Generative AI w/ Peter Cohan and His New Book: Brain Rush | builders podcast

0:00It can help you overcome that writer's block or the creator's block. Having said that, there's nothing that will help you compete if you do that, because this is accessible to anybody. Anybody can use it. In order to compete, you have to provide value for a customer that other competitors cannot immediately copy.

0:24Peter Cohen is a professor at Babson College. He's a strategy consultant and an author of a recent book called Brain Rush about generative AI. If you're at all interested in ChatGPT and how it will impact your business, this is an episode you won't want to miss.

0:47Today's episode is sponsored by Intuit's QuickBooks Payroll. I use Intuit every year for my taxes with TurboTax, and I'm excited to talk to you about this product for small businesses. So as your team grows, so does your payroll, not just in terms of expenses, but also in terms of the time it takes to manage. When managing payroll becomes a challenge, QuickBooks Payroll gives small businesses like yours an easier way to get it done. Pay your employees, file payroll taxes, and track time and attendance all in one place. Intuitive software and handy automation tools reduce the need for manual data entry, which frees you up to focus on more pressing tasks like running your business.

1:26And because of its same-day deposit capabilities, QuickBooks Payroll supports your cash flow by enabling you to hang on to your money longer without delaying your team's paychecks. That's great for employee morale and also your business's financial health. Visit quickbooks.intuit.com backslash payroll to learn more about how QuickBooks Payroll could make all the difference for your team.

1:55Hi, Peter. How are you? Great. I'm so happy to be on your podcast today. Yeah, me too. I'm really excited about the topic we're going to discuss today, generative AI. Probably many listeners of this podcast know what it is, but let's start by defining it, Peter, if you could do that. And then also, what inspired you to take this topic to write your new book, Brain Rush, which is about generative AI and how to invest in it and compete in a world of generative AI? Generative AI is different from other kinds of AI in that it allows a person to type in a natural language question. Instead of having to have complicated coding to get information, just say something that comes to mind in multiple languages, and it will return some creative response that will sound very cogent and some of it may be correct.

2:49It's very unique in that respect. There's no other technologies out there where anybody who can write a simple sentence can go into this computer and get this kind of creative response. And that's what generative AI is. And that's what makes it different. Yeah. And I've been I've been using it a ton, which I know I know you have. And it is really interesting to, quote unquote, talk to. You know, you might say something like, give me a restaurant recommendation for Chicago next week. And the answers are vibrant. They're creative. They might not always be correct, but it's a really, really interesting technology.

3:28What inspired you to write the book? Well, I have a long history in artificial intelligence. Many, many years ago, one of my first jobs, while I was in graduate school at MIT, I got a job at a consulting firm founded by some Sloan School professors, Sloan School being the business school of MIT. and five Sloan School professors started this consulting firm. I went to work there. I was incredibly happy working there. And one of the partners or one of the founders of the firm, a former MIT professor, decided to start an AI company. And he invited me to be one of the first people on the team. And I found this to be a very exciting experience.

4:09I mean, all the technology back then, this is decades ago, MIT was really heavily focused on artificial intelligence. There was this programming language called LISP, which is short for LISP, LISP processing. And it also makes me LISP when I try to say it. But it also was basically artificial intelligence many, many years ago. And that business that we started failed. And it was a business that was trying to do something called expert systems for personal financial planning. And expert systems, I think, require some explanation. I don't think people know what that is. Basically, what we did was we would find an expert in personal financial planning, which is essentially, I guess, helping people to figure out how to save and how to plan for retirement, things like that.

5:00There are people who actually spend their whole career doing this and they have expertise in it. So what expert systems meant you would interview these experts and then you would try to convert their decision rules into computer code so that you could basically give you a series of questions and take you through a process that would lead you to a financial plan. So that was what the point of the company was. And the company got funding from a large insurance company and essentially ran out of money because personal financial planners did not see any value in the technology. They may have been a little bit worried that it was going to take their job, but it really didn't create enough value to make people want to pay enough money to keep the company in business.

5:40So that was very early in my career. And I found that to be a very powerful insight, which is very common with technology companies, particularly companies that are founded by engineers who love the technology and maybe in some cases have less of a feel for how does that technology create value? um so that was sort of a very important reason why i wrote the book because i had already experienced uh artificial intelligence and i came at it with this a somewhat jaundiced view that unless it creates value it could be just pure hype and not not really be a business um the next there's another thing that happened which was that um basically i worked for a consulting firm uh after i graduated from business school founded by michael porter at harvard's business school who's a strategy expert and did strategy and wanted to help focus my attention on helping high-tech companies figure out how to grow.

6:37So I started my own firm several decades ago and have done about 150 consulting projects for high-tech companies, helping them to figure out how to take advantage of the growth opportunities created by new technology. So I've really focused a lot on that. And I wrote three books about the dot-com bubble, the dot-com bubble, starting in the 1990s. So I had this experience of doing this artificial intelligence startup. Then I had the experience of watching a very big wave of new technology back in the 1990s, which was the internet boom, and really observed it very closely to write the books. I had to do a lot of interviews and talk a lot of people.

7:20And then as soon as the third thing is that really triggered me to write the book was in May 2023, what happened was that, of course, NVIDIA, which we talked about earlier, NVIDIA announced a forecast for the second quarter of 2023 that was way, way more than people expected. So much more. And I realized I knew that ChatGPT had been out a few months, but until I saw how much people were willing to buy the semiconductors that NVIDIA made and how much interest there was in it, I didn't realize that this was going to be something that was real. And so as soon as I saw that, it was like I immediately realized this could be something as big as the dot com or bigger, or it could be just pure hype.

8:07But I immediately wanted to write a book about it. And writing a book, I've written 17 books, and writing a book takes a lot of sustained effort. And in order to have that sustained effort, I have to really care about the topic. And I really felt in my gut that this was a really great thing to write about. So I got a contract, started writing the book, and it came out a few weeks ago. And so I was at Stanford over the weekend, went into the Stanford bookstore to see all the books that were going to compete with my book, and there was nothing. No Stanford professors had written any books on the topic of the business side of generative AI, and there were basically no books on it at all, which I thought was really, really interesting and made me wonder, is this not an interesting topic?

8:54Or did I just publish the first book on this topic? Or possibly, I don't think my book would necessarily have been in the Stanford bookstore, but somebody's book would be in there on that topic, I thought. That is surprising to me. And maybe you were first to market. When you're talking about early in your career and how personal finance managers didn't want the technology or didn't want how it was presented, it makes me think a little bit about where I am now and where a lot of people are now with thinking about how generative AI impacts their career. So I'm a marketer by trade. And what you talk about in the book, investing and competing based on generative AI, resonate with me.

9:43I think a lot of us in marketing had kind of an existential question about our jobs when we saw the promise of ChatGPT. Could a CEO or CFO just say, write me a marketing tagline for NVIDIA and they don't need a marketing department anymore? So let's talk about the compete aspect first. Do you think people are competing with AI for their own jobs? I mean, I'm sure some are, but how do you respond to that topic? I definitely feel that if you are going to think you can kind of hope that generative AI goes away, so don't pay attention to it, you may be hurting your career. So I think it's important to engage with it.

10:24At the same time, having engaged with it extensively for teaching my students about strategy and coming up with lots and lots of interesting examples of how people are using it, what I've discovered is that there are some important limitations to what it can do, you know, specifically for marketing. It is good, I think, at overcoming. And I'll tell you, I developed this concept called the value pyramid. I'm really interested in the concept of value. And value to me means essentially the benefits to the customer for the amount that they're paying for the product. And so in order to win market share, you need to offer a customer a product that has more value than competing products, which is simple to say, and the complexity of it comes into the details, which is that every product, every customer involved in a complex decision has a different criteria that they're using to decide which of the competing vendors is the best one to go with.

11:22So I developed this concept of value pyramid. Bottom level, the most common use of generative AI right now, and I'll put it in terms of marketing, is overcoming creator's block. I don't say writer's block, but the concept is the same as writer's block, except that there's so many other media that it can help you overcome your block of getting started. It can help you write. It can help you produce videos. It can help you produce images. It can help you code all these things that can help you do. So, you know, getting that first draft is very hard. And a lot of people are so overwhelmed by the stress of trying to come up with that first draft.

12:00They don't do anything for a while. So generally, I can help you with that. It can help you overcome that writer's block or the creator's block. Having said that, there's nothing that will help you compete if you do that, because this is accessible to anybody. Anybody can use it. In order to compete, you have to provide value for a customer that other competitors cannot immediately copy. If you want a sustainable competitive advantage, you have to offer something that competitors cannot immediately copy, especially if it's successful. So I would say doing that is very common right now. I think many people are using ChatGPT for many things related to marketing, definitely.

12:42And I think that bottom level is sort of overcoming creator's block is useful, but not going to create a competitive advantage. The next level, there's three levels. The second level, which is less common, is using generative AI to improve the productivity of certain business activities, such as selling customer service. And coding is another one where you can improve the productivity of the people who are doing it. That sort of can get you into competing if you are using proprietary company data. But to give you an example of this, I've talked to many executives. Tell me how this technology is very useful for helping to make the early, you know, the first somebody who's just starting on the Salesforce, for example, who's been there for six weeks or a month or two months to to do the kinds of things that a very successful and experienced salesperson can do by sharing the knowledge base of that experienced salesperson with the new person.

13:49So they can essentially increase the close rate of a new salesperson and then bring up the overall close rate of an entire sales force. With customer service, it's the similar kind of thing. You have a new person who's, you know, I'm in the call center. They don't know how to deal with customers very well. instead of struggling without any guidance, they can have access to the successful responses to similar customer questions from the most talented of the customer service people. And that also raises the overall level of customer satisfaction. If that expertise is reflected in something that is unique to the company and the company has unique expertise, that gets you to something that potentially could give you a competitive advantage.

14:38But the ultimate top level, in my opinion, is something that just about nobody is actually doing right now, which is using generative AI to increase their revenue growth rate. One of the things that I have learned in I started going on CNBC decades ago when my first book came out and I've been studying what is it that makes people buy stocks? Why do stocks go up and down? I've been you know, I was taught something in business school about that, but it's not quite relevant. What is relevant is that you have to exceed investor expectations. Particularly, you have to exceed investor expectations on revenue growth.

15:17They really care a lot about that. So when you can use generative AI to improve the growth rate of your company in some way and do it in a way that exceeds investor expectations, that will raise your stock price, which will then create enough value so that in theory you could sell some stock and use it to pay back the investment that you made in the generative AI systems. And I'll give you an example of this because I just wrote about it earlier this week. I thought it was a very compelling example. It's sort of a marketing example. It turns out that consumers like to get weekly personalized recommendations of what they could be buying at a bargain at a discount from an online retailer or any retailer for that matter every week based on their previous purchases and based on their preferences and what other people with their preferences have bought.

16:15So, you know, in the olden days, they used to have these paper circulars, I guess, you know, go to a department store and they have these paper circulars with all the discounts for that week or something like that. This is sort of an online version tailored to each individual. And one of the examples that I did some research on, it turns out that on average, that example was increasing the revenue growth rate of the company using it by two to three percent, which is not negligible. I mean, it could be significant. It could be something that allows a company to exceed investor expectations. So this whole idea of using the technology in a way that, A, allows you to increase your revenue growth rate, and B, is something that is based on your proprietary information about your customers that the competitors can't copy.

17:07That is something that I would say is something that Generative AI could potentially create an opportunity for companies to compete and to win market share and to grow faster than competitors and to do better than competitors by using Genr of AI. Having said that, I don't know if there are very many of these applications actually being used right now. Well, I think it's important, and I want to hit in on this more for CEOs to maybe get some advice from you about where to think about investing in those areas. Yes. But before we get there, I want to back up a little bit and talk about the creator's block aspect.

17:46Because I agree, that is where most people are using it. And I also agree that if you're doing it in an undifferentiated manner, then you're not probably adding too much value to your organization. Google says something which I think is really interesting, that you're not competing against AI, you're competing against other marketers using AI. and um and the marketing that that either i've been doing recently or i've seen working combine generative ai with some unique proprietary insight something that i know that other people don't know and then when i can when i can inject that into a marketing message um and we're seeing this play out and we advertise certain things on facebook and and and what and whatnot um these ads that we recreate work better than the ones that Facebook auto-creates for us based on their machine learning.

18:44Yes. So, yeah, so it's really, really interesting. Your point about having proprietary data or unique insight that a marketing organization or a company has is super important. This is an idea that I've had for a long time. Many, many years ago when I was in business school, I worked on a consulting project for the New York Stock Exchange. And the New York Stock Exchange was essentially giving away for free real-time stock quotes information for traders. And they were trying to figure out, well, what can we do to actually make money doing this? And I did some research and I discovered that there was this company that was in an adjacent business that was averaging a 100 % return on equity, which is phenomenally high return on equity for that company.

19:33And I tried to figure out, well, what is it doing that is unique? Why is it able to do that? And the answer turned out that it was providing real-time price information on government securities. It had a monopoly. I don't know how it got it, but it had a monopoly on this information. And anybody who was trading government securities absolutely needed that particular company's product in order to do business. So I came up with this two by two matrix, one dimension of which is how how consequential is the decision you are making? How much money or how much risk is that is there if you make the wrong decision?

20:15So the consequence, the importance of the decision is one dimension. The other dimension is how unique the information is for helping you to make that decision. So the high, high box is what Tellerate was, which is the name of this company that had the 100 % return equity. It had information that was completely unique and better than anything else out there on the market for a very consequential decision about which government securities to buy. I mean, to me, there are more consequential decisions in the world, but it has a very capital intensive decision. And Quotron, on the other hand, was sort of in a, I would say, its information was not proprietary.

20:57Lots of people were able to get that real-time access to the New York Stock Exchange price information. The point is that that same two-by-two matrix pertains to generative AI, in my opinion. If you are giving somebody information that is related to a very consequential decision, and that information is unique and better than any other information out there, if you are in that particular box of the two by two matrix, you will do better than if you are in the other boxes. I want to dig into an investment, company investment in this technology. Before I do, I think it's important that we cover the level of trust either you have or someone should have, or I should have on generative AI to give you the right answer.

21:39We, just as some context, We're doing some some research in the in in advice given to older adults by generative AI overviews in Google, like asking it like pseudo medical questions or medical questions and then having doctors review the answers and saying, you know, is this advice that someone that someone should should take? And, you know, Google's gotten a little bit of scrutiny about this. Maybe a little is an understatement. Outside of topics maybe like medical advice, what's your level of trust with generative AI to give folks the right answer? I'm skeptical. It will always give you the right answer.

22:25I definitely think you need to have a human in the loop, as they say. Just to give you some examples of this, I mean, I think there's some pretty well-known examples, and I've had some examples from my own experience. There was a lawyer in New York who decided to use ChatGPT to write a legal brief, and he submitted it to the court, and then it turns out that ChatGPT just made up some cases that didn't exist, and a lot of false, you know, fake things. And he didn't even check it. And then he had to go in front of the judge and explain why he didn't do it. And, you know, it's not something that's a great advertisement for his professionalism, that he didn't even bother to check it.

23:05And that's, you know, that's an example. Another example is Air Canada offers a consumer a special discount on a ticket and then wants to go redeem it. And then Air Canada says, well, we didn't make this up. This was our chat bot that made it up. So we're not going to honor it. And then a tribunal in Canada forced them to actually pay for this. So basically, ChatGPT made something that was fake and the company was forced to honor it. Then my own experiences, just recently, I wanted to get ChatGPT to go through my book and extract the two or three most interesting stories. I think my theory is that people like to hear stories.

23:58And so I'm not very objective about what is the most interesting story in my book. So I asked it to come up with the most interesting story from my book. So it gave me a response that was totally made up. It was like there was nothing in the book about this at all. It was about somebody who kind of discovers some really clever way of painting and turns it into a massive business opportunity. And it's totally fantastical, totally, completely made up. So I said I responded and said, you know, this is not in the book. It's all made up. It's wrong. Now give me something that's really in the book. And it did exactly the same thing for a different topic.

24:36So complete, utter, made up nonsense to the question. And to be fair, I can't really defend it, but, you know, I thought that it would have access to the book and it would have been able to read through it and kind of and do that and kind of tell me what the most interesting was. But it just made something up. It didn't say, you know, it didn't say I can't answer it because I don't have access to your book. It just gave me this totally made up thing. If it if it had said, look, I can't get this information for you because I don't have the book, that would have been better. Yeah, the kind of hallucinations I think I've seen those colloquially referred to are pretty interesting.

25:20And it's not clear to me why ChatGPT or Gemini does that. Let's go through like a hypothetical. Let's say you're the CEO or a functional leader of a company. The company is a decade-year-old. So you have product market fit. You have a product or service that works. it's not a company about generative AI or an AI company itself. But you're thinking to yourself, how can I use this technology to make myself more efficient, to grow revenue, etc. How would you recommend a CEO approach that opportunity or that thought exercise? Well, I think to me, it's very much like a strategy project. And I teach, I'm the chair of the strategy course at Babson College where everybody in the undergraduate program has to take this course.

26:12There are plenty of other people who teach it, but I've been teaching it for a long time. And when I worked for Michael Porter, that's what I did. So I look at it as a strategy exercise. I think, you know, the first thing you need to do is to put together a cross-functional team. The CEO needs to put together a cross-functional team. It shouldn't be something that the CEO just sort of dreams up herself or himself, whatever. It has to be, you know, you have to sort of pull together the key functional executives that will be having to sort of execute on the strategy and contribute to brainstorming the strategy.

26:46And then I think what you need to do is identify the most important problem facing the company. I think because of the importance of this, I think you want to make sure you're putting an investment into something that actually solves the core problem. So one of of the things that I have noticed, I wrote a book called Scaling Your Startup. And one of the things that I observed is in the first stage of scaling, the most significant thing you have to do is identify the right problem and solve the right problem better than the competition. So this applies with using generative AI as well. You have to make sure you know what the right problem is, what is the most significant problem that your company is facing.

27:28And a lot of companies, I do a lot of case studies. The most common problem that companies face is their growth is slowing down. Perhaps their initial product, while successful, has matured and they don't have any new growth curve that they've invested in that will allow them to maintain their growth rate. So I look at strategy. A very important part of strategy is always investing in the next growth opportunity. So to me, there's a good chance that this problem that they're facing has to do with growth slowing down or potentially slowing down. There could be many other things, but just that's a pretty common one.

28:06And then you have to do what I call deep dive investigation. So the team should get together and sort of figure out, you know, what is the core problem and why is it happening? So really try to understand why it's happening. And it usually comes down to talking to customers, understanding their purchase criteria, understanding the things that they need and their needs change over time. So understand what their needs are, understand their purchase criteria for this category of products, and see how they perceive your company's product, how well it meets their purchase criteria, and try to figure out what kind of a new product or service would meet their purchase criteria better than the competing products do.

28:50Now, this is an area actually where ChatGPT can actually be quite helpful, is in brainstorming solutions to a company's problem. So this is, you know, I've used in my, for my students, I've used ChatGPT, done an example of, I asked it to create a service that would compete against Netflix and in 10 years, take over the leadership position in the online streaming industry. So when I did this, I found that one of the things that it was pretty good at was coming up with new ideas for a business. Not every idea, some of the ideas are pretty obvious, but some of them were like, hmm, that's kind of interesting.