The future of digital markets

0:00This is Stanford Engineering's The Future of Everything, and I'm your host Russ Altman. I thought it would be good to revisit the original intent of this show. In 2017, when we started, we wanted to create a forum to dive into and discuss the motivations and the research that my colleagues do across the campus in science, technology, engineering, medicine, and other topics. Stanford University and all universities, for the most part, have a long history of doing important work that impacts the world. and it's a joy to share with you how this work is motivated by humans who are working hard to create a better future for everybody.

0:37In that spirit, I hope you will walk away from every episode with a deeper understanding of the work that's in progress here and that you'll share it with your friends, family, neighbors, co-workers as well. We came up with a way of standardizing products in a scalable way where now you can compare directly products that are similar or equivalent and that increased price competition quite a bit. So as I think at the revenue of how, you know, technology can, you know, really help a public market make more efficient and, you know, we reduce in that case, you know, the budget by 8%, like the spend was reduced by 8%, which amounted to like tens of millions of dollars per year.

1:25This is Stanford Engineering's The Future of Everything, and I'm your host, Russ Altman. If you're enjoying the podcast, please follow it in whatever app you're listening to right now. That'll ensure that you never miss the future of anything. Today, Gabriel Weintraub will tell us about how markets work and how they change when you make them digital and when you add in AI. It's the future of digital markets. In today's episode, we're going to continue our brand new feature, The Future in a Minute, where I ask the guests a few rapid fire questions and they respond with some rapid fire answers. Also, before we get started, remember to follow us in whatever app you're listening to right now.

2:03That'll ensure that we notify you about all the new episodes.

2:13What's a market? A market is a place where people buy and sell things. It's basic. There were markets in ancient times. There's markets now. The difference is you'd like your market to be great. You'd like it to be able to have everything you need, and you'd like to be able to pay prices that are reasonable. In the last 20 years, we've moved from physical markets to digital markets, and now AI is in the mix. This makes market creation and evaluation more interesting, more complicated, but potentially more impactful. Gabriel Weintraub is a professor at Stanford University in operations, information, and technology, and an expert on digital markets and the impacts of AI on them.

2:53I'm going to ask him, what is a digital market? What makes it good? And how are they evolving over time? Gabriel, what drew you to study digital platforms as a major focus of your research at the business school? Well, hi, Ross. Thanks first for having me. It's a long story that I'll make short. So I was always interested in markets and designing markets. At some point, I was designing markets to buy school lunches in Chile for the government, which was quite interesting. And then, you know, sort of without realizing it, markets went online and digital and built like large companies that scaled up.

3:37And it turns out that the questions I was thinking when deciding school lunches, you know, auctions, not too different to the questions is digital platforms were asking to run efficient markets. So somehow like, you know, it was like a natural transition in this digital world. So it all started with a kid's school lunch. I love that. So for those of us who don't think about markets all the time, what are the features of a good market or a market that, because you say you design markets, and I find that very intriguing. So what are the features that you would like to see in a market that is well designed?

4:16And is this debatable or is it generally agreed upon in your field, like what makes a good market and what makes a bad market? That's a great question. I think what may be debatable is what the objective, what are you optimizing for? so you could optimize revenues, you could optimize economic efficiency, social welfare, consumer surplus. And so I think the first step is decide what's your objective. And maybe in different circumstances, there's going to be different objectives. Once you fix the objective, I think there's some well-understood best practices. You want to have markets that have liquidity, that are not congested, that if you think about having supply and demand, that is easy for supply to find demand, it is easy for demand to find supply, that there's trust, that people trust the market.

5:09And maybe those are like sort of how level best practice is. And then there's like a lot of engineering and a lot of details in terms of the operations, how you run these markets to achieve those best practices that ultimately are supporting your objective, whatever that is. It's like revenues or consumer surplus. Got it. Got it. That's very helpful. And so when you moved from the school lunch kind of, and not just school lunch, but all of the kind of old fashioned markets where there was a lot of face to face or there was a non-electronic communication, what are the big challenges that come up when people started building?

5:47And, you know, it's all happened in my life. I'm probably older than you, but this is almost 100%, to my knowledge, a phenomenon of the last 20 or 30 years. What are the special challenges of online or digital platforms? Yeah. So I would say the challenge is also the opportunity, which is that the disability that you have as a market operator is something we've never seen before. So basically, in a digital market, you have access to pretty much everything, right? Like you see all the transactions, you see all the prices, you observe all the clicks, what is the interest of users. So you have all this incredible amount of information.

6:30That's the opportunity. And the challenge is, okay, how do you best use that information to run your market? And how to use it to optimize different aspects of the market? What do you share with users? What type of information? To what extent you share it? So I think the vast sort of it is really like the IT revolution that it's like all over the place in these marketplaces. And so another thing that I know you're thinking about a lot is the now emergence in the last four or five years of these AI, LLMs and other AI technologies. I'm guessing that this really changes some of your calculus, but maybe not.

7:11So how has AI impacted these digital markets, the way you design them and the way they run? Yeah, I think that's still an open question in terms of what's going to be the impact. We've seen some use cases where, for example, it's much easier to summarize information, like say products, you have all these reviews, and so you can summarize information, you can personalize information. but I would guess that there's going to be like much larger changes. A lot of people are thinking about like search, you know, if you replace, you know, these traditional ways of search with more like unstructured ways or even agents that know your preferences and go out and find goods and services for you.

7:54And I think this is all pretty nascent and people are trying different things. So I think we're going to see a lot of, you know, a lot of action in the next say like three to five years, trying to find those successful use cases where we're using like Gen.AI or LLMs to facilitate search. I think that has been to be a very important aspect in these marketplaces. So as I think about our discussion marketplaces, one of the things that we probably should discuss is for somebody, again, who's not a professional, there's a few marketplaces that many of us interact with all the time, like Amazon, right, is one that's obvious.

8:28In your world, I'm guessing that there's a deep, deep set of markets that you look at. Maybe some of them are not even kind of aware. They're not in the awareness of the consumer because they're perhaps business to business markets. So can you tell us a little bit about what the other major markets are that are kind of ripe for either redesign or rethinking or are being revolutionized whether they like it or not by these technologies? Yeah, that's a good question. So I mean, yeah, we all know about like like Amazon, Uber, or D &B. And probably, frankly, like a lot of the research has been driven or, you know, motivated by these companies or like Google, like advertising markets.

9:11I think like the later generation of marketplaces that we're seeing, it's one much more verticalized. So it's, you know, it's not something like Amazon where you can like buy everything. Exactly. Exactly. Like your show. But it's more specific. And also the market operator, the platform is playing a more significant role in maybe selecting the supply and being much more active in terms of what's being sold in the marketplace. So, for example, an example that I've seen is a marketplace for nurses. So it's like much more specific. And then there's a lot of screening done by the marketplace. So there's like a guarantee of like trust and safety.

10:05And in principle, you know, you can hire someone that's going to be a high quality nurse. So we're seeing like more of those like verticalized, maybe more also personalized experiences rather than this more like generic marketplace. Yes. And as you describe these, it strikes me that there's going to be, I would guess that there are sometimes losers. As the market gets rejiggered, people who used to take advantage of position in the market or control of information, they might lose information. I'm thinking about, you know, as you know, I'm sure you know very well, the real estate industry is in the middle of this big churn because, you know, the 6 % kind of we paid it automatic for my whole adult life.

10:51You bought a house. There's going to be 6 % off the top for the agents, kind of whether you like it or not. And there's now a suggestion that some of these numbers might change. Does that fall into your work? Like, do you look at the way in which kind of power balance switches and that there's kind of some winners and losers as a new market emerges? Yeah, absolutely. So, you know, if you think about what is the role of marketplaces, what the value they provide is really reducing frictions, like reducing transaction costs. And, you know, if you think about something like, let's go to a traditional marketplace like Airbnb, right?

11:24Like, how would I know that there's like an empty room in the middle of Rome that would be like helpful for my next trip, right? So Airbnb is reducing that search and information friction. And that is basically what's doing is also reducing intermediaries, like some of the things you were mentioning, like people that maybe get a fee for finding you a place, finding you a product, facilitating a transaction. And now you have this platform that is kind of replacing all those intermediaries and facilitating the transaction. So maybe old fashioned, maybe older fashioned ways of, you know, for example, in travel, that's, you know, all these like agencies are the big losers and things like Booking.com or BNB.com, etc.

12:08are the big winners. And now sort of the platform becomes the intermediator, right? And of course, it's a cut, right? It gets a 15%, 20 % cut. So I think kind of technology is replacing these older ways of intermediating, which they also, of course, get a fee. But, you know, one would think, one would hope that because of technology, these are more efficient, more like faster ways of, you know, making demand meet supply. Yes. So I know that beyond looking at business markets, you've also looked at government and public sector issues, and especially in your home nation of Chile. So tell me about the opportunities for like these markets or like digital platforms in the setting of government or kind of not purely economic business transactions.

13:04Yeah. So we've done a fair amount of work with procurement agencies in Chile. So these agencies buy, purchase something like 3 % to 5 % of GDP every year. Wow. That's the amount of transactions. So any efficiency that you can get goes directly back to taxpayers. And this can be a fair amount of money. And this, without overgeneralizing, typically these markets are not as sophisticated as these tech platforms that we've been talking about. So there's a huge opportunity to introduce this type of tools like digital intermediation, now AI, to make these markets more efficient. So just to give you like a very concrete example, we were, you know, working with this type of markets in Chile.

13:59And basically the products were all described in natural language. And there was this like huge like Excel spreadsheets. And the thing, you know, was so unstructured that you couldn't compare two products that were identical. The system didn't know that this is a Diet Coke and this is another Diet Coke because they were like described slightly differently. So there was no price competition. And so basically using NLT and now more recently LLMs, we came up with a way of standardizing products in a scalable way where now you can compare directly products that are similar or equivalent and that increased price competition quite a bit.

14:41So that's, I think, a direct way of how technology can really help a public market make more efficient. And we reduced, in that case, the budget by 8%. The spend was reduced by 8%, which amounted to tens of millions of dollars per year. So from your experience with Chile and other governments, are governments ready to do this? So I'm imagining that there's a technology kind of comfort that you have to have if you're moving from your old fashion of getting bids. And we all I'm from New York. We all know that bids can be very complicated in the old days and lots of other factors. But are these governments ready?

15:23And what do you or others have to do to get them ready so that they can move to this kind of obviously more efficient way of doing business? Yeah. So I think there's like two challenges. One is that sometimes the technology is so obsolete that it needs like a pretty and so inflexible that you need like a pretty big overhaul just to make basic things work. I think that, but that's, I think, you know, that's an obstacle that I think you can jump over. But I think what's really key is to have the right leadership. So in every single case that we've been successful is because the leadership, like the very top leadership was super supportive, like really understood what was the opportunity of introducing technology and making these markets more efficient.

16:10And so we had a champion that had power and that was like a key partner to make this successful. So I think whether, probably ask more like which agency is ready and the agencies that I would say are ready are the ones that have leadership that is aware and has, you know, they're very conscious about the potential of these tools and approaches. So this 8 % savings is very impressive in the Chilean example. As you look at local, state, and federal governments in the United States, are there similar opportunities? Where is the United States with respect to creating such markets? Yeah, I think I haven't worked directly in those markets.

16:55So this is maybe a bit more superficial and second-hand knowledge. But my sense is that there are similar opportunities. If you go, there's inefficiencies in many pockets of government, local governments, federal government. And, you know, I have colleagues that have, you know, done work designing these markets to make them more efficient in different dimensions of government. And I think my experience is similar. Like you need the right leadership. It's just like the bureaucracy is just so big. And you need to get passed through that bureaucracy that without the right leadership and the right partner is just like very hard.

17:37But once you have it, there's like amazing things. Yeah, the picture that you painted that the mayor of the city or the governor of the state, they have to be the champion because they need to be able to clear the way for people who otherwise might be obstructive for kind of irrelevant or random reasons. Yeah, or even the director of an agent in our case has been the director of the National Procurement Agency or the director of the program of the school lunches that we were discussing at the beginning. Those were our big champions. This is The Future of Everything with Russ Altman. We'll have more with Gabriel Weintraub next.

18:23Welcome back to The Future of Everything. I'm Russ Altman, and I'm speaking with Gabriel Weintraub from Stanford University. In the last segment, we learned about what a market is, what a digital market is, and what makes a good market. In this section, we're going to talk about how governments can benefit from some marketing principles and also from generative AI. And we'll talk about what it means to have an AI strategy, either for government or for business, and how you should think about that. Don't forget, at the end of the interview, we're going to have the new Future in a Minute segment where I'm going to ask Gabriel some quick questions and he's going to provide some quick answers.

18:59But Gabriel, I know you also work with government. We touched upon government a little bit, still in the market arena, but you've worked on governments in a more broad way to think about AI, generative AI and the challenges. So especially in Chile. So tell me about some of the projects and what are the goals there? Yes, of course. Yeah. So I think this is like a huge opportunity in terms of making governments more efficient using Gen AI. And I think there's like a very good match with what Gen.AI can do is very good at, which is, you know, processing information, summarizing information, getting insights from information and a lot of the bureaucracy that happens in governments, which is if you think about regulations, it's like incredible amount of like paperwork and processing this, all this document.

19:45So there's like two projects we've been working on. One is to take a lot of, you know, most of the regulations involved in Chile for all the mining and energy projects, two of the key areas for economic development in Chile. And these are, you know, thousands of pages of documents. Each one of like for each project, there's this like obligations that they need to put in place to run the project. And we've used Gen.AI to convert all that unstructured information on actual data. So really quantify what is the number of obligations, how much they've been growing, what is the type of obligation. So this is the first time we get this kind of snapshot that is quantified, that is precise, that is rigorous of what exactly are we asking from firms and what the associated cost may be.

20:36there's this sense, there's this intuition that in Chile, like all these obligations have gone to the roof and they're like, you know, to the point that they're really, you know, making investment very slow and disincentivizing investment. And this is really like evidence. We're providing like, you know, strong evidence, quantitative evidence about this. So this is one project. So just so I can understand, are you analyzing the text generated by the government or the text generated by the projects that are being proposed or both? So at the end of the day, it's the text that was generated by the government.

21:14That's a result of the interaction between the company implementing the project and the government. And so then you can see, and it kind of cuts through all of the kind of fog, the fog of text and gives you a quantifiable, here's the project's deliverables, here's how much at costs and that's working. Is that deployed or is it a demonstration? We are launching this actually in two weeks from now. We've been a big event. Yeah. Whoa. Okay. So there'll be a trip to Santiago is my guess. I interrupted you, but what's the second project? And the second project is we're working with local municipalities that give construction permits for everything that you build.

22:01you need a construction permit for local municipalities. And these things would take like months. And sometimes it's like very inefficient and making, sometimes it's so long and makes a project that was profitable to unprofitable. So we are introducing Gen AI tools to speed up the whole process. What's interesting is we're introducing a bunch of tools for the project applicant and similar tools for the reviewers and the idea that they talk to each other to make the whole process more efficient. And we are going to launch this. We're launching a product like next week. And then we want to run a randomized control trial to estimate the effect.

22:42And hopefully that effect was like positive. We are starting with two pilots, two municipalities pilots, and we want to launch it to like tens of municipalities. So yeah, we'll see how that goes. Okay. So again, these are going to be bellwether projects that if they're successful, you can imagine, I can imagine getting a lot of attention and a lot of press really globally. Anybody who's done construction doesn't have to hear twice to know the potential value of anything that would speed up a, even a kitchen remodel is like a nightmare. So yeah, that's right. Yeah. So I think we're doing this in Chile, but I think like the use scale is very general.

23:20Yeah. Yeah. Yeah. Yeah. So I, so in both the examples that you've You've given, you talked about the markets and the businesses and you talked about the government. And it raises this issue that like the businesses and the governments, they might be thinking of this as like our AI strategy. Like, OK, now AI exists. How are we going to use it to make our world and our business or our government better, more efficient? And I know you've written about this and you've talked about this AI strategy. I'm guessing you get a lot of incoming requests to help with AI strategy. And how do you take a business, either big or small or medium-sized, and help them figure out what the impact of AI or the opportunity is for them?

24:00Yeah. So I think a key idea that it sounds like obvious that people in the middle of all this tech hype forget is that you should fall in love with the problem and not with the tech. So I think a typical dynamic is like the board says, oh, we need to do something in AI, but they don't know what. And then people start just implementing things that turn out to be useless. And I think the problem is that they started with the tech, not with the problem. So I think a basic pillar of any strategy is just basic strategy, which is start, what's the problem you're solving? How are you creating value? How are you capturing value?

24:37And then really understand what the technology is good at and make a match between the technology and your problem. And it may be that for some problems you don't need AI. For some, AI is a perfect fit. but really start from the problem like the value creation, the value capturing proposition that a business has. Yeah, I was struck because in reading some of your articles preparing for the interview, you basically made the point that it's not really an AI strategy. You should have a business strategy and you can see where AI helps that strategy. But the strategy is not focused on the AI. It's focused on whatever is like creating friction in your business.

25:14Right. I think AI is a technology. I think maybe it's different because it would really like impact like maybe every aspect of your business. But I think you should start from like, what is your business strategy? Yeah. Now, I also get asked this question every now and then. And one of the things I wanted to ask you is the degree to which there's a barrier because of the workforce. Right. So they can say, OK, we're going to do we're going to listen to you, Gabriel. We're going to think about our business strategy. Then we'll think about AI. Even if they do it the right way, what do you tell them about the workforce?

25:46Is it your sense that we have enough AI savvy people out there to help individual businesses? Or is this still a barrier for entry for the businesses? Just getting the talent to kind of use the LLM or install whatever needs to be installed. What are you seeing in these governmental collaborations or even in the business collaborations? I think people are getting more savvy with these tools. But I think, you know, depending where you go, what type of company, if it's maybe more like not a tech company, this is still a challenge. Yeah. So I think sort of training workers to be able to, you know, add value using these tools is something that is, yeah, that I think it's quite, you know, quite important.

26:34In another setting, we've given advice that you need to enable your workforce to play with these LLMs and not have it be like and be open about it. Tell your supervisor, I'm trying out LLMs. I'm doing some experiments to see if it can help us do X or Y. Because the only way you figure this out is by is by doing experimentation. And yes, you can bring in consultants. But really, if the people who are doing solving the problems day to day start to do these little experiments for themselves, they start to see what might be possible. Yeah, totally. know i think like that's the best advice is just learn by doing and i also think we're like it's a huge opportunity if you if you think about like small and medium medium businesses that and maybe like owners that are not very tech savvy uh i think it's like a huge opportunity because you know there's like amazing things that you can do and it's with natural language right you know you don't need to be a coder or a sophisticated technical person so so like train also like getting like training programs, going for like those type of business owners, like smaller, medium enterprises, I think that could also be very powerful.

27:36But it sounds just as my final question, it sounds like in all of these collaborations you've done in Chile, you have found the workforce to implement. Yeah, I mean, we found the leadership and yeah, we found the workforce and some of it has been a lot of like training like along the way. And I guess maybe one word of caution is there's also like a bit of a fear, right? Once you're implementing the systems, you know, there's the fear that, well, if the system works, I'll be replaced as a worker. So that's something that also we need to grapple with. We need to, you know, really be aware and, you know, hopefully design systems that work in tandem with humans to add value in whatever setting we're operating.