — is exactly what you need to explore new directions. Replace everyone with newcomers plus AI, and all you can do is make your existing mature business cheaper, grinding away in an ever-narrower lane.

So leaders need to figure out the money first. When AI takes what the experts know and teaches it to newcomers, the newcomers’ improved performance should include a share for the experts. Otherwise, under the old formula, sharing means losing — and nobody is going to do that.

Layoffs shouldn’t be rushed either. In February 2026, a U.S. outplacement services company surveyed 600 HR leaders. About a third said that after layoffs, critical skills and institutional knowledge walked out the door. More than half admitted they never even considered whether those people could have been reassigned before cutting them. Could the people being freed up be moved to revenue-growth roles instead? That’s worth more than cutting them outright.

One more thing — don’t pocket everything AI saves. A Wharton researcher said in January 2026: take some of that freed-up time and give it to employees to learn new things, to try new initiatives. That way they feel AI is helping them, not replacing them, right?

From the employee’s perspective, resisting this is futile — absolutely futile. The wheel of history will roll over eventually. A horse-carriage driver is better off learning to drive sooner rather than later. Guarding your carriage-driving skills and refusing to teach anyone doesn’t help. Better to take your knowledge of the roads — your familiarity with clients and the business — and apply it to driving a car. Those things are still valuable in the AI era, as long as you’re willing to pick up the new tools.

BCG surveyed 1,000 executives globally in October 2024. Among companies that achieved real results with AI, algorithms accounted for just 10%, technology and data for 20%, and the remaining 70% was spent on people and processes. I’ve said before that enterprise AI transformation is an organizational problem, not a technology problem — today we dug one layer deeper. Think about it: in enterprise AI, what matters isn’t how advanced your model is. It’s whether you can figure out the people and the money. That’s all for today. See you tomorrow. Bye.