Michael Olenick: Rebuilding The Role Play

Latest in a series featuring short interviews with faculty at various B-schools focused on the use of AI in business and business education. See also Learning Requires Friction (Wharton), The Deck Is Dead (Stanford) and We Teach MBAs To Apologize About The Best Tool They Have (INSEAD).

Cynthia Wang teaches negotiations, teamwork, and general management at Kellogg, where she is executive director of the Dispute Resolution Research Center and of Kellogg Teaching Resources. Sébastien Martin is a professor in Kellogg’s operations group. Jake Cohen is associate dean for the Executive MBA and executive education, who came to Kellogg recently after time as a senior associate dean at MIT and a decade at INSEAD.

Forty years ago, Kellogg’s Dispute Resolution Research Center helped invent the thing every MBA has now done: the role play. You’re the buyer, I’m the seller, twenty minutes, go. The DRRC turns 40 this year and is celebrating at the Academy of Management with all its former postdoctoral fellows.

It is also taking its own invention apart. Ask Kellogg its position on AI and you get the brand answer, which is collaboration, AI as a teammate. Then you look at what they built. A machine you compete against on creativity. A grieving colleague you have to console. A CFO who contradicts the CEO. Kellogg didn’t sit AI next to the student. It put AI on the other side of the table.

Below, we talk about all of that, plus a randomized controlled trial running across roughly 30 universities to find out whether any of this teaches anyone anything, and why the PDF may now be the hardest kind of case to produce.

CYNTHIA WANG, DISPUTE RESOLUTION RESEARCH CENTER & KELLOGG TRAINING RESOURCES

Q: If you had to name Kellogg’s angle on AI, what is it?

We’re using AI as a teammate, as a collaborator. If you know about Kellogg, we’re all about the team method, so it fits our mission of teaching students to collaborate. How do you collaborate with AI? And what we’re really trying to focus students on is when AI is beneficial versus when it’s detrimental.

Q: Start with Beat the Bot.

Brian Uzzi developed it with a former Kellogg PhD student now in Hong Kong. Students test their creativity against AI, and in some scenarios the AI performs better and in others the individual does. The key takeaway is that even the best AI can never outperform the best human, because AI is trained on the average human. You’re not going to see those outliers. Rely too much on it for creativity and you might be shooting yourself in the foot. But it also shows that a human team working with AI does see the benefits.

Q: Tell me about the empathy tool.

That’s my colleague Matt Groh. Students practice empathetic communication with an AI conversation. The AI has recently lost a job, or was passed over for a promotion, or there’s been a death in the family. Students respond in real time and have to validate the emotions and demonstrate understanding. He’s very good at making the AI human-like, which is the hard part.

Q: You could always do that with two students in a room. What does the AI add?

The feedback. If you and I are doing a role play and I’m a human saying, let me tell you about all my problems, you can react, but there’s no feedback mechanism other than that one individual. With this you’re getting research-based methods of how to respond. It’s the same with the negotiation bot we built. It’s chat-based, and we created a dashboard that pulls the strong negotiation moments out of the transcript and the weak ones. It will never replace human to human role play. It’s a complement, because you can dig into the process in a way you couldn’t before.

SÉBASTIEN MARTIN, OPERATIONS

Q: What actually broke?

Before, when you saw something written, it was a proof of work or a proof of thought. That’s not the case anymore. If you see an essay, you don’t know whether the student thought about it. It’s kind of the deepfake of education, and it’s a real problem, because we built a lot of how we teach around a fact that isn’t true anymore. But AI didn’t bring only problems. At first you only see the problems, and that’s why there’s so much negativity. The other question is what are the thousand things you can do that you could never have done before.

Q: What has happened to the cost of building a case or a simulation?

In operations we know that when the cost of doing something drops, you do more of it. If it costs nothing to change the color of a car, you offer a hundred colors. There are three formats now: the PDF, the interactive one, and the case where you talk to the characters. There’s no longer one that’s harder to build than the others. Actually the hardest may now be the PDF, because AI is not as good a writer as it is a coder. My own mix is about 80% cases where you talk directly to the characters. In my agents class the case is a company implementing AI, and students consult for them by talking to the CEO, the CFO, the team. Some of those people are afraid their jobs will be replaced, and students have to figure out how to talk to them. The feedback from students who came from consulting is that it felt much more like real consulting.

Q: And you’re running a randomized controlled trial on all of it.

It’s one thing to say as an instructor that AI is working. It’s another to prove it helps rather than hurts. The research literature mostly takes one tool, an AI tutor for chess say, and tests learning with and without it. But the question isn’t whether AI is good or bad for education. It can be terrible. The question is, out of the ten million ways you could use it, if an instructor really tries to find a good one, does that work? So with a colleague at Kellogg and two at Michigan, Jun Li and Andrew Wu, we built a consortium of around 60 professors. Each writes a document committing to a specific change and what they’ll measure. Then they toss a coin. Heads, they do it this term. Tails, they wait. That’s what gives you causality, like a drug trial. We’ve been running it a year and a half and the first paper comes this summer.

JAKE COHEN, EXECUTIVE MBA & EXECUTIVE EDUCATION

Q: You’ve now run programs at HBS, INSEAD, MIT, and Kellogg. What do executives need on this?

Kellogg has always been very strong in marketing, strategy, leadership, negotiations. Now we need to help executives lead AI transformation, and that means across the board, not as a separate subject.

Q: How does Kellogg differentiate against the rest of the M7?

I keep telling my teams we need to think of blue oceans, to find new opportunities and make the competition irrelevant. I don’t necessarily want to compete head-on with the other M7 schools in the US. The question I keep asking is what we can do that’s different, and that means looking for markets other schools aren’t in rather than fighting for the same ones.

And, finally, a closing thought from Sébastien:

“If you think AI is a fad, or that it won’t affect things much in whatever you’re doing, it means you haven’t used it enough. Give me any field and any story and I can find new ways AI influences it. There is no world where everybody has a genie and nothing changes.”

---

Michael Olenick, JD, is a former INSEAD research fellow and founder of VSTRAT.ai. He has spent four decades building AI and expert systems; his research has been used at INSEAD and, before then, cited by Congress and media on how to resolve the 2008 era financial crisis. Michael currently lives with his partner, Anastassia, in Austin, Texas.

© Copyright 2026 Poets & Quants. All rights reserved. This article may not be republished, rewritten or otherwise distributed without written permission. To reprint or license this article or any content from Poets & Quants, please submit your request HERE.