Adopt AI or get cut. But adopt the right thing.

The numbers are blunt enough that I'll just lay them out. A survey of 2,400 workers, including 1,200 executives, found that 60% of companies plan to lay off employees who can't or won't use AI. 77% say non-AI-proficient people won't be considered for promotions or leadership. AI "super-users" save about 9 hours a week and were 3× more likely to have gotten a promotion and a raise in the past year. "Adopt AI or get cut" has stopped being a spicy LinkedIn take and become a stated personnel policy.

So yes — the mandate is real, and pretending otherwise is bad advice. But here's the part almost everyone gets wrong, and where the same survey quietly disagrees with how it's being read: "adopt AI" does not mean "use the tools a lot." That's the shallow version, and it's a trap. The skill that actually gets rewarded — and that's hard to fire — is something else.

"Adopt" is being misread as "use," and that's the trap

The obvious reading of the mandate is: learn the tools, type into ChatGPT, be fast. And sure, do that — it's table stakes now, everyone will. But if your entire value is "I'm quick with the AI," you've made yourself easy to replace, because being quick with the AI is exactly the thing the AI is making universal. A capability everyone has is not an advantage.

The survey itself points past fluency. The well-paid version of AI skill, researchers note, is pairing the tools with judgment, critical thinking, and knowing when not to use AI — and employers have started interviewing for that specifically, probing whether candidates know when an AI tool is the wrong choice, how they verify its output, and what they do when it hands them a wrong answer. The premium isn't for enthusiasm. It's for skepticism that works.

The scarce skill is judgment about the output

Here's why that's the durable bet. As AI generates more and more of the work — the code, the copy, the analysis — the bottleneck and the value move from producing to judging. When everyone can generate a plausible answer in seconds, the rare, promotable person is the one who can look at that answer and know, quickly, whether it's right. That's the thread running through everything I write: the demo was never the hard part, the review is the real achievement, and an AI that sounds confident is most dangerous where it's wrong.

So the right way to "adopt AI" for your career isn't to use it more uncritically. It's to become the person on the team who can tell when it's lying. Use it constantly — and read every output like a skeptic, catch its mistakes, know the cases where you shouldn't trust it, and keep the underlying expertise that lets you judge it at all. That combination — fluent and critical — is the one earning the 56% premium, not fluency alone.

A warning, especially if you're early

There's a sharp edge here for people earlier in their careers, and the research names it: workers who learned to lean on AI before developing their own professional judgment are less able to catch its errors or work independently when it fails. That's the worst quadrant — fluent and defenseless. If AI does all your thinking from day one, you never build the judgment that makes you valuable and that lets you supervise the AI safely.

So the move isn't to avoid AI to "stay sharp" — that ship sailed, and refusing it is the other way to get cut. It's to use AI and deliberately keep building the expertise underneath: do the hard reasoning yourself sometimes, understand why an answer is right, learn your domain deeply enough to smell when the output is off. Use the tool to go faster, not to skip the part that makes you the person who can tell it's wrong.

The bottom line

"Adopt AI or get cut" is true, and you should take it seriously — refusing AI is a real way to become unemployable. But the people who get cut won't only be the refuseniks. They'll also be the uncritical adopters: fast with the tools, helpless when the tools are confidently wrong, indistinguishable from everyone else who's equally fast. Fluency is the floor now, not the differentiator.

So adopt — aggressively — but adopt the right thing. The durable, well-paid skill in an AI workplace isn't being the fastest user of the model. It's being the best judge of its output: the one who knows when to trust it, when to override it, and how to catch the expensive mistake before it ships. Become that person. The tool is easy to learn and easy to commoditize. The judgment is the part they can't automate, and it's the part that keeps you.