The real barrier to AI adoption isn’t capability. It’s identity.
When DevOps swept through software teams, the resistance surprised a lot of leaders.
The tooling was learnable. The case was sound. And still, developers dug in.
Why? Because for years those developers had been builders. They wrote the code, shipped it, and handed it on. DevOps asked them to become something else — builders and maintainers, on call for the thing they’d made, living with it in production. “You build it, you run it.”
That’s not a process change. That’s an identity change. And people don’t resist process nearly as hard as they resist becoming someone they didn’t sign up to be.
I think we’re watching the exact same thing play out with AI — and most leaders are misreading it.
The question underneath the resistance
When you roll out an AI tool and adoption stalls, it’s tempting to conclude the tool is too hard, or the training wasn’t good enough, or people are just change-averse.
But listen to what people actually say, underneath the surface objections, and you hear something different:
Will I still be good at my job?
Will I still be needed?
Will I still be… me?
The expert whose value was having the answer now watches a machine produce one in four seconds. The analyst who took pride in the craft of building the model now edits what the model spat out. The senior who climbed through years of hard-won mastery watches a junior with a good prompt leapfrog them.
None of that is a technology problem. It’s an identity problem. And you can’t train your way out of an identity problem.
This isn’t a hunch. It’s in the research.
Long before AI, I spent part of my academic career studying exactly this: what happens to people when organisational change asks them to shift who they are.
In one study, my co-author and I surveyed 181 employees in a nonprofit being pushed to trade its community, humanistic identity for a more corporate one — efficiency, budgets, commercial discipline. We wanted to know how that identity shift affected people’s openness to change.
A few findings have stuck with me for nearly two decades, and they map onto AI almost perfectly.
People whose identity already fit the new world were more open to change. People anchored in the old identity resisted.
Translate that to today: your AI-curious early adopters aren’t braver or smarter than everyone else. Their sense of self already fits a world of machine collaboration. And your resisters aren’t dinosaurs — you’re asking them to become a different person than the one who’s been good at this job for fifteen years.
The size of the gap mattered.
Where the new identity was a short step from the old one, people accommodated it. Where it was a leap — too far from who they understood themselves to be — you got dissonance, strain, and resistance. AI tends to ask for the leap. Fast.
And here’s the finding every leader rolling out AI should sit with: more information and more communication weren’t enough. The data was clear that navigating identity-based resistance needed something more sophisticated than simply telling people more, more often. Which is exactly why your town halls, your FAQs and your enablement emails aren’t moving the needle. They’re answering “how does this work?” when the real question is “who am I now?”
Why “more training” keeps missing
Training addresses capability. It assumes the barrier is can’t.
But identity resistance isn’t can’t. It’s won’t — and often it’s a won’t the person can’t fully articulate, because it isn’t a reasoned objection. It’s a felt threat.
That’s why you’ll see people who are entirely capable of using a tool quietly decline to. Competence was never the issue. Meaning was. When the new way of working seems to erase what made you valuable — or worse, seems to clash with what you actually care about — no amount of competence closes the gap.
In that nonprofit study, the people who held the community identity didn’t resist because they couldn’t adapt. They resisted because corporatisation felt like a betrayal of the very thing that drew them to the work. Sound familiar? It’s the same instinct behind “AI makes the work soulless,” or “this isn’t what good looks like to me.”
What actually helps
If resistance to AI is an identity shift, then leading it well is identity work. The research points to moves that look very different from another all-staff email.
1. Name the shift out loud.
Don’t dress an identity change up as a tooling upgrade. Say it plainly: we’re asking you to move from being the person who does the work to the person who directs and judges the work. Naming a threat shrinks it. Pretending it isn’t there leaves people to wrestle it alone.
Shrink the gap. Resistance climbs with the size of the leap. So build a bridge from the old identity to the new one, and make the continuity visible: the judgment, the taste, the domain expertise that made them good — those don’t disappear with AI. They become the thing that makes their AI output better than anyone else’s. You’re not asking them to abandon who they were. You’re showing them how it still counts.
2. Use the group as a buffer.
This was the standout finding for me. In the study, people’s identification with their team — their local peers, their site — buffered the strain of the bigger identity shift. Some stayed only because of the people beside them. The lesson for AI: don’t make people renegotiate their identity alone, at their desk, in private fear. Move them through it in cohorts. Let teams experiment, fail and learn out loud together. Belonging is what makes a frightening shift survivable.
3. Make it a renegotiation, not a decree.
Identity isn’t a switch you flip; it’s something people actively work out through sensemaking and conversation. So create genuine dialogue — the real kind, mutual inquiry rather than persuasion wearing a consultation costume. People need room to author the new version of themselves, not just be handed it.
4. Protect the purpose.
Resistance spikes when the new identity feels at odds with what people value. Anchor AI to the purpose they already hold rather than against it, or you’ll trigger the “this is against what we stand for” reflex — and that one runs deep.
5. The reframe that changes everything
When AI adoption stalls, it looks like a technology problem. Then it looks like a training problem. Underneath both, it’s almost always an identity problem.
The leaders who get this stop asking “how do we get them to use the tool?” and start asking “who are we asking these people to become — and how do we help them get there without losing themselves along the way?”
That second question is change leadership. It’s slower, more human, and far more uncomfortable than buying licences and booking training.
It’s also the only thing that works.
Helping people through identity-level change is the heart of what we teach. It’s why we built Leading AI Change — the Agile Way — for leaders who’d rather lead their people through the shift than drag them across it. Start here.
