View full episode transcript
As AI tools spread through classrooms, many educators see a “cheating factory.” Kellogg Professor Sébastien Martin sees things differently: as a chance to reinvent how and what we teach. In his operations management class, students trade PDFs for lifelike chatbots, log conversations with Socratic AI tutors, and use AI to explore the kinds of social and political dynamics that students normally have to learn on the job. But Martin also grapples with harder questions: What about public schools with fewer resources? When should kids use AI at all? And how can students stay hopeful and true to their own interests in a wildly uncertain future? (Part 1/2 in our back-to-school series.)
Further Reading:
AI Is Teaching the Next Generation of M.B.A.s the Classic Case Study
Episode Transcript
This is a computer-generated transcript. While our team has reviewed it, there may be errors.
Jess Love: This is Life, Automated, the show that explores how to live, work, and make decisions in a world increasingly shaped by machines. By now, I’m sure you’re aware education has been hit hard by AI. It’s not exactly a secret why. You remember being a student. Maybe you are still a student. You work, you learn, you get your grade, your degree, and then you leave school better prepared for life.
Or at least that was the deal. Today’s students essentially have the world’s greatest cheating machine right at their fingertips, and that part about needing to learn to prepare for life, says who? If AI can do it, will you ever even need to? So maybe it’s no surprise that students are some of the heaviest adopters of AI, a trend that AI companies are very happy to facilitate, thank you.
And this puts educators in a tough place. I want to be upfront about this. Education is changing in ways we still don’t understand, and educators in universities, high schools, middle schools, are on the front lines. Many are not happy about AI. They didn’t ask for it, and they have real concerns that it will harm their students.
Next episode, we’ll go deep into the challenges they face. But today, something different. Today, we’ll be talking to someone who looks at generative AI tools like ChatGPT, Gemini, and Claude, and sees mostly opportunities.
Sébastien Martin is a colleague of mine here at Kellogg, an associate professor of operations. He studies AI, he teaches MBA students about AI, and in his operations management classroom, he’s replaced a lot of pen and paper-based activities with AI-powered simulations. In these simulations, rather than AI making it easier for students to finish whatever assignment is on hand, it slows them down, makes them work even harder.
I got to experience this firsthand.
Jess Love: I feel like maybe you’re not doing your part.
Sébastien Martin: Oh, she’s not happy.
Jess Love: She says, “I am doing my part. That’s a job. What I’m not gonna do is pretend the existing system works the way headquarters imagines.”
Sébastien believes that AI has already made him not just a better instructor, but a happier, more creative, more inspired one, and his students agree.
Kellogg’s graduating class chose him as the winner of the 2026 David Bisenko Outstanding Professor of the Year Award. I’ve come to see Sébastien as a virtuoso of a strange new art form. And amidst the big, even existential challenges facing education, I think it’s important to tell these kinds of stories, too.
I’m Jess Love, and this is Life, Automated, a project from Kellogg’s Ryan Institute on Complexity, distributed by KQED.
My conversation with Sébastien began with him telling me how, as soon as ChatGPT came on the scene, he got busy experimenting, trying to see if it could help him improve his operations management class.
Sébastien Martin: In a way, you know, when you have 140 students at the same time, for example, or even more, 200, it’s hard to have individualized conversations with everyone with my time, and I only have this time where it’s me with all the students at the same time.
But if they all talk to an AI that personalize the conversation to their own background, I can task the AI to connect the class content with their background.
Jess Love: The idea is that a student could feed their CV into an AI chatbot that had been trained on Sébastien’s course materials, and the AI would ask some probing questions, ultimately connecting the dots between a student’s individual work experience and interests and what the course would cover.
And this was kind of, was it to just get the student excited about the class? Was that kind of the point of this exercise?
Sébastien Martin: So the point was for them to know even before they take it why we think it’s good for them to be there. Which changes everything because if you have the attention of people as you even just get started, then communication just flows a little bit better. And if you think about it, there’s no way for me to do this without AI.
Jess Love: Sébastien was such an early adopter of the technology that he was using it before most of his students to solve problems that had nothing to do with AI at all. But it didn’t take too long for AI to start causing problems, too.
Sébastien Martin: You know, AI came like a wrecking ball. And, you know, like, what’s even the meaning of a take-home exam anymore? You know, you can trust the students not to use ChatGPT, but we are at the point where you can just take a screenshot of the exam on your phone and get A+ for everything technical. And so it’s just unfair to ask the students to do it this way. And because, like, how do you even grade it, because some students would respect the rule and maybe a few will run out of time and ask the help of AI.
Jess Love: It’s been well over three years since Sébastien started experimenting with AI in his classroom. And these days, nearly everything about his course looks pretty different. Some of these changes have been spurred by AI making it really, really easy for students to cheat. But Sébastien says that other changes are more about problems with the old ways of doing things.
Sébastien Martin: Think about our old ways to evaluate, right? I have students coming from so many different backgrounds that care about so many different things, and I have to have one exam that’s graded by people that don’t necessarily understand their background. And so what does it push us to do? It pushes us to standardization, to always asking, like, very abstract questions about things, or to focus on math instead of, like, deeper understanding about the world, because these can be evaluated more easily. I love math, but it becomes just an exercise in itself, like, almost like a Sudoku. At the end of the day, the exam is just asking them to manipulate numbers because this is what I can evaluate.
Jess Love: So let’s get concrete here. With simple problem sets basically out the window, what does Sébastien’s class look like today? Well, for homework, Sébastien might require his students to converse with an AI tutor for maybe 30 minutes, walking through concepts and connecting them to their own experiences. That tutor might be trained in the Socratic method so that rather than giving answers like a standard chatbot, it pushes students to do the thinking themselves. Then, instead of turning in answers, students would upload a log of their chat to be graded pass or fail. Or maybe the tutoring goes the other way.
Sébastien Martin: The homework I just gave, like, a few weeks ago is the, student has to teach the AI what they learn in the last class.
Jess Love: Ah, interesting.
Sébastien Martin: You see that we say learning by teaching is, like, one of the best ways to learn. And as a professor, I can say that’s very true because I created a new class on AI and I learned a lot just by teaching it.
Jess Love: Of course, this technology is changing fast, and there’s really nothing stopping students from building their own AI agent or autonomous AI tool to interact with Sébastien’s AI, and turning in a log of that conversation.
Sébastien Martin: That’s one of the main obstacles our instructors want to get into AI, because it just feels it’s constantly evolving, and whatever I do now, I fix a problem, another problem will appear later. So it’s a very, very real concern.
Jess Love: Well, and this is hard because I’ve done, you know, way back in the day I taught some classes myself, and the first time you teach something, it can be kind of a train wreck, right? And then you kinda get better over time, and you learn what works. So how do you as a professor or a teacher keep up?
Sébastien Martin: I would turn the question around, and this is why I try to put my students in the loop as much as possible. It’s like, “Here’s the deal,” like, “What is good for you?” You know? “Why are we doing this homework? Why are we doing all of that?” The goal is not just to go for the grade. That would be the mistake.
Like, the goal is at the end of the day, you know, you learn something. And, you know, if AI is becoming more and more of a commodity that can do anything for us, our deep expertise and our personal thoughts and what makes us unique has even more value, because that’s what differentiates all of it, rather than the things that AI can just automate, you know?
And so I think at the end of the day, the answer to this question is also, it’s not only on the professor’s side, it’s also on the students. Like, why are you at Kellogg? What do you wanna get out of all of this?
Jess Love: Some of the changes that Sébastien has made to his class, they aren’t planned. He’ll be in the middle of teaching, he’ll get an idea for how to make something better, and just use AI to make it happen.
For example, just a week before our interview, his students were in small breakout groups using class time to discuss a case study, that staple of business school education. The time was flying by. Sure, the students were having wonderful discussions in their small groups, but Sébastien was getting frustrated, realizing there would not be enough time for the entire class to reflect on the ideas that the individual groups had generated.
So while the students were still in their breakout groups, Sébastien explained the dilemma to Claude Code, a popular AI coding tool.
Sébastien Martin: And so Claude Code was like, “No problem. Let’s do it,” and created something, and by the time students had sent their reply, we had a custom-made software, which was kind of like a constellation of ideas that popped up on the board, and we could see quotes from all the students and the ideas that they had in common and not in common, and this enabled a much deeper conversation because I was able to share all this information instead of just, like, talking to, you know, like, three, four students that were willing to share their ideas.
You know, it’s just a small thing, but it’s changed the way I taught this class, and it did use AI, and at the end of the day, it’s just about your creativity. Like, it’s a tool that is so powerful that, you know, sometimes, you know, as an instructor, you go, “Oh, it’d be so good to have this,” and then you just use it, and it creates a much better experience.
And you see here in that case, it increased the communications between humans. Like, of course, AI can sometimes take away from what we do, but in that case, I think it added to what we do.
Jess Love: Now, there are tools already on the market that can do this kind of thing. But what was unique was that Sébastien was able to create them on the fly, in that moment for that class. And the best example of how AI has changed Sébastien’s course has to do with the experience of reading the case itself.
Sébastien Martin: So we turned it into, instead of me giving them a PDF, I tell them that they can interview, “People” from the company and ask any question they want. And I give them access to the CEO, to the CFO, to, like, a frontline worker, to anyone, and then they have to understand the company deeply, kind of like consulting. You interview people, and you understand their relationship, and they all have very different point of views, and some of them don’t even wanna talk to you, and so on.
Jess Love: Well, and “people” are, of course, chatbots with, like, personalities and information that you’ve given them. This is the activity that has gotten Sébastien the most attention, including a big write-up in The Wall Street Journal. I wanted to try it, and Sébastien was kind enough to mentor me through his simulation. We pull it up on his desktop.
Okay, so I am a consultant at Proxima Health Systems?
Sébastien Martin: Yeah, you’re an AI expert, and they called you because they wanna have a big AI project, and they want it to be successful. And your point of contact is Maya Chen. She’s the leader of this After Visit AI program. Like, she has an AI agent she wants you to build.
Jess Love: Okay, and so I can just start talking to her and say, like, “Hey.”
Sébastien Martin: Yeah, she’s just waiting for you.
Jess Love: There’s a whole bunch of little avatars representing different people at the company, and I can pick one to have a text chat with. I see the avatar for Maya, my point of contact, and I begin to type.
‘Hi there. I’d love some clarification on what you want me to accomplish.’
Maya tells me that I, the AI expert consultant, will be building her company an AI tool that helps automate the after-visit forms that sales reps fill out after they visit their customers’ hospitals. And after a bit of back and forth with Maya about some of the players involved, including Marcus, a sales manager, and a bit of help from Sébastien, I start to get a sense of the nitty-gritty details involved in my assignment.
Sébastien Martin: And so for this, we have a CRM system, a customer relationship management system, which is like this big database where rep have to enter information. But they hate it. It has to be formatted, made in a very specific way.
Jess Love: Ah, I hate it already.
Sébastien Martin: They spend hours every week. And then so this is why, like, there is this sort of tension because Marcus needs them to enter this information, but they’re like, “Yeah, it doesn’t represent what we see on the ground. Why do you want me to spend all my time on this instead of with the customer?”
Jess Love: Okay, got it. Maya, Marcus, and others in the company want to automate the after-visit forms that the salespeople fill out, apparently to make things easier on the sales team. But then I chat with a member of the sales team, Dana, and Dana doesn’t necessarily think AI will help. She thinks the whole customer relationship system is messed up.
‘Dana makes a number of points. She says it’s like whiplash. She can spend 30 minutes in a unit with a nurse manager who’s stressed out of her mind. She catches the real issue. They come to a conclusion, and then she’s supposed to translate that into 12 little CRM boxes that don’t fit how hospitals actually talk, and then the managers wonder why the notes are vague.’
And Dana is a little suspicious of me, the AI consultant. Here’s how she puts it.
“What exactly are you trying to solve? Time, accuracy, or the manager chasing us for missing fields?”
Jess Love: Ooh.
Sébastien Martin: You’re in the middle of it right now. I am in the middle of it. It’s hard to help her without more information, understanding everything.
Jess Love: I decide to try to get Dana in line with some tough talk.
‘All right, Dana. I hear what you’re saying, but you do realize that you are part of this big organization that has a lot of really important goals? I feel like maybe you’re not doing your part.’
Sébastien Martin: Oh, she’s not happy.
Jess Love: She says, “I am doing my part. That’s the job. What I’m not gonna do is pretend the existing system works the way headquarters imagines.”
Sébastien Martin: How dare you question that I’m doing my job?
Jess Love: ‘Thanks, Dana. I appreciate the straight talk. I love straight talk. What would you like to see?’
And now I’m really starting to understand this scenario. It’s not just about coming up with some kind of AI-powered solution to data collection. It’s about understanding the social dynamics in the company.
Sébastien Martin: ‘Cause she’s very afraid. She says, “Not cutting heads.”
Jess Love: Not cutting heads.
Sébastien Martin: She’s really afraid that you’re coming with AI trying to take over her job.
Jess Love: Oh, yeah. That’s a good point.
Sébastien Martin: You know? Because we’re automating a part of the job that she and her teams do.
Jess Love: Once I understand that there’s anxiety in this company about AI taking jobs, and that this anxiety is driving so many problems, I start having more productive chats with the different characters, and we start making progress.
This is what Sebastien says this kind of AI-driven exercise is designed to explore. How to keep asking questions until you understand what’s actually going on, the problems behind the problem, the stuff that only rarely ends up on any official brief. I think there is a risk in education where I’m like, “This is the truth.” This is the answer to this problem.” And, you know, like, for example, if you go on Instagram and you have your newsfeed, you’re not asking for the next thing you wanna see. We’re just so used to just information being provided to us and not to have to think. And then you get into the middle of this and you have a “real person” you talk to, you actually need to decide what’s your next step.
But this is actually where all the thinking happens, right? You’re like, you need to be aware of what you understand and what you don’t understand. You need to, like, plan ahead. You need to also manage your relationship with them. And I think this is where I think the most. Actually, it’s what I feel like it’s the most important thing to think about, in life, you know, in general.
And I think we don’t have many tools that are not meant to just, like, provide us with the answer, but are meant for us to ask the right question and for us to learn more. And I think AI is a wonderful tool for that.
Jess Love: Here, the AI tool isn’t being used to make things easier on students. It’s not making it faster to read the case or offering up a streamlined summary. It’s slowing students down, making them think more. Or that’s the idea. So are his students learning more using this AI-powered simulation than they would’ve the old way, discussing PDFs or writing short essays?
Sébastien Martin: It’s hard to answer this question for sure. You can only capture it with exams and stuff like that. But AI cases, in the context of this class, were not something that was graded. They were something meant for conversations in the classroom. So then the second best I can give you is I had the most fun I’ve ever had teaching a case. Like, the participation of the students was through the roof.
They had so many ideas.
Jess Love: Sébastien acknowledges that it’s hard to know for sure. Just because students enjoy an activity, recognize it with a teaching award, doesn’t always mean that they’re learning more. But the students tell him they’re learning more because the scenarios feel real.
Sébastien Martin: The students say that it was very useful to them, especially the students who came from consulting. They say it’s exactly the type of thing I had to be good at. It’s great to see it in the context of AI and so on.
Jess Love: After the break, this strikes me as the dream scenario for AI in the classroom. But just how common is this dream scenario? Is it limited to the classrooms of super AI savvy teachers, or does Sébastien’s optimism extend to other classrooms, too?
I, like a lot of people, think about this technology as having the ability to both empower people and to disempower or overwhelm them. And it seems like for you, the primary experience has been one of empowerment. But I wanna point out, you are kind of a special case. You are somebody who studies AI, is very, very familiar with using algorithms, and was even before ChatGPT came on the scene, and you also teach highly motivated students at a top business school.
And this is not true of most educators. So my husband teaches high school at Chicago Public Schools, and he assures me that the general consensus among public high school teachers is not one of empowerment. I think a lot of schools are strapped for resources. They don’t always have the opportunity to train instructors in how to manage this technology, and a lot of them have just come to see this as a widespread cheating factory. And so I’m just kind of curious how you think about that.
Sébastien Martin: I mean, first, thank you so much for asking this question. Something I think about a lot, and first it’s very important to highlight that I come from a place of privilege in so many dimensions, as you mentioned. And, you know, because I, so I speak more and more publicly, like right now, about this topic, and I’m always like, “What should you say?” Because, you know, as you saw in this conversation, I choose to highlight the positive because I know that people are very aware of the negative. And I think seeing a bit of light at the end of the tunnel can help.
And it is true, you know, that I was lucky to start early and lucky to be in the position I was, and to iterate over time, and I made many, many mistakes. There are things that worked, things that didn’t work. I would say the message, and I really don’t wanna deny at all what public education is going through. I’ve worked a lot also with public education and I know how hard a job it can be, but also how rewarding. What I wanna say is give it a chance.
You know, this technology, for good or bad reason, is most likely here to stay. And if you iterate a little bit with Claude, ChatGPT, whatever, you’ll quickly realize after you overcome this barrier of, you know, not resentment, but this barrier or, like, this technology is really making my job complicated, you quickly realize that it’s a neutral technology.
Like, it can do a lot of harm, but it can also allow you as an instructor to create new things and to solve some of the problems. And at first you start very small. You know, you have all of this, like, I forgot how you were saying, like, this cheating factory thing. But then you’re like, “Oh, I can maybe use it that way.”
And actually that had a small positive impact. And it looks like a drop in the ocean. But then because you open this door, then curiosity gets in. What if I could do this? And what if I could do that? And all I can say, I know it’s a very minimal point, is there are good things. And yes, they take time, and they take trial and error, and all I can do is show examples. But I think there is value in just knowing that you can use this technology for good.
Jess Love: So, I am a parent. My kids are a little younger than the students we’ve been talking about, two elementary school-aged kids. And I personally do not want them using AI right now, partly because I think they already spend enough of their lives staring at screens. That is the perennial parenting battle of our age. But also because I want them to learn some really core skills about how to go through the world, and think about problems themselves before I provide them with the scaffolding that AI would provide. And so I wanna ask you, do you think there are some skills that people should have before they turn to AI to learn something?
Sébastien Martin: I mean, of course take everything that I’m about to say with a grain of salt, because I’m not an expert in that. You’re way more of an expert than me.
Jess Love: We’ll bring in some folks to rebut you point by point.
Sébastien Martin: But, you know, and also, I wanna highlight that we hear so much about AI these days. I think what’s super, super important to say is we don’t know. You know, like they say, wisdom is to know when you don’t know. Nobody knows where AI is going, okay? Like, but then it’s still good to talk about scenarios, and your analogy with the screens or the smartphone is a great one. You know, there has been a lot of debates, on not allowing screen times and, like, this leading to, children learning more to interact with the world and with each other.
And I think it could be great to do this with AI. I mean, that sounds like a great strategy to me. Then, the one caveat I wanna say on this is one pet peeve that I have is I don’t think AI means anything. It’s kind of you saying humans or electricity. Like, is electricity good or bad? Should we forbid kids to access electricity?
I think there are ways to use AI that could be net positives and there are ways to use AI that could be a disaster. So when we say allow kids to access AI, often we think ChatGPT, Claude, Gemini, like the basic thing you can ask all your questions. But for example, I was just talking to a parent and they were playing with their children to create their own song for a story that they invented.
And they used AI to like, put their lyrics that they co-created into a nice song and they were singing it together. Is it a bad use of AI or does it increase creativity? I heard another parent has ChatGPT to, you know, create like, these frames of drawing that you can draw into on things that were very personal. Like they just went on vacation somewhere and let’s actually draw what we went through. And you know you cannot buy a book that’s so personalized.
Jess Love: Like a coloring book of their exact vacation.
Sébastien Martin: Exactly.
Jess Love: Oh, that’s, that is kind of cool.
Sébastien Martin: You see? Like another example and maybe more the fringe of what’s acceptable is there’s a lot of, there are ways to turn ChatGPT into something that don’t think for you. So the way AI works is there are hidden instructions, and AI follows these hidden instructions.
Jess Love: That’s right. If you didn’t know, most chatbots have some version of hidden instructions to guide their use; and as a user, you can change the default settings for these instructions, specify how the AI should respond to prompts, including how much information it should provide.
Sébastien Martin: In the hidden instructions called a system message, you can ask the AI, “Look, you’re talking to a child. Here’s what we really want to avoid. We want them to feel for, to think for themselves. We want them to learn how to be independent. We don’t want them to learn how to rely on you instead of thinking.”
But maybe you can encourage their curiosity. And then you let the children talk to this version of ChatGPT, and the AI will just reply to their questions with other questions, and what do you think about this, and what do you think about that? Maybe it’s better than just staring at an iPad on YouTube, you know? Maybe it actually leads you to, to think a bit deeper.
Jess Love: For the record, I will not be tasking ChatGPT to chat with my children anytime soon, even in appropriately Socratic only ask questions mode. But that coloring book idea, it’s actually kind of a good one, and I take his point. We really don’t know. We don’t know how helpful or harmful AI will ultimately be as an educational tool. Stepping back a bit though, I wanted to know what advice he would have for students, the people on the other end of the educational process.
Sébastien Martin: I mean, the first one, which is, you know, the one thing I want us to remember from this podcast is, like, stay hopeful. Because, you know, there are reasons to despair from time to time, and especially as humans, we are really not good at dealing with uncertainty. Like, second, staying hopeful and being aware that we tend to only see the downside instead of all the unexpected upside that can come later.
Everyone is in the same boat. Like, it feels like it’s me that doesn’t know what the next step to choose and so on, like what studies should I go for, how should I learn and so on. But it’s not only me, it’s the whole society that has to adapt at the same time. And I think that at least for me, it helps me feel a bit better about it, like, the idea that it’s not me that doesn’t know what to do next and it’s all of us together.
Jess Love: Yeah, it’s like we’re all collectively gonna go through this crazy period together.
Sébastien Martin: Exactly. And so the more we can, you know, like help each other and so on, the better it is. And, finally, what do you do? You have this huge uncertainty. I’m saying, okay, take it with a grain of salt with all the feedback you receive about what you think you should do and so on, and all of us together.
But I think a flip side of that is, if you’re, it’s pure uncertainty and we don’t exactly know where things are going, maybe the best thing to do is to follow your heart. I know it sounds a bit cheesy, but you know, at the end of the day, I think it’s better, like, try for all of us to maybe not conform as much as possible and what’s expected of us and follow more, like, our instinct. Maybe there has never been a better time to do that because the uncertainty of the system is the system is gonna change so much that maybe the best path to follow is just what we really like to do.
Jess Love: Thank you so much.
Sébastien Martin: Thank you.
Jess Love: AI is forcing students and educators to adapt. And while a lot of attention has been focused on cheating and outsourcing thinking, I think Sébastien makes a compelling case that this adaptation will have some real upsides, that new kinds of learning experiences are now possible, including ones that ask more of students, not less.
But of course, AI is not the first technology to barge into schools. Next episode, we’ll look at how previous technologies have changed education, particularly for younger students, what those technologies delivered, what they failed to deliver, and what this could mean for today’s high schoolers, middle schoolers, elementary school kids growing up alongside AI.
I’m Jess Love. This is Life, Automated, a project of the Ryan Institute on Complexity at the Kellogg School of Management at Northwestern University, distributed by KQED. A special thanks to today’s guest, Sébastien Martin. We’re produced by Jesse Dukes, additional production by Nathan Ray, music by Stephen Jackson, recording by George Christensen and Will Feeney, marketing help from Ananya Mallapragada, administrative support, recording help, and wise counsel from Stacia Sliger.
