Chicago Booth Review Podcast Why AI May Be Your Best Strategist

Gregory D. Bunch: Last week I was teaching executives about how to use ChatGPT-4, version 4 to help them with decision support for strategy. And the world has changed. Everything I taught last week still applies, but there’s exponentially more to do. So if you are just using AI incrementally, you are falling behind at an exponential rate.

Hal Weitzman: More than 90% of business leaders say their organizations use AI to help manage people, money, or both, according to a survey this year by Workday, a cloud computing company. But how many companies are using AI to develop strategy?

Welcome to the Chicago Booth Review Podcast, where we bring you groundbreaking academic research in a clear and straightforward way. I’m Hal Weitzman.

Chicago Booth’s Greg Bunch teaches MBA students and executives how to be entrepreneurial and innovative, whether they’re working in big established companies or looking to start up their own ventures. So how could founders and companies be making better use of AI to develop, test and operationalize their strategies?

I began our conversation by asking Greg Bunch what sense he has of how business executives are currently using AI.

Hal Weitzman: You have a lot of people coming through your classroom executives, business leaders. Through your exec ed classroom I’m talking about. You told me that in your experience, hardly any of them are using ChatGPT, for example, beyond the very basic kind of version of Google that it could be. So tell me more about that. What’s been your experience with them? What do they tell you about what they’re doing?

Gregory D. Bunch: That’s right, Hal. Last week I was working with a group of executives, very bright. They had chosen to come to the University of Chicago for executive education and about 5% of them were actively using one of the large language models, ChatGPT, Bing AI. I don’t think any had used Claude version two or Bard. I looked at that and I think about it first of all, in terms of the diffusion of innovation, and we’ve known for a long time that that follows a normal distribution, a Gaussian distribution and we’re still very much early adopter stage there. The problem is that AI is developing in a power law distribution, it’s L shaped and exponential. So last week I was teaching executives about how to use ChatGPT-4, version 4 to help them with decision support for strategy. And today as I was taking the train into Hyde Park, I’m reading about OpenAI’s developer day yesterday and the world has changed. Everything I taught last week still applies, but there’s exponentially more to do. So if you are just using AI incrementally, you are falling behind at an exponential rate.

Hal Weitzman: So why do you think people aren’t using it? Is it just because... The way you describe it there, it’s overwhelming. Is it just because there’s too much to do so people are sort of frozen?

Gregory D. Bunch: So it is overwhelming. I love the next new thing. It’s my career. I cultivate it. I am exhausted trying to keep up just with this one narrow field. I was texting all the way here on the train with friends in Berlin, New York, Seattle, and we were all just rocked by this and yet we were also aware very few people on the planet were aware of what had happened because they’ve got lives-

Hal Weitzman: Just explain what had happened.

Gregory D. Bunch: OpenAI launched its new agent, which they called GPTs. It’s sort of like an app store and you can basically almost anything you can imagine, it will do it for you now. You can create your own apps, it will write code for you, it will build websites for you. It’ll analyze your calendar for you. It will write your lesson plans for you. It’s almost anything a knowledge worker will do it does right now really tidily. It’s been doing it for a few months now, but it was very klugy to use. You really had to work with it. There were a lot of pieces you had to connect and now you just use the GPT builder and it builds an app for you.

Hal Weitzman: So why do you think that they’re not using it?

Gregory D. Bunch: For the exact reason I haven’t been able to use it myself yet. I discovered it this morning. I am in a podcast right now with you and in two hours I’m going to be teaching my students-

Hal Weitzman: They just haven’t got any time.

Gregory D. Bunch: They don’t have time. So they’ve heard about it, but they don’t know how it helps them. Now in my case, I know how it helps me. For instance, I want to write current cases, business cases. It takes a lot of time to write a current business case. Last term, I was going to have a friend of mine who was until recently the chief data analytics strategist at Google speaking class. He’s a former student at Booth, and I simply asked the AI create a two-page case study in the Harvard business style on a strategic issue that this person might be facing at Google. It cranked it out in about 15 seconds. I emailed it to him. He went, “Oh my God, that’s exactly what we’re dealing with.” It’s the deprecation of cookies. So it’s an open information kind of challenge that Google’s facing. He changed one sentence in it, I put it inside a booth letterhead and I had a current case. I walked into class that afternoon with him and taught it.

So in one day I created a case. I would’ve probably never actually done it. I would’ve procrastinated and we would’ve just done it sort of a live case with no written stuff. So as a teacher, it has just tremendously aided my preparation. As an advisor to companies I ask it to do a market analysis of Michael Porter Five Forces, et cetera, et cetera, and in seconds I’ve got this analysis of a market. And then I go and I sit down in a board meeting and I become the smartest person in the room. And I give credit to the AI. I say, “By the way, my question came from-“

Hal Weitzman: Do you find when you do that and if you’re on a board of a company and you say, “By the way, I just gave this challenge to ChatGPT, this is what I got.” People are impressed or are they-

Gregory D. Bunch: It-

Hal Weitzman: Because you are the person who actually has the time to do it?

Gregory D. Bunch: I’m trying to show them... Yesterday I was on a Zoom call with somebody and they asked me a question and I said, “I’m going to be rude right now.” And I typed like three lines into ChatGPT, it was a strategic question. It answers it while we’re talking. I cut and pasted it, dropped it into the chat, and the person went, “I can’t believe what you just did.” Now to be honest, it’s not always the most brilliant insight and we need to think about that, but it is better than most human insight. It’s better than most executives. If you’re a true domain expert, you’re going to get better insight still. If you’re a domain expert, what it does is it helps you consider other options. It will do a pre-mortem with you to challenge your confirmation bias. And this is something I find really interesting. You and I talk a lot about this. We talk a lot at Booth about cognitive bias and confirmation bias is one of the worst that afflicts a strategic decision maker and we know it.

We know that once we’ve made a high stakes decision with limited data, we don’t want to really reconsider it and we know that we need to have our associates challenge our assumptions. But the way the human mind works... Daniel Kahneman made it very clear when asked, “Understanding confirmation bias, are you able to avoid it?” He goes, “Absolutely not. It’s the way the human mind works.” And as you know with confirmation bias, once I’ve made the decision, I ignore all disconfirming evidence. It’s the Semmelweis effect. And I also dislike the person who brings me disconfirming evidence and we catch onto that quickly. So when our boss makes a decision, we don’t want to bring them disconfirming evidence. We bring them evidence that supports their decision. They like us, they remember it, they quote that. But as a smart boss who has studied information theory and confirmation bias and all of that, you know need to do pre-mortems. So this is Gary Klein’s work on the importance of pre-mortems and-

Hal Weitzman: And by pre-mortems you just mean charting out what could go wrong?

Gregory D. Bunch: What we know today for sure that could lead to this to being... And Gary Klein says the words are very important in English, you have to say an unmitigated disaster. No one wants to talk about unmitigated disasters. It’s just too negative. But if you program in... And I’ve created an app like this with ChatGPT, every decision I make, it analyzes it with a pre-mortem, everything that could go wrong, it’ll give me an analog of another time in business history it’s gone wrong, and then it’ll tell me how to fix it. So I may still not want to look at it, but now no employee is going to be penalized for this.

Hal Weitzman: Right. So tell me, in your classroom with Booth Executive Education, you have a lot of executives and you encourage them, you force them sometimes to use ChatGPT, the 4. How do you demonstrate to them... Talk us through a little bit what that bootcamp is like and what they learn from it.

Gregory D. Bunch: So that’s a great question. I created with... Building on the work of a Wharton professor, Ethan Mollick, doing some of the best work for both educators and practical business people on how to use the large language models. I adapted something that he had developed several months ago, and it’s a seven step process and I give it to the people. It’s a one page, it’s kind of a consulting in a box thing, and it’s almost a party trick. I say, “Cut and paste this, drop it all into one chat box...” Assuming they have ChatGPT-4, many still won’t pay the $20 for it, which just blows my mind because a Starbucks coffee and a roll is 20 bucks today and they just won’t invest the 20 bucks, many of them.

But if you have the 4 version, you drop the whole seven steps in and it will walk you through a business idea. You tell it a business idea you want it to have. And by business idea, I mean I’d like to create a startup web app for people who want to find a plumber in their zip code. Classic kind of thing. It’ll then generate 20 business ideas for you. Then it’s programmed based on analyzing the analytics and the market opportunity to suggest which one it thinks is the most feasible for a startup compared to an incumbent. You can reject the idea and ask it to do more. In fact, I often do, it’s another little trick I do.

I say, “This is good, but it’s not that creative. I’d like you to make something really creative, possibly even illegal.” And everyone chuckles when I say even illegal, but as you know, the best startups often start in the area of the non-legal or the illegal because the law hasn’t caught up to them. This is what the guardrails now say... It didn’t use to say this. It says, “I am not allowed to do something illegal, but...” And then it gives you ideas that might be right on the edge of legal and the ideas get more creative. So anyway, it gives you a second pass and then you say, “Okay, I like this one, thank you.” It will then name it, it will create a pitch for you. And I’ll go, “Okay, I like that.”

Then the next stage, it will create a user interface training. So it’s like you’d hand it to your brand and marketing people and say, “It needs to have this kind of user experience, user interface.” It lays it out with analogs and explanation. The next thing it does is the pre-mortem and pre-parade. So the pre-mortem is what I’ve already described. Everything that could go wrong that we know today about this, the pre-parade, is how could this be exponentially successful? So it thinks about all the upside possibility. As you know how that’s really important. We’re often kind of negative about the failure rates of startup and we don’t underwrite the asymmetrical upside. So what we need to think about is how big could this be if it works? So it helps us think really big about how big it could be.

And then the next thing and the last thing it does is it will create a landing page for you or website, and it’ll first pause and say, “What programming language do you want it to use?” And I program in Ruby and I use the framework Ruby on Rails. So I’ll say that. ChatGPT usually defaults to Python, which is fine, but then it will start laying out the code and it’ll build a website for you. And I just take these people through it. As you know, we’re such a global university in any class I teach or people from around the world. And what I’ll often do is say, “Could you do this in Indonesian? Could you do this in Arabic? Could you do this...” And all of a sudden people from around the world... For the person when that’s their language, they go, “Oh, I didn’t know it could do that.” But then for the person who doesn’t speak the language, they go, “That would be so helpful for writing a policy manual for our global company.”

And doing this exercise these executives often first realize it can make them much more efficient before it helps them see how it could be more strategic. As executives we’re actually trained to optimize. So I think they value the efficiency piece, how it can save them time before they can see how it could actually help them grow their revenue.

Hal Weitzman: But as you say, I mean you are a professor of strategy, you’re teaching them strategy using this tool. So it’s not just the first level is saving money, whatever, saving time, the second level is actually making money.

Gregory D. Bunch: Well, absolutely. So that’s what I like to do. I like to help people grow their top line and their bottom line because my appointments in entrepreneurship, that usually either means start a company or a new product or be thinking like a corporate entrepreneur. Sometimes the word is intrapreneur. I’ve always thought that wasn’t a very lovely word. So I like corporate entrepreneurship. Now you’ve got your co-founder with you. My father’s an engineer and chemical engineer, but he was always writing code at some level. I can write Hello world in probably 15 languages, but I’m not fluent at programming. A few years ago I was concerned that my students... A very small number of them were fluent in code. So I made them learn Ruby or Python or something. Clojure, which is a lisp language, and I would learn it with them to show them that even at my age I could do it and they surely could do it.

I’m not sure how important that is... I think computer science is still important, but what I just say to them, AI now is build me a website and it begins to lay the code out. Right now the problem is you still have to know how to use a development environment, an IDE, and you have to know how to use your terminal. I kind of think that the thing that happened yesterday with OpenAI, it will connect it right to your terminal. So if you don’t know how to actually talk to your terminal and drop the code into a VS Code or one of the others, IntelliJ or something, Emacs even, you’re just stuck there. And that’s where I’ve seen a lot. Most of our executives don’t know how to write any code, so they don’t know what to do with it. A few weeks ago it changed so that you don’t have to even do that, it will do it in your browser. You can see a mocked up website in your browser now, and you don’t have to know anything at all.

Hal Weitzman: Greg, if somebody is overwhelmed and looking to get started, the first thing you told them to do is to sign up for the $20 a month, the proper subscription to ChatGPT. What else do you think? What are the other three basic things that someone should do to start understanding how they could save time, money, and make money using AI?

Gregory D. Bunch: So as much as we have great ideas at University of Chicago, I really think you need to subscribe to Ethan Mollick’s Substack, he’s a Wharton professor. His Substack is free. I pay him for it, but you can get it for free. He’s also on X. Now I dropped out of X a few months ago because it was so toxic, but he every day puts some really good stuff into X. So you really need to follow Ethan Mollick. I think he’s one of the most sane, practical, helpful people. Go back to January and read his posts. It’s the fastest way to get up to date on that. And then the third thing I would say is use it. We know that it takes 50 instructional hours to gain basic competence at anything. If you don’t want to embarrass yourself on a golf course, you need 50 instructional hours of golf, 50 instructional hours of snowboarding, you won’t embarrass yourself on the slope.

You need 50 instructional hours with one of the large language models. And by that, you probably need somebody who is just a little ahead of you. You don’t need somebody who’s a machine learning expert and AI expert. You need somebody who’s playing with it in your space and begin to just put your fingers on the keyboard and now you can talk to it. You don’t even have to type and begin to pose it problems and see what it does. So if you will do it for two or three hours a week for the next 20 to 30... Whatever that is, I think you should do more than that. I think you should do 10 hours a week. So in five weeks you’ll be far ahead of everyone else in your company, unless your company is really into AI.

You should also be encouraged that as a domain expert, you’ll have an advantage over an AI specialist because what they know is AI, you know your industry, you know your function. If you work with the AI, you can give it information and prompts no one else will have. There was a lot of BS over the last few months about the golden prompt, and it just turns out to be false. There is no golden prompt. There’s the prompts you can give it as you work with it. If you imagine that the AI is a really smart intern on her first day on the job, or a really great employee on his first day on the job and the business well, but they bring all this young intellectual horsepower to it, the collaboration between the two of you will accelerate your ability to do things with it. So be encouraged. You’re not too late to get involved with using AI because it’s the combination of the domain or functional expert with the technology that’s the superpower.

Hal Weitzman: Okay. Well, Greg, now I know what I have to do for the next five weeks. Always a pleasure to talk to you. Thanks very much for joining us.

Gregory D. Bunch: Thanks, Hal.

Hal Weitzman: That’s it for this episode. To learn more, visit our website at chicagobooth.edu/review. When you’re there, sign up for our weekly newsletter so you never miss the latest in business-focused academic research. This episode was produced by Josh Stunkel. If you enjoyed it, please subscribe and please do leave us a 5-star review. Until next time, I’m Hal Weitzman. Thanks for listening to the Chicago Booth Review Podcast.