Solutions · AI Role-Play
AI Role-Play for Leadership Training
Rehearse the conversation before it happens. With someone who argues back.
AI role-play lets a manager hold an unscripted conversation with an AI character who resists, deflects, and remembers what was said, then get feedback that quotes their own words back to them and scores the conversation against published leadership research. It is the practice that workshops, courses, and coaching never have time for.
Browser-based · Voice or text
Written by Frank Basinski, founder of LeaderCoreAI, after 20+ years in executive roles and years of running leadership workshops.
Featured in Training Magazine, May 2026
Updated October 2026
What AI role-play is, and what it is not
AI role-play is a live, unscripted conversation with an AI-played character, followed by feedback on how you handled it. The character is an underperforming direct report, a peer who is blocking you, a team that has gone quiet, or a board that has already made up its mind. You say what you would say in the real meeting. They respond to it, not to a script.
That rules out a lot of what gets sold under the name. A video with a quiz at the end is not role-play. Neither is a branching exercise with three buttons and one correct path, nor a general-purpose chatbot with a leadership label stuck on top, which is what much of the market has shipped. Two questions tell them apart: does the character get harder when you handle it badly, and does the feedback point to the exact moment where you did?
Why practice at all, when managers have already been taught what to do? Because taught is not the same as retained. Without practice, 70 percent or more of leadership training fades within months. E-learning retention drops below 20 percent within six months. Workshops combined with ongoing simulation practice hold above 70 percent at 24 months. The content was never the problem. The repetitions were.
What years of workshop role-play taught us
Before LeaderCoreAI, our team designed and delivered leadership workshops for years. Role-play was always the part participants dreaded. People played it safe in front of colleagues. Senior leaders found it undignified and quietly declined. The feedback afterwards was polite, because the person giving it had to work with you on Monday. Everyone learned something, and almost nobody changed how they handled the next hard conversation.
Workshops still do things AI cannot: facilitated reflection, peer learning, shared context. We say so on our comparison with traditional workshops. But a single day in a room cannot produce the volume of practice needed to change behaviour under pressure. That is the gap AI role-play fills: private, repeatable, honest, and available at nine in the evening before the meeting.
Every scenario in the library is grounded in a situation that I, my teams, or executives I have worked with handled firsthand: a leadership crisis during a product launch, a team conflict threatening a strategic initiative, stakeholders pulling in opposite directions. More on that background on the about page.
How an AI role-play session works
Four steps, 20 to 40 minutes, in a browser. No app, no headset, no scheduling.
Choose a scenario from the library, or describe a real conversation you have coming up. Set one of five difficulty levels. At Simple the character is cooperative and tells you what is on their mind. At Extreme they are evasive, emotional, and testing whether you will fold.
The character has a personality, a stake in the situation, and something they are not saying. They remember everything said so far in the session and dynamically respond to your communication style and approach, not to a script. Type, or turn on voice and talk. If you get stuck, or the difficulty turns out to be more than you bargained for, step out to the AI coach without ending the session. The coach exchange is left out of the grading. A typical session runs 20 to 40 minutes.
A separate grading pass scores the transcript against the research framework for that competency. The report quotes your own lines back to you, names what worked, and names what to change next time. Download it as a PDF or compare it with your earlier attempts.
The character will not react the same way twice. Try the approach the report suggested, raise the difficulty, or build your own scenario, with the challenge and characters that you will have to face tomorrow. Repetition with variation is what turns knowledge into behaviour under pressure.
See the interface, the difficulty selector, and a sample report on the features page.
What happened when senior leaders were given AI role-play and no mandate to actually use it
One of our business partners (CTG Group) embedded LeaderCoreAI as the practice layer in a five-month Executive Leadership Development Programme at Leonardo UK. Using our platform outside of the training workshops was entirely voluntary.
135
voluntary sessions by 18 VPs and Senior Directors
34 min
average session, 28 exchanges each
75 hrs
self-directed practice outside the classroom
The most-practised scenario was a conflict over three million dollars of project spend, attempted 24 times across the group. Leaders moved from Moderate into High, Ultra, and Extreme difficulty on their own initiative. One VP returned 25 times in 14 weeks.
“What I didn't expect was how much they'd use it voluntarily. They kept going back to test different approaches between workshops. At that level, people simply don't do that with tools they don't find valuable.”
“My initial reaction was that I didn't see the value. But then the scenarios became relatable to real life situations, and I found them genuinely useful.”
The full data, the scenario list, and quotes from the delivery partner are in the Leonardo UK case study. In a separate deployment with the management of the EU-CONEXUS European University consortium, one participant called it “the most effective emotional intelligence training I've ever encountered.”
What separates AI role-play from a chatbot with a leadership prompt
Seven things to check in any AI role-play tool, and how LeaderCoreAI handles each.
The AI counterpart gets defensive, goes quiet, changes the subject, or agrees in the room and disengages afterwards. If you ask a vague question you get a vague answer. If you lecture, they stop listening. The realism is in the resistance.
Scores are not a chatbot's opinion. Leadership simulations are graded against Kouzes and Posner, Goleman, and Edmondson. Influence simulations against Cialdini, and Cohen and Bradford. The same rubric applies at every difficulty level, so your scores compare across sessions.
No colleagues watching, no facilitator keeping score, no HR record of the attempt where you lost your temper. That privacy is why senior leaders use it voluntarily. It is also why the feedback can be blunt: no colleague has to soften it.
Workshop role-play has a facilitator who steps in when you freeze. Here, an AI coach is available on demand during the session, in text or voice, for the moment you are stuck or the difficulty has become overwhelming. It coaches the way a good coach does: it shares what it observes and offers a fresh angle on the situation, rather than handing you a line to say. What you say to the coach is excluded from the grading, so asking costs you nothing.
Every compatible scenario can be performed aloud with distinct character voices. Interrupt mid-sentence and playback stops. Practising the words out loud is a different exercise from typing them, and for some conversations it is the one that matters.
Describe the person, the history, and what you need from the meeting. The platform builds a private scenario around it. Managers use this for the performance review they have been putting off, the peer who keeps missing deadlines, and the stakeholder who has already decided.
Organisations see aggregate practice patterns: which scenarios a cohort gravitates to, where it is improving, and where gaps remain. This is the behaviour-change data most leadership programmes cannot produce.
How the grading is built and enforced: the research behind LeaderCoreAI.
AI role-play scenarios
Dozens of scenarios across the LeaderCore products, each at five difficulty levels. A sample:
Difficult conversations
- Coach a struggling performer who thinks the problem is you
- Deliver feedback to a high performer with a blind spot
- Run a performance review that includes a compensation disappointment
Conflict and alignment
- Mediate between two team members who have stopped speaking
- Align cross-functional leads around competing priorities
- Broker a decision when the loudest voice is not the right one
Change and pressure
- Navigate resistance during an enterprise-wide change
- Delegate a critical task mid-crisis without micromanaging
- Handle a public stakeholder crisis with incomplete information
Influence and first-time management
- Win cooperation from a peer you have no authority over
- Manage former peers after a promotion
- Persuade a busy senior sponsor in the four minutes you have
Who uses AI role-play
- HR and L&D directors who need a practice layer under an existing programme and evidence of behaviour change for the budget conversation. See LeaderCoreAI for HR directors.
- Training providers and coaches who want participants practising between sessions, with cohort data they can bring to the next workshop. See partner programmes.
- Individual managers with a conversation next week they would rather not get wrong. See LeaderCoreAI for managers.
When it is the wrong tool
- You need to transfer knowledge, not change behaviour. A policy briefing or a compliance course does not need role-play.
- You want a scripted branching exercise with one correct path. Our characters are unscripted and there is rarely a single right answer.
- You are replacing facilitation entirely. AI role-play is a practice layer. It does its best work between workshops and coaching sessions, not instead of them.
AI role-play: questions evaluators ask
What is AI role-play?
AI role-play is practice in which you hold a live, unscripted conversation with an AI-played character, such as an underperforming direct report, a resistant peer, or a stakeholder with a hidden agenda, and then receive feedback on how you handled it. Unlike a video, a quiz, or a scripted branching exercise, the character responds to what you actually say, so no two sessions run the same way.
How is AI role-play different from role-play in a workshop?
Workshop role-play puts you in front of colleagues, which makes most people play it safe and makes senior leaders opt out. AI role-play is private, repeatable, and available the evening before the real conversation. It does not replace a facilitator or peer discussion. It gives you the practice repetitions a one-day workshop cannot, and the two work best together.
How does the AI grade a session?
A separate grading pass scores the transcript against the research that defines the competency being practised. Leadership-development simulations are scored against Kouzes and Posner's Five Practices, Goleman's emotional intelligence work, and Edmondson's psychological safety research. Influence simulations use Cialdini, and Cohen and Bradford, instead. Every score points to the moment in the conversation it came from.
Is it realistic enough for senior leaders?
The best evidence is what senior leaders do when nobody is making them. At Leonardo UK, 18 VPs and Senior Directors completed 135 voluntary sessions during a five-month executive programme, averaging 34 minutes and 28 exchanges per session. Twelve of the 18 came back for more without being asked, and the cohort moved from Moderate into the High, Ultra, and Extreme difficulty settings on its own initiative. People do not keep returning to a simulation that feels fake.
Can I practise a real conversation I have coming up?
Yes. Describe the situation in your own words, including the person, the history, and what you need to achieve, and the platform generates a private scenario for it. You can run it as many times as you like before the real meeting.
Does AI role-play work with voice, or is it typing only?
Both. With voice on, each character speaks with its own consistent voice and you reply by push-to-talk. You can interrupt a character mid-sentence, the way you would in a real meeting. Typing remains available for anyone who prefers it or is practising somewhere quiet.
What happens if I get stuck during a session?
You ask the coach. An AI coach is available on demand throughout every session, in text and in voice mode, for the moment you cannot find a next move or the character has become harder to handle than you expected. It does not tell you what to say. It shares observations and helps you look at the situation from a different direction, so the next move is still yours. The coach exchange is excluded from the grading, so the report reflects the conversation you held, not the help you asked for. It is the equivalent of the facilitator stepping in during workshop role-play, without an audience.
Who sees my sessions?
Your transcripts and reports are yours. Organisations that run cohorts see aggregate analytics: which scenarios are being used, where the cohort is strong, and where the gaps are. They do not read individual conversations.
What does it cost?
Pricing depends on two things: how many named seats you purchase at once, and the duration of the subscription. More seats and longer terms bring the per-seat price down. For a quote, ask our sales team through the contact form. A free trial is available for individuals and for teams evaluating a pilot.
Try one AI role-play session before the next hard conversation
Pick a scenario, set the difficulty, and see what the character does when you handle it the way you usually do. The report will tell you the rest.
Runs in the browser · Reports in your language