Generative AI in Retail: Challenges and Opportunities for Retail CEOs with Peter Cohan

[00:00:07] You're listening to The Retail Razor Show, where your expert hosts and their guests cut through the clutter in retail and retail tech to shape the future of retail.

[00:00:19] Hello and welcome to season four, episode six of The Retail Razor Show. I'm your host, Ricardo Belmar.

[00:00:25] And I'm your co-host, Casey Golden.

[00:00:27] Welcome to Retail's favorite podcast, where we cut through the clutter and give you sharp insights on the retail industry and commerce technology.

[00:00:35] It's a show for retail executives, e-commerce specialists, store operation leaders, customer experience officers and everyone else in retail and retail tech alike.

[00:00:46] So last episode, we focused on one of my favorite topics, retail media, and looked at how measurement is transforming retail media networks for retailers.

[00:00:55] Plus, we had a great introduction to an amazing startup, Venity.

[00:00:58] This week, we are hitting the gas on AI and turning our attention to a topic I just know you love hearing about, Casey.

[00:01:05] I can't deny that AI is reshaping industries and has the potential to enhance our everyday experiences.

[00:01:15] I haven't had the best experience myself with the current applications, but that's why I'm really excited about this conversation specifically today.

[00:01:26] It is being used to drive progress, boost creativity, and I'm excited to see the impossible become possible.

[00:01:35] Right. So I think we know what we need to do.

[00:01:37] We need a guest on the show who not only gets AI, understands it because he's been part of the AI ecosystem since, honestly, it's very early days.

[00:01:46] But someone who's a proven management consultant and strategic advisor to many, many brands and startups and can deliver an academics viewpoint on the topic with a very sharp eye towards what type, what's real, what's really possible for retailers to accomplish here.

[00:02:01] How does that sound for a plan?

[00:02:03] Well, I mean, I'd have to say it sounds amazing, but I also already know who our guest is going to be for this episode.

[00:02:10] But what a way to build them up.

[00:02:12] And the suspense is the one, Ricardo.

[00:02:16] Ricardo.

[00:02:17] Yep.

[00:02:17] Well, so obviously, Gen.ai is one of our three themes for the season.

[00:02:20] So this may be our first independent view, I think, on the strategy behind how you should use Gen.ai and coming from someone who's actually guided many retail leaders in the most meaningful ways to get high value from investment in Gen.ai.

[00:02:34] Yeah, even with my dose of skepticism, I'm looking forward to diving into a deep strategy discussion that isn't afraid to talk about how this technology can help us versus hurt our business growth.

[00:02:49] So let's meet our guests for the week.

[00:02:51] Peter Cohen, a professor, strategy consultant, startup investor, book author and columnist.

[00:02:58] Yes, Peter is an associate professor of management practice at Babson College.

[00:03:02] He is the faculty lead for the core undergraduate strategy course and has created popular electives on startup cities and scaling strategy.

[00:03:11] His strategy consulting firm has completed over 150 projects to help companies identify, evaluate and profit from growth opportunities created by changing technology.

[00:03:22] He's invested in seven startups, three of which were sold for over $2 billion and one of which, SoFi, went public in 2021 at an $18 billion valuation.

[00:03:36] He's the author of 17 books, most recently Brain Rush, How to Invest and Compete in the Real World of Generative AI.

[00:03:44] And he's a senior contributor at Forbes and a contributor at Inc.

[00:03:50] Rethink Retail named him a top retail expert every year between 2001 and 2024.

[00:03:58] I have no doubt that 2025 is very near.

[00:04:03] And chose him as a top AI leader for 2024 this year.

[00:04:08] Yeah, an impressive background.

[00:04:09] And of course, we always enjoy having fellow top retail experts on the show.

[00:04:13] Absolutely. So let's get to the conversation and learn how to unlock generative AI the right way for retail.

[00:04:21] Let's do it.

[00:04:27] This is a discussion that I'm very excited about because as many of our listeners,

[00:04:33] I'm not the biggest cheerleader of everything AI, especially in retail.

[00:04:38] I have my concerns, even though I admit that there's a place for this technology in many retail businesses and several different capacities.

[00:04:45] But fortunately, this discussion is going to tackle most of my concerns head on.

[00:04:51] And I hope to shine a light on the right and wrong ways for AI to be used in retail.

[00:04:59] How C-suites should look at generative AI and traditional AI.

[00:05:02] We are privileged to have Peter Cohen with us today, associate professor of management practice at Babson College.

[00:05:10] And as we mentioned in the intro to the show, an impressive career in management and strategy consulting, startup investor, advisor, and author of 17 books.

[00:05:22] We'll get the latest one, Brain Rush, how to invest and compete in the real world of generative AI in just a moment.

[00:05:30] Peter, welcome to the show.

[00:05:32] Well, thank you.

[00:05:33] I'm delighted that you invited me.

[00:05:35] Peter, we just listed a number of impressive items in our intro from your background.

[00:05:40] Why don't you start just by giving us, in your own words, a brief tour of your background and how you became a top AI expert and leader and how that resulted in writing this latest book.

[00:05:52] Okay, great.

[00:05:52] Well, I'll start off by saying that when I was in college, I was a freshman in college.

[00:05:56] I told my father I wanted to be a poet.

[00:05:58] He said, look it up in the yellow pages.

[00:06:00] It wasn't there.

[00:06:00] So I decided to shift to something else.

[00:06:03] I won't bore you with all the other things that I considered.

[00:06:05] But by the time I was a junior, I thought, okay, I want to be in strategy consulting, helping companies to grow.

[00:06:11] And since I started learning how to program at a very young age, I thought maybe it would be interesting to be involved in strategy for companies that are in the computer industry.

[00:06:19] So after I graduated, I set off to graduate school at MIT and computer science.

[00:06:24] And I was looking for a job and I got a job at this consulting firm founded by five former Sloan School professors.

[00:06:31] And this was really a fantastic place.

[00:06:35] I loved it.

[00:06:36] I met all these great people.

[00:06:37] And one of them was a professor, a former professor who ended up starting an AI company.

[00:06:42] And so when I was in grad school at that time, I decided to go work in this startup.

[00:06:49] And to make a long story short, it failed.

[00:06:52] And that was my introduction to artificial intelligence.

[00:06:55] And I think the most important lesson I learned was that you can have super brilliant technologists with a very admirable piece of software technology.

[00:07:05] But if it doesn't create a sufficient value for the people you're trying to sell it to, it will fail.

[00:07:10] There's no amount of intellectual brilliance that will solve that problem.

[00:07:13] Unless you are offering a product that actually helps somebody with what I call a quantum value leap, namely a huge leap in value over existing products for the price that you're paying.

[00:07:25] People will say, well, this is a startup.

[00:07:26] It's probably going to fail.

[00:07:27] I don't want to get involved in this.

[00:07:29] Give me something that will solve a problem that is not being solved right now and do it better than anything else that's out there.

[00:07:35] Something that's really causing me pain.

[00:07:37] So that's a lesson I learned pretty early.

[00:07:39] And I'm going to kind of speed up a little bit here.

[00:07:41] During the dot-com era, which many of your listeners may know about, I wrote three books about that.

[00:07:47] And it's very interesting to contrast the current generative AI era with the dot-com era.

[00:07:51] And the key thing is the number 2,888, which is the number of companies that went public during the period from 1996 to 2001.

[00:08:00] And compared to zero, which is a number of generative AI startups that have gone public in the last couple of years.

[00:08:07] But I find that to be very, very interesting.

[00:08:09] And in thinking about that, I was just struck by how much of the generative AI movement has been controlled and dominated by essentially very, very large tech companies.

[00:08:20] But specifically, the idea of finding ways to build out and use cloud services.

[00:08:26] This is a tremendous growth engine for cloud services providers, which are very large companies.

[00:08:32] And yes, there are a lot of startups, but they are in some respects subsidiaries of these large companies that are pouring huge amounts of capital into them and driving up their valuations.

[00:08:41] Why haven't any of these companies gone public?

[00:08:43] I find it rather puzzling.

[00:08:45] And I find that there's a value at having all these startups that are driving the industry.

[00:08:49] And compared to the dot-com era, I think there were a lot of failed dot-com startups, but there's a lot more innovation because the startups did not have a vested interest in the way things are now.

[00:09:01] So it was, in some respects, more creative and more disruptive.

[00:09:05] And so now you've got these products that are going out there that are not necessarily delivering a huge amount of value, but there's a lot of large companies with money behind them.

[00:09:14] So having said that, this kind of is the frame that I brought to something that happened to me in May of 2023.

[00:09:20] I was looking at what happened with NVIDIA's earnings announcement, and it was sort of the shot heard around the world.

[00:09:26] Even though ChatGPT had been launched in November of 2022, it had been out for a couple of months.

[00:09:32] In the first two months, it got 100 million users, which is pretty impressive unless you compare it to the roughly 5 billion users that Facebook has and Instagram.

[00:09:40] It was still a lot of growth really fast.

[00:09:43] And then NVIDIA said, well, we had a decline in revenues in the first quarter, but our revenues are going to increase manyfold, two, three, fourfold in the next quarter of the next year.

[00:09:53] And all of a sudden, I just realized a bell is ringing.

[00:09:56] It's telling me that this is a fantastic opportunity, that this could be something that is significant, like the dot-com era was, the World Wide Web.

[00:10:06] And I'm going to write a book about this.

[00:10:08] So that is what launched me into a very intensive, year-long process, which began in July of last year and ended in July of this year with the publication of my book, Brain Rush, How to Invest and Compete in the Real World of Generative AI.

[00:10:23] Yeah, it's an amazing background.

[00:10:24] Thank you, Peter.

[00:10:25] So with that, let's jump in and start with a little bit of a level set on the impact of AI and Gen AI are having in retail.

[00:10:33] Peter, what would you say is how Generative AI is reshaping retail landscaping?

[00:10:38] Could you share some of the most interesting or perhaps groundbreaking applications that you've come across so far that you would hold up as positive examples?

[00:10:46] Well, yes.

[00:10:47] There's clearly companies are using Generative AI to personalize offers to individual consumers.

[00:10:54] And a company like Amazon, which is mapping out how does this particular user go through the entire online shopping process, the browsing, deciding what to buy, what to purchase, and kind of figuring out how to customize the recommendations for each individual for what they should buy and how that process is going to work.

[00:11:13] And with the idea that this actually is something that drives, on average, a roughly 2% to 3% increase in revenue for the companies that are providing this kind of a service.

[00:11:22] So this strikes me as an example of a service that will have what I think is the kind of impact that Generative AI has to have in order to be worth the investment that goes into Generative AI.

[00:11:34] This year, I saw an estimate that Sequoia Capital estimated about $150 billion would be spent on infrastructure for AI this year.

[00:11:44] And Goldman Sachs estimates over the next several years, that'll be a trillion dollars.

[00:11:48] If you look at it from the perspective of how can we generate a return on investment, namely the profit from all this is going to end up being at least equal to the amount that we invested.

[00:11:57] The question is, what will drive that return or what will drive the value?

[00:12:02] And in my opinion, the value will come from faster revenue growth.

[00:12:05] So this first application is an example of that.

[00:12:08] Another example, which is much more niche in my opinion, is companies like Ikea and also Wayfair using a sort of virtual reality and generative AI to take a picture of your room that you want to furnish.

[00:12:21] And then allowing you to imagine using, putting the furniture that they are selling into your room and seeing what it will look like.

[00:12:29] From what I understand for people who buy furniture, they are feeling anxious about whether the furniture will look right once it gets into their house.

[00:12:36] And they're kind of reluctant to pay for it, have it delivered, and then it gets there and it just doesn't quite work.

[00:12:41] So this is another thing that I find is a fairly powerful and valuable application of generative AI.

[00:12:48] And a third one that I think is also useful for retailers in general is the idea of 24-7 customer service.

[00:12:55] However, that one is very helpful if it gets at the right answer because you're not waiting online, you're not on hold.

[00:13:03] I can say that that is really helpful for people.

[00:13:05] People really want that.

[00:13:06] On the other hand, I have seen examples of it working and examples of it not working.

[00:13:12] And if it doesn't work, it's extremely bad for the company.

[00:13:15] I talked to the CEO of a software company in the Boston area, and he had two case studies that he experienced himself.

[00:13:23] One of them was that he had a problem with his wireless bill.

[00:13:27] He used the chat bot for the wireless carrier.

[00:13:30] Within three minutes, it had a copy of his contract with the company and figured out that he was right,

[00:13:36] and they made the correction to the bill within a few minutes.

[00:13:39] Which sounds like a very powerful, positive experience.

[00:13:43] At the same time, the same executive who had contact lenses, has contact lenses, got the wrong prescription.

[00:13:49] He tried to interact with the chat bot on his provider's site.

[00:13:55] Basically spent a lot of time, three different interactions.

[00:13:59] I don't know all the details.

[00:14:00] But it was so bad that he decided that he was not going to do business with that provider of contact lenses anymore.

[00:14:07] Totally destroyed his relationship with that company.

[00:14:11] So this is a very risky move to put customer service on that chat GPT or any kind of AI.

[00:14:20] If it goes wrong, it could be very damaging.

[00:14:23] And you never know whether it's going to do that or not.

[00:14:26] There's not enough solidity in the context of what's being produced to be sure that it's going to work all the time.

[00:14:33] Yeah, I definitely feel when it comes down to sales versus customer service, there's a distinction.

[00:14:39] But also just with physical goods versus maybe something service-oriented.

[00:14:44] I will say that service is something that if you look at the purpose of a business is to get and keep customers.

[00:14:50] The keeping of customers, service is a very important aspect of that.

[00:14:54] Keep customers you need to have good service.

[00:14:57] Whenever they want to use it, it has to be excellent.

[00:14:59] It has to make them really happy about continuing to do business with you.

[00:15:03] And so in that sense, it has something to do with perhaps not losing revenue and customer retention, which is a positive component of the whole growth picture.

[00:15:12] Yes, it's not the same as a new service.

[00:15:15] To me, all the things that I've just said are, how shall I say, relatively milquetoastish when it comes to what you might hope,

[00:15:23] which is that generative AI could allow you to create entirely new rapid growth curves.

[00:15:28] So I've been talking in my strategy classes at Babson this week about the idea of the product lifecycle.

[00:15:35] A very simple concept, but very important and powerful, which is that whatever product you have, even if it's doing really fast growth right now, it's eventually going to slow down.

[00:15:44] Look at NVIDIA.

[00:15:45] NVIDIA is the poster child for the entire generative AI industry.

[00:15:49] And last year, it was posting 200%, 300% revenue growth.

[00:15:55] For the current quarter, it's forecasting 80% revenue growth, which is phenomenal for any other company, but it's like a certain of what it was last year.

[00:16:04] And so even NVIDIA is under pressure, and perhaps even more so because the semiconductor industry changes so fast, to come up with new products that are going to drive new growth.

[00:16:13] And so you essentially have this S-curve, and you want to invest in those new growth engines before the existing S-curve kind of slows down.

[00:16:21] So to me, this is Casey's point is spot on that to some extent, these applications that I've just mentioned are not enough to create a very exciting new growth curve as in yet.

[00:16:34] So how do companies decide when to embrace AI, and what are the right voices that should be in these decisions, part of the decision process?

[00:16:46] Given that there's a mix of fear and FOMO and leadership, businesses have to be able to navigate the reality out of maybe 200 experience or just really starting with their first and feeling comfortable and getting between the fear and the FOMO.

[00:17:05] What would be your recommendations like specifically for brands and retailers?

[00:17:11] Yes. As we talked about earlier, not in this conversation, but in an earlier conversation, there are two approaches, theoretical approaches to how companies develop new products.

[00:17:21] And one of them I would call the relay race approach, and the other one I would call the rugby approach.

[00:17:27] And I think most companies are doing something more like the relay race approach.

[00:17:32] And so let me talk about what I mean when I use these terms.

[00:17:35] My relay race approach, and this is a concept I wrote about in my first book many years ago, was that in a traditional technology company, and I think this may be somewhat relevant to retail as well, is that you had an engineer that designed the product, and they developed a blueprint for that product.

[00:17:54] And then once they were done with their design, they would go to the head of the manufacturing division and say, okay, now build 100,000 of these or whatever the number is.

[00:18:02] Build a lot of these. And then the manufacturing people would do it, but they would give feedback to the engineers and say, this is expensive to manufacture.

[00:18:09] The materials don't work very well, and the quality is going to be kind of low, but this is what you asked me for, so I'm going to do it.

[00:18:16] And then they build it, and they have all these, whenever the widgets are that the engineer wanted to design, they're sitting on the loading dock.

[00:18:22] And the head of manufacturing calls in the head of sales and says, okay, now go sell these. We need some cash.

[00:18:28] And the salespeople looks at the thing that's on the loading dock ready to go out, and you know what? It's too expensive.

[00:18:34] It doesn't do what the customers want. It's not really that different than what the competitors are offering, and it's going to be very hard to sell.

[00:18:42] So that is the relay race approach to product development.

[00:18:46] The alternative and superior approach, in my opinion, is the rugby approach, which essentially is to answer Casey's question about who's in the room.

[00:18:55] The relay race approach, the way that most companies, I think, are using AI, and I think this is true in retail, because I was sitting in on a conference call with some executives from the retail industry a couple weeks ago,

[00:19:05] and it seems to me that what's happening is that the CEO is saying, basically, they don't necessarily brag about Davos in front of people, but they went to Davos.

[00:19:12] Everybody was talking about AI, so they're going to do something with AI, but they don't understand it, so they're going to say, I'm delegating this to my chief technology officer.

[00:19:21] Maybe I even went out and hired a chief AI officer, and they're going to figure it out.

[00:19:25] And the chief technology officer or the chief AI officer are probably really smart when it comes to technology and what's the right tools to use to get the problem solved.

[00:19:35] But what they lack is something the CEO has, which is a knowledge of how the company gets and keeps customers.

[00:19:42] How does it interact with customer segments, individual groups of customers?

[00:19:46] How does it identify unmet needs?

[00:19:48] How does it develop a prototype that actually meets those unmet needs better than competing products do?

[00:19:54] And all of that, all those things, that sort of cycle of understanding the unmet needs and building something that better meets their needs and delivers a quantum value leap,

[00:20:03] that is something that a chief AI officer, a chief technology officer cannot do.

[00:20:08] But in the sort of rugby approach, all the different functions are assembled together by the CEO.

[00:20:14] And I guess in a retail organization, if you're developing a product, you've got a designer, you've got people who are supply chain, you've got people who are doing marketing,

[00:20:23] you've got different people, depending on what particular retail industry it is.

[00:20:27] You get them all in the room and you say, we are going to find together early adopter customers who have problems, unmet needs that we can solve in a unique and creative way.

[00:20:39] And so we're, as a team, we're going to find those individual consumers or customers who need something that nobody else is satisfied they need for.

[00:20:48] And we're going to listen to them and we're going to develop a prototype and we're going to use generative AI as one aspect of the solution.

[00:20:56] And we're going to give that, those early adopters, the opportunity to use that technology and give us feedback on that prototype.

[00:21:03] Well, I like these two things, but these three things are not right.

[00:21:07] And you're missing these four other things.

[00:21:09] So try again.

[00:21:10] And so you go back and you do it again.

[00:21:11] And basically, after you've done that a few times, you get a product that actually really does a spectacular job of meeting a need that nobody else has met in the industry and does it better than competing products.

[00:21:25] And that is really what you need to do to make this technology effective and to generate new sources of revenue.

[00:21:33] It's a process of discovery.

[00:21:35] I do not think it involves looking at what other companies are doing and copying it.

[00:21:39] I think it involves you being willing to innovate.

[00:21:42] And to be fair, there are some executives that can do that.

[00:21:46] Often they are people who have started companies and grown them and maybe even taken the public and people who have actual founding and entrepreneurial experience.

[00:21:54] This is an entrepreneurial way of thinking.

[00:21:56] And there are certainly executives in the world who lack that experience and will struggle to do that.

[00:22:02] And so I think if you are in the retail industry and you are more like the relay race approach rather than the rugby approach, you are going to be vulnerable to people who have that true rugby approach to developing new products.

[00:22:17] And by the way, the term rugby, I guess, has something to do with a scrub.

[00:22:19] I've never played rugby, but apparently in a scrub, you've got everybody to come together.

[00:22:24] Right, right.

[00:22:25] It's a good analogy for a collaborative effort.

[00:22:27] I think one of my takeaways are what you just described.

[00:22:30] People in technology haven't necessarily gotten along very well under the roof.

[00:22:34] Also in retail, we talk about silos and siloed or departments and groups within the retail business.

[00:22:40] I think if I take one thing away from what you said, maybe two things.

[00:22:44] One is that you absolutely can't succeed if you're going to work within those silos in this approach.

[00:22:48] You need to have that true.

[00:22:49] That rugby approach really defines collaboration throughout the organization.

[00:22:53] And I think the other thing I take from that is from the CEO's perspective, I think you're really talking about accountability, right?

[00:22:59] The CEO has to maintain that leadership role.

[00:23:02] They can't just hand this off.

[00:23:04] To your point, right?

[00:23:05] Yeah, well, I mean, this gets to what I think of as the emotional side of it, which is, or maybe the psychological side of it, which kind of seems like a bipolarity.

[00:23:13] You've got these very powerful opposing forces.

[00:23:16] You've got the fear of missing out, and you've got the fear of AI going wrong and costing your company its reputation, not to mention legal bills.

[00:23:25] That is where the typical large company may be working on 200 different AI experiments, and it will only roll out maybe 10 within the company.

[00:23:34] So even they view it as risky to launch certain things in the company, but maybe 10 of them can be shaved into a position where it can be used internally, and maybe one or two at most outside to the customer.

[00:23:48] So there's this whole process of, because of this sort of competing forces acting out of the CEO, doesn't want to be fully accountable for doing anything that actually happens,

[00:24:00] because they want to be able to blame somebody that goes wrong to preserve their job, which is another reason why they're delegating responsibility for it to this person who's a technology person who, frankly, may not understand retailing, may not have a lot of retail experience,

[00:24:14] may not have any experience actually coming up with new products in the retail space that generate a lot of revenue, which I think is, to me, the primary requirement of a successful executive is the ability to create new growth curves.

[00:24:29] That, to me, is the way I look at why they should pay a CEO so much money is because they have the skills to look at everything that's going on in the industry and lead the industry to continue to sustain expectations beating growth.

[00:24:41] It's a lot of pressure.

[00:24:42] It's a hard job.

[00:24:43] And I think the reality is that if you can't do it, you don't really last that long unless you have extraordinary voting control over the company.

[00:24:51] Yeah.

[00:24:52] Yeah.

[00:24:53] So in this process, apart from owning that leadership role and defining and guiding these AI projects, if you're trying to stay within that rugby approach, as you described it, who are some of the other roles?

[00:25:05] Obviously, there'll be some technology people, but to your point, you have to include the people who understand the business, right?

[00:25:11] And those business leaders have to be part of this process.

[00:25:14] Or who would you typically say are, for this to succeed, who are the other roles that have to be part of that rugby approach?

[00:25:19] Well, I will say that one of the lessons that I've learned is that whoever you leave out of the process that could have any effect on the success of the product, from the concept of the product to the sales and the service of the product, all those people need to be involved.

[00:25:36] Because whoever you leave out, there's a risk that whoever you leave out of the process is going to intentionally or unintentionally sabotage the success.

[00:25:44] So you have to figure out, there's a judgment that has to go on in terms of how do you engage people and at what level do you engage them in the process?

[00:25:54] If you have every different function represented, it's going to really slow things down.

[00:25:58] But if you leave out important people, it's going to be a problem.

[00:26:03] And I'm sure it depends on the industry.

[00:26:05] It depends on what the basic product is for that company.

[00:26:08] But I would say pick an example of an industry or as part of the retail industry, and I will try to answer the question more specifically.

[00:26:16] So this is everybody, please go and take your teams out for a spin class or a jog through Central Park and get to know people who work in different departments and different buildings and in your headquarter town for sure.

[00:26:31] Yeah, and that's another interesting issue is that companies should not be so siloed.

[00:26:38] And it's really ultimately the responsibility of the CEO to encourage communication among people in different departments.

[00:26:44] Because this is a classic management problem of what's called suboptimization, which is just a technical term of saying that the marketing people think that the only thing that matters is the marketing issues.

[00:26:55] And the engineering people think that the only thing that matters is the engineering issues.

[00:26:58] And they all need to subsume their priorities to what is ultimately to what is going to create a competitively superior value for the customer.

[00:27:08] So I think the CEO needs to create a lot of coordination.

[00:27:12] This is, I wrote this book on how do you scale a company.

[00:27:15] And one of the key things you have to scale different functions, you have to coordinate different functions.

[00:27:20] So they are all acting in a coordinated fashion to make sure that what they're doing is going to lead to better, ultimately creating better performance for your products in the minds of the customer.

[00:27:31] So I think it's good to have people who can do their functional jobs well.

[00:27:36] But I agree with you, any kind of team building exercises you can do that reinforce the importance of doing things well for the customers.

[00:27:44] I did a six-month education program for Procter & Gamble.

[00:27:48] And Procter & Gamble is a very traditional company with category management and people who have different functional specializations.

[00:27:56] And getting them to work together on behalf of the customer is really the critical thing that has to happen.

[00:28:02] And so one of the things that we were doing was just getting them all to talk about strategy and understanding the customer and what technology you could use to solve their problems better.

[00:28:11] And having all those people in the room, everything, I remember we had purchasing, we had manufacture, we had advertising, marketing, sales, all the different functions were all in that room.

[00:28:21] All thinking about ways that they could work together to come up with news products that would solve unmet customer needs.

[00:28:28] So that kind of process is something I think the CEO should be leading.

[00:28:32] They should not just sit back and assume that all the different functions will know how to work well together.

[00:28:37] Because if you leave them alone, they will all be trying to put their functional interests ahead of those of the company's position in the marketplace.

[00:28:45] What are some of the best practices?

[00:28:48] I would say, you know, the best practices that I've seen are coming from founding founders, CEO founders.

[00:28:55] So I was really, really intrigued when, and this is not really a retail industry, but it's kind of retailish in the sense that this was a company that was a founder.

[00:29:06] I've talked to over the last 15 years, and he's grown the company to the point where it's probably going to hit a billion dollars in revenue in the next year or two.

[00:29:14] It's called, and basically the CEO is a company that offers technology for recruiters, for high-tech recruiters.

[00:29:22] So there's this whole industry of people who recruit engineers and software developers for the high-tech industry.

[00:29:28] I'm in Boston area, so there's a lot of these people there.

[00:29:31] And basically he is using AI right now in a way that is really, in my view, the best practice,

[00:29:37] which is he's looking at his customers who are these recruiting firms, who are pretty substantial privately held businesses.

[00:29:45] And he is figuring out ways of using AI to allow those people to grow faster.

[00:29:53] So he's thinking about how can we deploy AI in a way that will allow our most experienced,

[00:30:00] the expertise of our most experienced salespeople, our most experienced recruiters,

[00:30:05] to be available to that junior person that we just hired, who spent the last six months as a wait person at a restaurant,

[00:30:13] knows nothing about how to do this.

[00:30:16] And we will create a generative AI system that is tuned to our company's own salespeople

[00:30:22] and basically taps into the expertise of the best salespeople that we have and makes that available to a junior person.

[00:30:30] So all the minute, detailed decisions that they make, such as which specific resumes should we be sending to our clients

[00:30:39] so they can review them to see which one will fit their job,

[00:30:43] that the experienced person knows how to do that very quickly.

[00:30:46] The junior person gets confused and sends too many resumes, too many of which don't fit.

[00:30:52] That's just one example.

[00:30:53] So essentially what this does is it makes on the average level of sales productivity higher.

[00:30:59] So if you have a finite number of people and they can close a sale,

[00:31:03] they can basically match up a candidate to a company much more efficiently,

[00:31:08] that will inherently, if they just work the same number of hours, they're going to increase their revenue.

[00:31:13] So to me, that was very clever.

[00:31:15] You essentially looked at how can we re-engineer the process of matching a candidate to a company for their customers

[00:31:23] and doing that using a generative AI.

[00:31:26] So I would say that is a best practice.

[00:31:29] So one of my professors at MIT was one of the co-inventors of the concept of re-engineering,

[00:31:34] which some of your listeners may have heard of.

[00:31:36] Basically, it was in the 1990s.

[00:31:38] It was a huge management change.

[00:31:41] The companies were looking at their business processes from the perspective of customers

[00:31:46] and figuring out how to create more value and take out excess things that happened.

[00:31:51] And there are plenty of them that probably still do happen, which is individual departments duplicating the work of the department that was supposed to do it

[00:31:58] and basically adding 44 steps to a process that can be done at six steps.

[00:32:02] So that whole thing is something that's been around for decades.

[00:32:05] The concept has been around, but generative AI kind of allows you to perhaps solve some of those problems in a way that essentially can allow you to drive revenue.

[00:32:16] And I think perhaps the most important thing I would say is what the best, most important best practice is to stop flailing around and trying out generative AI.

[00:32:26] But rather than flail around and try out different experiments and see what works, which is directionless and basically will just lead to a lot of long conversations about why we can't do something.

[00:32:38] Focus your attention on understanding your customer's business or your customer's needs and creating a solution for those needs that delivers more value than the competing products out there.

[00:32:53] Using this technology as one part of the re-engineering process.

[00:32:58] To me, that is what the CEO should be driving the company to do and not just say, well, we can't.

[00:33:05] Here, look at all these hundreds.

[00:33:06] Let's try all these different things.

[00:33:07] Don't tell me about the process.

[00:33:08] You know, focus our attention on what's really going to pay on.

[00:33:11] Because one of the things that I learned when I started talking on CNBC many years ago was that I didn't know anything about what makes stocks go up and down.

[00:33:20] And they were asking me to pick stocks and I was determined to try to answer that question someday.

[00:33:26] And I finally think that I have the answer, which is, and it's not really an answer, it's a quasi answer, which is that a company has to grow faster than investors expect.

[00:33:37] That's the whole game in a nutshell.

[00:33:39] The reason it's a quasi answer is because how to do that is not as easy as saying those words.

[00:33:45] That's right.

[00:33:45] That's the trick.

[00:33:46] It's very hard to do, but if a company can do that consistently, it's surely going to create value.

[00:33:53] So that comes from driving revenue, not from cutting costs.

[00:33:57] So any kind of applications that cut costs, whoop-de-doo, I cut costs.

[00:34:00] That's not going to really necessarily drive revenue.

[00:34:02] In fact, focusing on cutting costs almost inevitably distracts you from innovation and driving revenue.

[00:34:09] I think that's important.

[00:34:11] Yeah, I agree with that one.

[00:34:11] I think we've often said, and I've probably had other folks on the show say similar things, like for when we see news reports of retailers whose business is not doing well, you always immediately hear their CEO and CFO start talking about cost cutting.

[00:34:25] And to your point where you can't cost cut your way to growth.

[00:34:28] You can't cost cut your way to success.

[00:34:30] That's a short-term band-aid in a sense, right?

[00:34:33] That might help you here and now, but it's not going to produce that long-term value that you're talking about and illustrating how to do it.

[00:34:39] And the other thing that I'm hearing here as well from what you've said too, that a lot of the hype that started around AI, especially Gen AI, is that it's just going to solve all your problems.

[00:34:49] But it's just one more tool, right?

[00:34:51] It's another technology tool, like every technology tool before it, that can be, if you apply it the right way in certain ways, it will help you solve challenges.

[00:35:00] And some challenges is probably better suited to solve than others, but it doesn't automatically mean it's the technology that's going to solve all of your challenges just by having it.

[00:35:08] You still have to find...

[00:35:09] Right, you can put another 200 problems on your plate.

[00:35:10] Exactly right, you can just create another set of problems along the way.

[00:35:13] If you don't have that direction.

[00:35:14] Yeah, and so listening to you describe that, so I've also seen you talk about a framework that you describe the value pyramid.

[00:35:22] And I just want to quickly describe the three stages that I've seen in that that you already outlined just a moment ago.

[00:35:27] But you start with this idea of overcoming a creator's block at the bottom of the pyramid, which I hear that as, that's where I need the right definition of what is the challenge I'm trying to solve.

[00:35:37] And how I'm going to figure out what's the right way to go about solving it versus just throwing a lot of spaghetti on the wall.

[00:35:42] And then as you move up, you're trying to boost productivity.

[00:35:45] You give a great example of the recruiting example you gave, which I thought was fantastic about how the AI solution there is actually augmenting what the people are already doing, but it's helping them do it faster and better.

[00:35:56] And then that is what leads to the top of your pyramid where you're enabling that revenue growth.

[00:36:01] Because you became more efficient, you became more productive.

[00:36:04] Now you're actually able to generate new revenue that way.

[00:36:06] Yes. And frankly, even that is within that second level of the pyramid of increasing productivity.

[00:36:13] It's basically, it does allow you to do your current approach, your current business model and get more revenue out of it, which is good.

[00:36:20] It will increase your revenue.

[00:36:22] But what will really, in my view, make the difference is that pinnacle part of the pyramid, which virtually nobody is doing right now, which is they are literally using the technology to invent new revenue streams that don't exist.

[00:36:33] And they're fast growing new revenue streams.

[00:36:36] To me, that is what companies should be focusing on trying to do.

[00:36:41] And productivity increases is, it can be, are good, but unless you can use this technology to create new business lines.

[00:36:50] Yesterday I was talking about Netflix and we all know about the DVD by mail business, but the ultimate vision was online streaming.

[00:36:57] Which was fundamentally, in some sense, the same business of people sitting in front of some screen watching a movie in their homes.

[00:37:07] But with online streaming, it becomes a completely different set of capabilities that you need to create a totally different source of new revenues, which fundamentally had disrupted the movie and television industry.

[00:37:20] And all these other established companies were trying to come in and try to replicate what Netflix had done.

[00:37:26] But Netflix had an amazing lead, which it's managed to sustain for a long time.

[00:37:31] So I guess what I'm saying is that kind of new business line is what I'm really saying is at the peak of this value pyramid, the pinnacle of the value pyramid, where there is an infinitely small number of companies even thinking about trying to do that.

[00:37:44] But that is where I think companies should be investing.

[00:37:48] I don't think that if you do overcoming writer's block, which is the bottom layer, everybody can do that.

[00:37:54] There's no competitive differentiation.

[00:37:56] Everybody can get a free subscription to ChatGPT or pay $20 a month and everybody can do the same stuff.

[00:38:03] No, there's nothing to keep anybody from copying it.

[00:38:05] There's no unique sustainability.

[00:38:07] If anything does work, everybody else is going to be able to do it.

[00:38:09] You will not get an end.

[00:38:10] Right.

[00:38:11] You will not gain market share.

[00:38:12] If you do the productivity increases, you may be able to use your internal expertise, like I was talking about with the headhunting, the tech recruiting firms, to have some kind of advantage over your competitors until they start doing the same thing.

[00:38:27] But ultimately, what I think will keep people ahead is being able to create these new lines of business where they will own temporary monopolies.

[00:38:36] It's like in the semiconductor industry, it used to be that every two years you had to develop a new, fundamentally new product that had twice as many chips on it, it cost half the amount.

[00:38:45] That kind of level of periodic innovation is the extreme.

[00:38:49] It's a level of innovation that you need.

[00:38:52] But I think retailers should be thinking about this as well.

[00:38:54] I always thought of certain aspects of the retail business fashion was you had to come up with something really new every season, I guess, in order to keep people buying.

[00:39:03] I think that that whole mentality is critically important.

[00:39:07] And, you know, if this retail is such a broad category, but certainly there's a lot of things that people can do to use this technology.

[00:39:15] But fundamentally, it's not an end in itself.

[00:39:18] And I've engaged with it.

[00:39:19] When I wrote my book, I did all try to all search of different ways to use it.

[00:39:23] And frankly, I found that it was at the point where it's like, I don't feel like I need to use this every day because there's so many things that don't work right.

[00:39:32] But then I just, it's like, yeah, not really helping me that much at this point.

[00:39:36] Let's talk a little bit more in depth on the technology because you have had some interesting points I've seen you make as well about the particular ingenerative eye.

[00:39:44] Starting with, of course, it's only as good as the data it's built on, right?

[00:39:46] So you're, it's a garbage in, garbage out kind of situation like with so many other technologies.

[00:39:51] So how do retailers ensure they're feeding their AI systems with quality data and avoiding that garbage in, garbage out dilemma?

[00:39:58] And specifically, I want to ask you about smaller, more focused language models.

[00:40:02] Small language models versus large language models that things like ChatGPT are being built on a large language model that's being fed by the whole internet.

[00:40:09] How do you advise retailers to leverage their internal data versus that collective internet data to produce the better solution that they need, which may not be a large language model?

[00:40:21] Yes.

[00:40:21] Yeah.

[00:40:21] Large language models are not going to be useful to business because first of all, let's just talk about the basic concept behind generative AI, which is that you have, you type in or talk a sentence in English or whatever your language is.

[00:40:38] And it will spit back a response in the same, whatever format you want.

[00:40:42] And it will sound cogent.

[00:40:45] But what all of that's going on here is that it's a, what's called a neural network.

[00:40:50] It's basically a distribution of analyzing what word is like most likely to come next in a sentence, giving the context of that word.

[00:40:59] So this is a key thing, this transformer technology that was developed by some Google scientists, just about all of whom left Google, by the way.

[00:41:07] But basically it's, it was a, is a model that says, if you see this word in this context, the next word is likely to be this.

[00:41:15] And basically there might be a probability distribution.

[00:41:17] There may be a couple of other words that could potentially be the next word, but it's going to tell you the, what is the word was the right, the highest probability.

[00:41:25] And it's, so it's basically guessing it's producing a guess and it's producing a guess that will be more accurate.

[00:41:33] If all the information that's used to train it is trusted information that you are controlling.

[00:41:38] If you are a company and you are trying to use this technology, it's going to have, it's going to be like polluted by all sorts of garbage.

[00:41:46] For instance, everything on the internet.

[00:41:47] In fact, what I've read is that they've run out of new data to train this.

[00:41:52] They have such a voluminous appetite for new training data that they now have to create data.

[00:41:57] So AI is creating data to train itself, which means that over time, the data will be useless and the models will be useless.

[00:42:06] So, you know, as the models get bigger and bigger, and as all this, people are paying all this money to build these models and buy all these NVIDIA chips, they're just getting worse and worse.

[00:42:15] So companies are saying, I want, instead of the large language model, I'm going to do the small language model.

[00:42:19] I'm going to just train it on data that I trust and keep it inside the company and just use it for the purposes I have inside the company, which will hopefully make the answers more likely to be accurate predictions because there's no garbage in there.

[00:42:34] A lot of guarantee.

[00:42:35] I can imagine it misfiring despite that.

[00:42:37] Second of all, you have the risk, on the risk side, you have the risk of employees letting proprietary company information out into the training of chat GPT, which basically is a legal and compliance risk.

[00:42:51] But if you keep it all inside in a small language model, in theory, it would be not only better compliance and lower legal risk, but also more likely to have an accurate prediction.

[00:43:02] Especially when it comes to using other software tools that are using chat GPT or LLMs, but when their customers or the brands or the retailers are using it with customer data and dealing with PII data privacy issues, things of that nature, just trying to keep that personal information inside of your organization.

[00:43:24] Yes, for an industry that has been hailed the land of Frankenstacks.

[00:43:31] We have spent many of discussions, many of years trying to take a more unified approach, not only to the business, but to how we manage data, how we store data, our supply chains, and trying to get those tech stacks to be more modern.

[00:43:48] I think we still have many companies that are still using AS400.

[00:43:53] And it's very sad to say that, but there is also still carbon paper that is often used.

[00:43:59] And so...

[00:44:00] Tech machines?

[00:44:01] Yeah.

[00:44:02] Yeah.

[00:44:03] Yep.

[00:44:03] Tech machines are like super regular.

[00:44:06] I, in fact, funny story, I made about $600,000 in my first couple of weeks at a brand I will not mention.

[00:44:16] Because all the customers were, I was calling asking for orders.

[00:44:21] They said that they already sent them in and they sent them in and they sent them in.

[00:44:26] And then finally, one of the buyers said, I faxed it.

[00:44:30] And I'm like, oh, whoa, whoa, whoa.

[00:44:32] We have a fax.

[00:44:34] Found a fax machine unplugged with no ink above the refrigerator in the kitchen.

[00:44:40] Plugged that baby in and I had $600,000 in orders.

[00:44:44] I was a hero.

[00:44:46] Wow.

[00:44:47] Wow.

[00:44:48] That is amazing.

[00:44:51] I'm like, oh, a fax.

[00:44:53] So cool.

[00:44:54] That is, yeah, yeah, that is amazing.

[00:44:57] So yeah, these basic things about getting your data so they can be used to train.

[00:45:02] That's a huge technical problem.

[00:45:05] And the big benefit here is that as much pain as some of these processes might be for brands,

[00:45:11] they're hundreds.

[00:45:12] It's very easy for them to remain hundreds of years old.

[00:45:17] They can hit 30 years, 50 years, 80 years, 125 years.

[00:45:22] They've done it.

[00:45:23] They don't have to put all this investment in to disappear.

[00:45:27] Like any amount of incremental improvement, it has this opportunity to have exponential benefits

[00:45:36] because the company is still going to be here for many, many, many years.

[00:45:40] I think it's just so imperative to really be looking at all of the ways that these companies

[00:45:47] are managing their data.

[00:45:48] And if you've been listening on 2X, turn it off and go back to the beginning of Ricardo's

[00:45:53] question and re-listen to that.

[00:45:56] Because I think it was just very much speaks to the dream of our business and the nature

[00:46:02] of our business and a lot of shifts that are happening right now as we start breaking

[00:46:05] down these silos.

[00:46:07] And I think it's very, very compelling proposition to help anybody who's listening that works at

[00:46:13] a brand to really bring those initiatives forward and talk about the difference between

[00:46:19] what kind of LLMs they're using or what kind of software tools or bringing it in-house

[00:46:24] and cleaning up their data structure.

[00:46:27] Clean data.

[00:46:28] Clean data.

[00:46:29] Looking ahead, Ricardo and I can go down a rabbit hole with you.

[00:46:34] We could be on the phone for, I mean, I'm down for the whole weekend retreat.

[00:46:38] I'm sure.

[00:46:41] But, you know, looking ahead, where do you see AI driving the most change in retail?

[00:46:48] And what should these businesses be preparing for the next wave of AI investments?

[00:46:53] I may have touched on it a little bit with data.

[00:46:56] A lot of the brands have a really difficult time knowing their customers.

[00:46:59] Everything is really seems that all money is driven into marketing, not necessarily on

[00:47:04] customer experience, not in building understanding customers or CRM data.

[00:47:10] It's really on marketing and top line funnel.

[00:47:14] Fill the top funnel for sales rather than going into the bottom of the well.

[00:47:19] This is my heartbeat of my core.

[00:47:21] So I don't know if I'd put AI necessarily in there yet because the risk is so high.

[00:47:27] Yes.

[00:47:28] But for everybody who's listening that works at a brand or a retailer, where do you think,

[00:47:34] maybe even the CEO, where do you think AI could really drive the most change in retail

[00:47:40] so they can just think about the ways that they would apply that and maybe sink their

[00:47:46] teeth into it?

[00:47:47] Maybe not, but maybe hopefully.

[00:47:48] One of the things that I did not mention is that AI can be used to go through all of the

[00:47:54] customer interactions between a retailer and the customer and try to understand, for example,

[00:48:00] if they have a record of all the texts and all the emails and all the chats, all the conversations

[00:48:06] between customer service people or salespeople and customers, all the customer, the reviews

[00:48:13] out there of the service, of the quality of the products, of the selection, all that feedback

[00:48:19] is out there, massive amounts in all these different structured and unstructured forms.

[00:48:23] And I think that AI can help forth.

[00:48:27] This is one of the things that AI is really good at.

[00:48:28] If you can get all this information into a chat, into a large language or a small language

[00:48:36] model, you can say, okay, now tell me what are the newest and most surprising changes that

[00:48:42] have taken place in our interaction between company and our customers in the last three,

[00:48:49] six months, 12 months?

[00:48:50] What are some emerging changing needs or things that people are concerned about that they might

[00:48:56] not have been concerned about six months ago?

[00:48:58] And try to extract some unique insights about where customers' needs are going to evolve in

[00:49:05] the future.

[00:49:06] And try to use it to build sort of new psychographic models about what customers are going to be

[00:49:11] wanting to buy and what kind of changes in customer service they're going to want.

[00:49:16] So I see this as the raw material for ideation about what new products or new services you

[00:49:21] could offer that would uniquely benefit your existing customers.

[00:49:26] Another thing you'll probably do is look at what's changing in the demographics of your

[00:49:30] customers.

[00:49:31] Are you still serving those same customer segments?

[00:49:34] Or are you seeing a growth in some product categories of new kinds of customers flowing

[00:49:40] into the organization?

[00:49:41] And why are they doing it?

[00:49:43] So, and maybe are we losing customers?

[00:49:45] Are there just a certain group of customers that we're losing?

[00:49:47] And why are we losing them?

[00:49:48] So essentially, all this data is there.

[00:49:52] It's in structured and unstructured formats.

[00:49:54] If it could get into a large language model or a small language model that's specifically

[00:49:58] for your company, you could use that to understand the broad changes that are going on that you

[00:50:06] may not be able to see because there's just, there's so much detail out there.

[00:50:09] You just can't get above it and figure out what the implications are.

[00:50:13] So to me, I would see that as pretty powerful.

[00:50:15] There's another thing that I wrote about in the book, which is agentic AI, which is essentially

[00:50:20] AI solving a series of problems for you rather than, for example, one of the examples is you

[00:50:28] say, I want to increase the number of high quality leads that are coming into our sales

[00:50:33] force by 20% in the next year.

[00:50:36] And the agent, the AI agent will figure out all the things that need to happen in order

[00:50:41] to deliver on that increase in high quality leads.

[00:50:44] That is the vision.

[00:50:45] And frankly, I see that as being one that would take a long time to implement, but it

[00:50:51] is definitely something that many people in the industry are talking about is these agents

[00:50:55] that will solve a more than just the first step of the problem, but just the entire chain

[00:51:00] of activities that need to happen in order to deliver on the goal that the management has

[00:51:05] set.

[00:51:05] That is another possible thing that could happen in the future.

[00:51:08] So, you know, maybe I want you to develop a new line of shampoo that will generate another

[00:51:15] billion dollars in revenue for us.

[00:51:17] Go figure out what it would look like and then design that, design it, package it, manufacture

[00:51:24] it, market it, you know, all that kind of stuff.

[00:51:26] So Ricardo, I'm going to throw us over here, but I'd like to go back to your example of

[00:51:31] reviews.

[00:51:32] I think that's something that is information that's completely unable to be digested at

[00:51:38] HQ, but so important.

[00:51:42] How and being able to back out maybe your influencers and paid people to get customers.

[00:51:49] However, I think this is where a lot of executives I talk to find a challenge.

[00:51:56] Whose responsibility and where do you feel the responsibility should be is, is it the review

[00:52:04] software's responsibility that they use to build this or should they build it in house?

[00:52:11] Because all of that information is, should be in an archive, should be used for future

[00:52:17] small language models within their organization in different departments, regardless if they

[00:52:24] use that review software company in six months or five years.

[00:52:29] This is kind of where a big challenge kind of is, is whose responsibility is it?

[00:52:34] Yes.

[00:52:34] Should they be doing it in house and using that as a first experiment?

[00:52:38] Yeah.

[00:52:38] Yeah.

[00:52:38] I would say that I don't really know who the providers are of the review software, but I

[00:52:43] would think that if I were a review software company, I would want to be working with companies

[00:52:50] that are willing to co-invent a service that would do what you just said.

[00:52:54] In other words, in every industry, there's probably a small number of companies that say,

[00:52:59] I want to be the best company and I'm willing to take some chances here.

[00:53:03] And I want to be the first to use the technology.

[00:53:05] So what I'm trying to say is it would be better if no one.

[00:53:10] It's very, nobody wants to be the first.

[00:53:12] Nobody wants to be the first.

[00:53:13] So basically that is, to me, if I'm a startup, I'm saying there's your opportunity.

[00:53:18] Very ripe opportunity.

[00:53:19] There's the opportunity.

[00:53:20] I'm going to, there are retail startups.

[00:53:22] A couple of years ago, H&M was a new thing.

[00:53:25] All these, Zara, they were basically taking a new approach.

[00:53:28] The people who have been in the industry probably know how state it is and how locked in its

[00:53:33] ways it is.

[00:53:33] So to me, I teach in a school that's known for entrepreneurship and I just look at people

[00:53:39] who are locked in.

[00:53:40] I even wrote, I wrote even, it's in my book about the distinction between cognitive hunger

[00:53:44] and cognitive locket.

[00:53:46] This is a distinction that I developed because I was reading an article about the fact that

[00:53:51] only 0.4% of founders of startups are running the company after two years after it goes public.

[00:53:58] And I, and the article was really great because it had that 0.4% number, but what was missing

[00:54:03] was why, what is it about these people that's different than the people who don't last?

[00:54:08] And so I said, geez, I know some of these people.

[00:54:11] In fact, I know about 30 of these people.

[00:54:13] So I started interviewing them and to make the long story short, the punchline is that

[00:54:17] these people have cognitive hunger and the people who are not surviving have cognitive

[00:54:23] lock-in.

[00:54:24] Cognitive lock-in being when I was starting off in my career, this was what worked.

[00:54:29] So they kind of get locked into this, whatever it is that they thought worked for them.

[00:54:35] They might not even have, it might not have been the actual right thing that worked, but

[00:54:38] they think it is.

[00:54:39] So they're locked into that and it's something called a confirmation bias.

[00:54:43] It's the only tool I have is a hammer.

[00:54:45] Then every problem is a nail.

[00:54:46] Every product can be solved with a hammer.

[00:54:46] Yeah.

[00:54:47] Yeah.

[00:54:47] Yeah.

[00:54:47] Right.

[00:54:48] Yeah.

[00:54:48] So basically any entrepreneur will say, okay, this is an opportunity.

[00:54:53] I can predict how these large companies are going to react if I do something different.

[00:54:58] Yeah.

[00:54:58] They will basically say, this is going to fail and I'm not going to worry about it.

[00:55:02] I'm not going to do anything about it.

[00:55:03] And that creates a huge opening for the entrepreneur to take advantage.

[00:55:09] So for example, in the fashion industry, the fancy designer would design the thing and go

[00:55:15] on the runway and they would say, now we're going to roll it out.

[00:55:19] And maybe the average person they're trying to sell it to doesn't want to buy that, doesn't

[00:55:24] afford it.

[00:55:24] So why not figure out starting in a completely different way?

[00:55:27] And there was a startup that I talked to that decided to do this, which is basically,

[00:55:32] let's figure out what the customer wants, design it for them.

[00:55:36] And then once they agree to buy it, I'll build it.

[00:55:39] I'll make it.

[00:55:40] And so instead of putting all this stuff on the shelf that the designers guess is what

[00:55:45] customers want, you actually give them something that they say they want and we're willing

[00:55:50] to pay for.

[00:55:50] And then totally rejigger the entire process of executing on that.

[00:55:55] So that is actually an opportunity for some startups to say, I understand how people are locked

[00:56:01] in to the way they do it now.

[00:56:03] And they aren't going to change.

[00:56:05] And I don't know how bad things are going to have to get for them to make them change,

[00:56:09] but that is what's going to happen.

[00:56:10] Awesome.

[00:56:12] I love that you offered like a third option, which is just building a co-building with your

[00:56:17] software vendors, maybe choosing a startup to pull into the conversation with you and

[00:56:23] your teams.

[00:56:24] That's a great third option there.

[00:56:27] Yeah.

[00:56:27] I mean, because most companies, I'm guessing most retailers do not consider themselves to

[00:56:31] be experts at data management.

[00:56:32] In some respect, it would be better to have somebody who actually does that for a living

[00:56:36] rather than trying to do it yourself.

[00:56:38] But you raise a very important point, which is if you can't get somebody else to do it

[00:56:42] right for you, then maybe you have to do it yourself.

[00:56:45] But it's going to be a hard sell internally, I would guess.

[00:56:48] Yeah.

[00:56:48] Yeah.

[00:56:48] Yeah.

[00:56:49] And it definitely requires having the right talent and the right way of thinking about it.

[00:56:52] I agree.

[00:56:53] I think most return organizations don't have that yet.

[00:56:55] I think the leaders in the field are probably the ones who are recognizing that they may not

[00:57:00] have that and they're working towards figuring out which of those options is the right

[00:57:03] one for them.

[00:57:03] Do they need to have it in-house?

[00:57:04] Do they need to work with the right partner to create that capability and then later bring

[00:57:08] it in-house?

[00:57:09] Or do they just need to work with a startup, work with an established company, but figure

[00:57:13] out exactly how to address it.

[00:57:14] I think those are the ones we see leading the market for retail right now.

[00:57:18] Well, this has been an incredibly illuminating discussion on AI.

[00:57:23] Thank you.

[00:57:24] I enjoyed it.

[00:57:24] I did as well.

[00:57:25] I think we all learned quite a lot.

[00:57:27] And as Casey said before, we went through a lot.

[00:57:29] So you may want to, if you were listening at 2XP like so many podcast listeners do,

[00:57:33] go back, turn it back down to 1X and listen again, because you probably missed a few really

[00:57:38] important nuggets.

[00:57:39] And in fact, if you didn't pick them all up then, I think you should all just run out and

[00:57:43] get a copy of Brain Rush right now.

[00:57:44] Because we'll highly recommend it to get all of Peter's insights on how you should be looking

[00:57:48] to attack the capabilities that AI could give you if you follow the right approach.

[00:57:53] And if you prepare for it and think long-term, I think is another way that one of my takeaways

[00:57:58] here is don't look at this as a short-term fix for everything.

[00:58:02] It's an investment.

[00:58:03] You're going to use it as a long-term capability that's going to help you solve challenges you have,

[00:58:08] but it's not inherently the solution itself.

[00:58:11] You've got to know how to do it.

[00:58:13] And maybe the one repeating theme we had throughout this episode is that it doesn't work in silos.

[00:58:18] The CEO has to take that accountability and you have to take that rugby approach.

[00:58:22] It's a collaborative effort.

[00:58:23] It's not that relay race approach as you described it.

[00:58:26] That's really going to put you on the path to success.

[00:58:29] Yes, Peter, thank you so much for joining us today.

[00:58:32] I maybe have turned the corner with the Gen AI.

[00:58:35] Really appreciated how you dug into some of these use cases.

[00:58:38] We'll have to see how I feel about it tomorrow, but baby steps.

[00:58:44] Okay, well, I'm glad you're thinking about it.

[00:58:47] That's really good.

[00:58:48] Yeah, thanks again, Peter.

[00:58:49] I'm looking forward to having you back on the show soon to keep this conversation going.

[00:58:52] Thank you.

[00:58:53] Have a great rest of the day.

[00:58:54] I'd say that's a wrap.

[00:59:02] If you enjoyed our show, please consider giving us a five-star rating and review on Apple Podcasts,

[00:59:08] Spotify, or GoodPods.

[00:59:09] Remember to smash that subscribe button in your favorite podcast player and tune in on YouTube

[00:59:14] so you don't miss a minute.

[00:59:16] A big thank you to our GoodPods listeners for helping us continue to stay in the top three

[00:59:21] spots of the indie management and marketing charts.

[00:59:25] I'm your co-host, Casey Golden.

[00:59:27] Please connect with us and share your feedback on LinkedIn, Threads, Instagram, Facebook,

[00:59:31] Twitter, and Blue Sky.

[00:59:32] And if you want to preview the best highlights from each episode right in your email inbox,

[00:59:37] subscribe to our Substack newsletter for full episode transcripts and bonus content.

[00:59:41] I'm your host, Ricardo Belmar.

[00:59:43] Thanks for joining us.

[00:59:45] Until next time, keep cutting through the clutter and stay sharp.

[00:59:49] This is the Retail Razor Show.