Your Team Isn't Split by AI. It's Split by You.

When I took over the team, three different people warned me about the same engineer before I'd even met him.

He was brilliant, they said, and impossible. And he had drawn a hard line on AI: he would not use it. Not "hasn't gotten around to it" — would not. As the rest of the team started leaning on these tools and pulling ahead, his refusal was starting to look like a problem someone was going to have to manage out.

I didn't manage it out. I never told him he had to use anything. I just did my own work the way I do it, which is to say with AI woven through most of it, out in the open. A few weeks in, he started asking questions. A few weeks after that, he was the most thoughtful AI user on the team — not a vibe coder shipping slop, but someone using it to sharpen work that was already sharp. Today the rest of the team goes to him to learn how to do it well.

I've thought about that engineer a lot since I read this year's big tech-worker survey.

The finding everyone is going to quote

The survey asked 5,332 working tech professionals a deceptively simple question: how has working with AI changed how you see yourself as a professional? Just under half — 49% — said "amplified": I can do more, and better. About one in five landed somewhere darker, between "destabilized" (I'm less sure where I stand) and "diminished" (I feel less essential). The rest sat in the middle, watching their role change shape without a clear verdict on it.

Here's the part getting all the attention. Which bucket you fell into predicted how you felt about your entire career better than anything else they measured — better than your role, your seniority, your company's size, or your pay. In a regression pitting every variable against each other, that AI-identity stance was the single strongest predictor of both career optimism and whether you'd recommend your field to someone starting out, stronger than role, level, and company size combined. The gap between the "amplified" and the "diminished" came out around three times the size of the well-worn finding that founders are happier than everyone else. It was, by a wide margin, the biggest effect in the data.

The tidy conclusion is that AI has cleaved the workforce in two, and where you land is some fixed fact about you — your relationship with the technology, your adaptability, your wiring. That's the version that will get quoted in a hundred LinkedIn posts this month.

I think it's mostly wrong about the cause.

The arrow points the other way

Look again at what that question actually measures. Not your skill with the tools. Not your output. It measures a feeling about yourself — how AI has changed your sense of who you are at work. And feelings like that don't form in a vacuum. They form in an environment, under a manager, inside a particular stretch of weeks.

I've watched some version of this split for twenty years, and it long predates AI. I saw it when we moved to Microsoft SQL Server. I saw it with new versions of Windows Server, with configuration-management tooling, with every new language that turned up promising to change everything. There was always a camp that got energized and a camp that got defensive and dug in, and the ratio had almost nothing to do with the tool. The technology changed every eighteen months. The pattern never did.

The research here is fairly settled, even if the AI conversation talks right past it. A 2024 review of 63 studies on why employees resist digital transformation found that resistance comes mostly from context — perceived threats to job security, identity, status, and autonomy, and the way the change is managed — not from the tools themselves, and not from some workers being "resistant" by nature. Everett Rogers made a version of this point decades ago in Diffusion of Innovations: adoption runs on the social system and the people modeling the behavior, not on a technology's feature list. And Amy Edmondson's foundational work on psychological safety shows that whether people take the risk of trying something new is driven by the climate their leaders create, not by their personality.

Put plainly: the thing the survey captured as an identity is, to a real degree, a readout of the room. When five thousand people tell you how they feel about "tech," a large part of what you're actually hearing is how they feel about their manager and their Tuesday.

I want to be careful here — this is a hypothesis from twenty years of watching it, not a coefficient the survey measured. The survey found a correlation, not a cause. But the direction fits everything we know about how people actually take up new tools, and it fits every team I've walked into.

What actually flipped him

Which is why the engineer flipped, and it wasn't magic. Two things changed. First, the coercion came off — nobody stood over him with "use it or you're out," which is the fastest known method for turning a skeptic into a resenter. Second, someone he respected used the thing well, in the open, without turning it into a loyalty test. Safety, plus a credible model. That's close to the whole recipe, and it's the exact opposite of how most AI rollouts are running right now.

Because look at what the survey's angriest quotes have in common. The people who described tech as "kind of suck[ing] right now" weren't furious at the technology. They were furious at "use AI or you will lose your job — and then people get fired anyway." That is not a technology problem. That's a leadership problem wearing a technology costume, and no amount of tooling will fix it.

What this means if you run a team

If your team is separating into the thrilled and the shaken, the survey invites you to read that as a fact about your people. Read it instead as a verdict on the conditions you built.

The uncomfortable version: AI adoption is a sorting mechanism, and you're running one whether you meant to or not. Push it as a mandate, bolt it to fear, skip the modeling, and you will manufacture your own resistant camp — and then you'll be sorely tempted to conclude that those people were simply resistant, which conveniently absolves you of having built the thing that resisted them. Take the coercion off, put genuine safety under experimentation, model good use yourself, and you'll be surprised how many of your "refusers" were just waiting for the room to feel safe enough to try.

That is the actual job, I'd argue. Not "drive adoption." Grow people through a change that is genuinely disorienting, rather than standing back and letting it sort them into keepers and casualties. It's most of what I'm writing about in the book I'm working on now, What a Leader Owes Their Team — the obligations a leader takes on that never show up in an org chart. When we treat a struggling employee's fear as their problem to get over, we've quietly set down the part of the job that was ours to carry.

And if you're the one who's resentful

One more, because the survey named a group it called the resentful — the burned-out, checked-out people who feel pressured into all of it and have started to hate it. If that's you, here's the reframe I'd offer: the problem is probably not AI, and it's probably not you. It's more likely the room you're standing in.

I know that one from the inside. I started my career in a department run by a genuinely bad manager, working eighty-hour weeks, and I was miserable in a way that bled into everything — I was convinced I hated the city I lived in. Then I left for a good company and discovered I liked my hometown just fine. The job had been coloring my read on my entire life. So if AI feels like the thing making your work unbearable, it's worth asking, honestly, whether AI is really the thing. Sometimes the most productive move available to you is a different room.

None of this lets any of us off the hook for our own craft — that's a separate obligation, and one I'll come back to later in this series. But the split this survey found isn't destiny. It's a set of conditions. And conditions always have an owner.