The mistake most new managers make with AI is treating its arrival as a tools problem. You find the right tool, you run a quick training, you tell the team to start using it, and you are confused when adoption stalls and a couple of people get quietly resentful. The tool was never the hard part. The hard part is that you asked people to change how they work, in a year when “AI” and “your job is at risk” are spoken in the same breath everywhere they look, and you did it without addressing the only question they actually cared about: what does this mean for me.
Leading your team through the AI transition is a change-leadership problem wearing a technology costume. The friction is not about the software being hard to learn. It is about fear, identity, and trust, and those do not yield to a better tool or a louder mandate. They yield to a manager who leads the change deliberately, names what is really going on, and goes first.
This article lives in the AI-for-managers hub, but it is also the deep-dive for the tenth leadership skill, leading change, in the leadership skills for new managers pillar guide. The pillar makes the case that leading change integrates the other nine skills and lays out four moves at the new-manager altitude. This article applies those four moves to the specific change most likely to test you this year, and gives you the concrete artifact and conversations that make them real.
One honest place to start is with yourself. Before you lead the team’s AI use, it is worth checking your own, because the way you use AI is the model the team will copy. The Is AI Making You a Better Manager or a Lazier One? free quiz is a quick gut-check on whether your own AI habits are the ones you would want your team to inherit.
The real source of the resistance
When someone drags their feet on adopting AI, the surface reason they give is rarely the real one. “I don’t have time to learn it” or “it makes mistakes” or “my work is too nuanced for it” are often true and also a polite cover for the thing underneath: if this tool can do part of my job, what happens to me. People do not say that out loud at work, because saying it feels like admitting vulnerability, so it comes out as foot-dragging, skepticism, or quiet non-adoption instead.
If you treat the surface reason as the whole story, you will spend your energy on the wrong problem: more training, more tool demos, more “but look how good it is.” None of that touches the fear, and the fear is what is actually stalling the change. The manager who names the real thing, out loud and early, gets further in one honest conversation than ten enthusiastic rollouts.
The four moves, applied to AI
The pillar lays these out in general. Here is what each looks like pointed directly at the AI transition.
1. Be honest about what is changing and what is not
Vagueness is what manufactures the anxiety. “We’re going to start using AI more” tells people nothing and lets each of them fill the silence with their worst guess. Specific is calmer than reassuring. Say what is actually changing (these tasks, this workflow, this expectation) and, just as importantly, what is not (headcount, who owns what, the standard the work is held to). If you genuinely do not know whether AI changes the headcount picture, say that too, because a manager who pretends to certainty they do not have loses trust the moment reality contradicts them.
2. Name the legitimate fear before you solve it
Every change has a real cost to the people going through it, and the AI transition’s cost is unusually loaded: it touches job security and professional identity at the same time. Name it before you reach for the upside. “I know the honest worry here is what this means for your role, so let me address that directly” earns more trust in one sentence than any number of productivity statistics. The instinct to skip past the fear to the exciting part is the instinct to resist. The fear does not disappear because you did not mention it. It just goes underground and comes back as resistance. This is the servant-leadership move of making the change about them, not about your rollout, covered in the servant leadership in practice cluster.
3. Move first, visibly
If you are asking the team to integrate AI, you go first, in the open. Show your own prompts. Show where it saved you time and, more importantly, show where it produced confident garbage and you caught it. A manager who quietly avoids the tool while mandating it for everyone else has zero credibility, and the team notices immediately. Visible, honest use (including the failures) does two things: it makes the change safe to attempt, and it models the discernment you actually want, which is using AI without outsourcing your judgment to it. The when to use AI versus ask a human guide is useful raw material for modeling that line, and how to use AI as a new manager covers the practical habits worth demonstrating.
4. Build the new equilibrium on purpose
Change does not end at the announcement. It ends when the new way is simply how things are done, and that is months of small reinforcement, not a launch event. Keep AI use a live topic in 1-on-1s and team meetings. Celebrate good applications. Correct the failure modes gently and early. The transition is complete when nobody is talking about “using AI” anymore because it has dissolved into the normal way the work happens. Until then, your steady, repeated attention is what holds the new behavior in place against the gravity of the old one.
The artifact: write a team AI charter
The single most useful thing you can produce is short, and almost nobody does it: a one-page team AI charter that removes the ambiguity people are anxiously filling in on their own. It does not need to be formal. It needs to answer the questions the team is too uneasy to ask. At minimum:
- What we use AI for. The tasks where it is encouraged, even expected.
- What we do not use it for. The work that stays human, and why (client trust, judgment calls, anything where a confident wrong answer is dangerous).
- How we check its output. The expectation that AI drafts are reviewed, not shipped raw. Nobody gets to blame the model for an error that reached a customer.
- What this means for roles. The honest, current answer on how AI changes (or does not change) what people are responsible for. This is the section everyone reads first.
- How we share what we learn. A norm for passing around good prompts and useful discoveries, so the team levels up together instead of each person hiding their tricks.
Writing it forces you to get specific, which is the whole point. The ambiguity is the anxiety; the charter is how you remove it. Draft it, then refine it with the team rather than handing it down, because a charter people helped write is one they will actually follow.
The failure modes to watch for
Four ways this goes wrong, all common:
- Mandating without modeling. Requiring AI use you do not practice yourself. Fatal to credibility. Fix it by going first.
- Pretending it is only a productivity tool. When people fear for their jobs and you keep selling efficiency, you signal that you are unwilling to discuss the real thing. Name the fear instead.
- Banning it. Prohibition does not stop AI use; it drives it underground, where it happens without review, without shared standards, and without you knowing. Shadow AI is worse than managed AI. Set norms, do not set a ban.
- Over-trusting the output. The opposite failure: treating AI’s confident answers as correct and letting quality quietly erode. The standard is that a person owns every output, AI-assisted or not. The your AI isn’t underperforming, you’re undermanaging it piece covers the discernment this requires.
How to start this week
Two moves. First, the self-check: take the quiz linked above and be honest about whether your own AI use is the model you would want copied. Second, draft the AI charter, even roughly, even just the five bullets above filled in for your team. Do not publish it yet. Bring it to your next team conversation as a starting point and build it together.
Then have the one conversation you have probably been avoiding: ask each person, directly and privately, what the honest worry is about AI and their work, and listen without rushing to reassure. You will not have all the answers, and you do not need to. What builds trust through a transition is not having every answer. It is being the manager who was willing to name the real question instead of pretending it was about the tool.
Leading change is the skill that integrates all the others, which is why it sits last in the leadership skills pillar, and why the AI transition is such a complete test of it. The deeper calibration of when you are operating in execution mode versus stepping up to genuinely lead is in the manager vs leader cluster. Leading your team through this well, over months, will shape how they experience you as a leader more than any single decision you make this year. The good news is that the hard part was never the technology. It was the willingness to lead the people through it honestly, and that has always been available to you.