Have You Tried AI Role Play?

Have you tried role playing with an AI yet? Not asking it to write something. Asking it to be someone. The skeptical employee in the restructure conversation. The sponsor who keeps saying yes and doing nothing. The manager who has to cascade a message they don't believe yet.

I tried it. I expected a gimmick. What I got shocked me.

I had asked it to play a sponsor, and it pushed back almost immediately. No warm-up, no easing in. It stuck to the sponsor's story and would not let go of it. At one point it looked me right in the argument and said, "What do you need from me? I did the town hall."

It made me stop and examine my own communication style, specifically the gap between how open I believe I am when someone pushes back and how open I actually am. That's the part nobody tells you about AI role play. The surprise isn't what the machine knows. It's what it reflects. A deck never argues with you. A comms plan never gets defensive, changes the subject, or goes quiet the way a person does. An AI playing the person does. And when it responds to you, you hear your own style coming back at you, sometimes for the first time.

The Cost of Never Practicing

Think about what practice has always cost in our field. Rehearsing a hard change conversation meant a facilitator, a workshop slot, a role-play partner reading from a card, and one nervous attempt in front of peers. Most managers got zero reps before the conversation that mattered. And the gap shows. A Harris Poll survey found 69% of managers are often uncomfortable communicating with their employees at all. A Leadership IQ survey of 689 HR executives found HR would trust only 35% of managers to handle a performance conversation with a truly difficult employee without HR in the room. DDI assessed over 70,000 manager candidates worldwide and nearly half failed to demonstrate effective conflict management skills. Only 12% showed high proficiency.

Now the reps are unlimited, private, and available at ten at night before the real meeting at nine the next morning. Announce the restructure to the machine four times. Get it wrong the first three. Nobody saw.

The practice payoff is documented. PwC ran the largest enterprise study of immersive soft-skills practice and found learners were 275% more confident to act on what they learned, and finished training up to four times faster than classroom learners, partly because they could repeat the scenario as many times as needed with no fear of being judged. That was VR simulation rather than AI role play, but the mechanism is identical: safe reps, no audience. Workers get interrupted 275 times a day. Practice makes people 275% more confident. The same number sits on both sides of the problem.

Consider, too, who is already doing this, because it isn't the laggards. The people rehearsing with AI are the engaged ones, the learners, the managers who genuinely want to get the conversation right. In our ChangePro work, that's exactly who shows up: managers, HR partners, and team leads who are personally inside a change while steadying other people through it. If your L&D strategy hasn't met them there, they went without you. That should change how we build training: less content to click through, more conversations to survive. A module tells people about resistance. A role play lets them feel their own reaction to it, and that reaction is the thing that shows up in the real room.

In AIM terms, this is the tool the deck could never be. Communication addresses information. Role play builds ability and confidence, two readiness conditions no announcement has ever moved.

The Rehearsal Is Not the Performance

Now the caveat, and it's the whole second half of the point. The rehearsal is not the performance.

You can live in the AI world, practice endlessly, refine your phrasing until it gleams, and still change nothing, because nothing changes until you have the conversation with the actual person. The machine can't read the silence in the room the way you can. It can't notice that your best team lead won't make eye contact this week. It can't build the trust that makes someone tell you the real objection instead of the polite one. Practice that never leaves the simulator is just a more sophisticated way of hiding, and we already have plenty of those. I've written about them: the decks, the plans, the beautifully targeted emails nobody reads. The avoidance data makes the stakes clear. 67% of managers regularly avoid or delay giving critical feedback. Unlimited private practice only matters if it ends the avoidance instead of becoming a more comfortable version of it.

Practice on the Machine, Deliver to the Human

So the sequence matters. Practice on the machine. Deliver to the human. Use the AI to find out where you get defensive, which phrases land wrong, what the pushback will probably sound like. Then close the laptop, find the actual person, and have the actual conversation, better than you would have without the reps. I said something on an ACMP webinar that I'll stand by here: I have not found AI ready to replace practitioners, but it makes me better at work I could not produce alone. The role play version of that sentence is simple. The machine improves the conversation. It does not have it for you.

The skills were always the point. Now there's finally a gym.

Use it before the conversation that matters, not after.