Les startups d’IA no tenen moat? Salvador Mas sobre SaaS, banca i avantatges defensables

Let's talk about professional future, career, development, business strategy, one of the interviews that I was most looking forward to bringing to Montocástico. We have a great guest. Well, we finally have you here, Salvador Más, interviewed on Monos Estocásticos. And well, it's an interview that I was really looking forward to doing because I've been following you for years: your career, your activity, what you've shared, the interviews you give...Well, for those who don't quite know who you are, well, Inverta, Invertia, sorry, Open Finance, Fin Matrix and now GPT Advisor; in other words, a person who, well, has basically been in the, let's say, in the engine room in this whole part of banking. Asset management and especially the technological implementation in that entire industry. And well, now let's say that you had already made the previous technological leaps, right? The point like, uh, digitalization, digital transformation of banking, well now you're also riding the other wave which is the integration of artificial intelligence. Well, I wanted to bring you in as an expert on Leonard Coin, which is basically what I was most interested in, as a big movie buff, but well, we're going to have to focus a little on the topic of the podcast and well, if you want, we can start at the end. Tell me a little bit, Salvador, tell us about this whole part of GPT Advisor, which is precisely a venture in the wave of AI, finance and technology and how you've approached it and what the idea behind the project is . Yes, the truth is that I was in the industry. I was working at All Fans, which is the leading B2B fund distribution company in the world, of investment funds in the world, and I was the Global Head of Digital; but I was a little bored, as this often happens in some corporate stages, right ? And that coincided with the fact that when we saw the GPT chat in November 2022, we said, "Well, here's a company, right? Here's a startup." and it just so happened that it was a time when I was almost leaving a corporate business , and that window just happened to come up, and I immediately jumped in and set up GPT Advisor, which we gave this opportunistic name that has also been widely reported by Open AI. Yes, yes, yes. And so we said, well, how is generative technology going to affect everything I've been doing for almost, I wouldn't say 25 years, but almost 30, right? Which is doing software projects or digital projects for savings and investment. So, it was a very good moment, a very good opportunity. By April 2023, we already had the company set up, the partners, and so on. Very good, everything went great. And we immediately had clients. There are a lot of clients who come because of FOMO because many entities have a lot of innovation departments that had to do things with AI and things like that. Well, it's always good to be in something that's trendy, right? Sometimes it's bad, sometimes it's a bit too bad because you do things simply to satisfy that hype, right? Without much depth. The check of I have it, the check of I have a website, right? Exactly, exactly. And that's it, the truth is that we are doing very relevant projects, we have projects in production with many financial entities, we are B2B and we have projects, I don't know, with Santander, Anbank, Bank Inter, Banco Sabadei. Well, we are setting up a bit,

eh, I think the first phase of the two that exist here in AI in a company like mine, the first was like creating copilots or creating private GPT chats for the entities, because I think the biggest challenge that existed in the early phases of AI and still exists is to create systems that are not, they don't have to be very wow . The wow part already comes with the AI part. It is that they are reliable, right? I mean, the concept of hallucination is something that some bank executives can't understand. For whatever reason. Exactly. No, what do you mean it got wrong? What do you mean it got the wrong amount of money the customer has in the account? And then someone comes along and says, "Well, with a 99.1% probability, we have reached a 99.1% probability that it gets it right." No, no, no, that is intolerable, it has to be 100%, right? So, in the end it was about how to make the AI use the LLM part, the linguistics part, so that it understands things and makes queries and then makes the queries where it has to make them and, again, use the LLM to build the answer. But in reality, the LLM has to be kept very short and takes away a lot of its play. And that was the first part, and it continues to be, right?,of our work the first year, the first two years, was practically all about making projects that weren't mind-blowing. With that you could already have a GO. Without that, it was impossible. Well, we've done a lot of projects there and there are many users, for example, I don't know, there are banks whose bankers or advisors are constantly asking GPT Advisor, "Hey, which fund is best for this client?" Or, "Explain this client's portfolio to me so I can explain it to them, etc." We're also working, which is also very relevant, with the regulator, both with the Spanish regulator and with the English regulator, with the CNMV. In fact, next week I have a meeting with the Bank of Spain and with the Minister of Economy, uh, or rather, Corpo is the vice president, right? Carlos Corpo, Carlos Corpo, uh, to explain what things are being done and all the good that this is bringing and all the dangers that it has. And all the dangers that it entails. It's a lot of fun. I'm having a lot of fun. Then I think there's another phase, there's another disruptive phase that you also put very well in your writing, which was that you said that this has stopped being a stochastic parrot, right? It's, uh, I think it was last summer or September, December, January, and that's it, and that's where there was a disruptive change in my company. And now, in the company, we're actually becoming not just someone who provides agents who, who offers agents, agents to search for products, agents for such and such and who integrates them, but also the trendy word now, which is the "forward developer engineer" and such. In the end, uh, what happens is that since the cost of making programs recently, recently, that is since January, I would say, uh, is so low, now it's like you can do anything, right? Anything you know how to do, anything you know. And there and there we've done interesting projects. For example, the latest management company that has been launched in Spain , which is a management company called Silverway, which is owned by a guy who wants to, who comes from Miami and who

is setting up a management company in Spain, in Luxembourg. I met the head of operations for coffee and he said to me, "Hey, what software should I buy for the back office? What software should I buy for the middle office? What CRM? Sales Force or HSPOT or any of the other questions that an operations manager has to ask when he has to buy a lot of software." And so, while having coffee, since we were already here messing around with Cloud, I told him, "Hey, man, don't buy anything, let's give it a week, let's try to do everything from scratch." And he said, "No, no, man, that's impossible." And I say, I don't know if I say, it seems possible, but it scares me. I mean, that's why, I mean, eh, let's try to see if it's possible. And in a week, before a week, we already knew for sure, Antonio, that everything swept everything, it swept all the software. I mean, with Cloud, someone who knows what they have to do, like this guy, this guy who is Álvaro, who is a guy who knows perfectly well the functionality that it has. I mean, what is very important is knowing how to be the boss, being a good product manager. That doesn't take that away from you. CL for now, for now, for now, But from there everything, man, everything...I thought: wow, in my previous companies this would have taken 6 months, 15 development guys, the consultant, the other guy, the so-and-so...Well, with three very good junior guys (because that's another one that we can also talk about, these junior profiles, who are also being blocking entry into companies with the excuse of AI, when they are the salvation for productivity problems), they are not going to come only from the, but from this. We did a project, of course, a marvel. So, the first breakthrough was Chat GPT from our point of view, in November 2022, which is when we set up the company and now we have a partnership with Anthropic, we are talking all the time with Anthropic, we are there developing anything. is that you have brought up 14 topics for, not like 14 threads that could be pulled, saves, which seem super inspiring to me. Look, there is one that I think will be of great interest to the business maker part of the audience, which is basically, uh, in addition, you have explained a little about the strategic movement of the company, right?,from a company that is a bit of a SAS product, suddenly, to a kind of technology consultancy that puts consultants in the client. I don't know if your vision is of a platform, that is to say, uh, in what position or in what way of working Do you think it will make sense with how the AI power of the big labs is growing and how it will grow, let's say, which I think is what conditions it the most? But, at the same time, what is the set of the platform-application model? Because right now I have very little clarity on , hey, is the functional part going to be MCPs or is it going to be creating skills or is it going to be, uh, creating specialized models? That seems less and less clear to me, uh, or is it going to be creating specialized jars or where is the path both at a strategic level and almost technologically for a station, right? Which in the end is what GPT Visor is also. Well, I don't know. I mean, I navigate with all those doubts and I navigate with that. I'm an entrepreneur that Peter Ciel wouldn't like at all,

Peter Ciel says, "You only have to decide on one thing, right? I don't know and I'm going to and you're learning as you go because there's no other choice. Yes, I think there will be something for everyone. For example, SAS, this project I'm telling you about , to set up a management company from scratch with all the operations, integration with brokers, all of this that we've done in two months. This is also because the head of operations there is very smart and he's a guy; I mean, I think there's a part of the industry, the corporate industry, that will continue buying SAS because, in the end, buying SAS for them is like not worrying, because they don't know; they tell you how many executives work, right? CRM. So if you were to work with Cloud, you would say to them," Okay, no, but which one do you want? Sales Force or Habsport or I don't know what. "That's the typical question. So if I say," Well, what CRM functionality do you want? What do you want? Sending an email to customers when it's their birthday? "Remember onboardings, many managers don't know about all that kind of small functionality because their way of working is buying a CRM. They're going to keep buying a CRM, they're going to keep buying sales force and such because what you're going to need is no longer that manager who outsources. Uh, what happens? Until the managers change, those will also continue to exist and maybe they'll never change, right? Because I've been in this job for many years now and managers are more or less the same. So, maybe they never change and there's a way to SAS. So, in the end, what you have to do, for example, what we can do is a SAS for management companies, imagine that case, to have the entire back office and mid office of a management company. And to another client who is not so specific, I can sell that packaged. like SAS, because in the end SAS also provides a good service. I mean, SAS also makes sense. So, I think there are going to be many different markets. What's new is not that SAS will disappear, because I probably don't think it will disappear completely. What's new is that it can be done from scratch, that you have the option to do it. That's the new factor, but that doesn't mean that it will always be done from scratch. So, I have my doubts because I have my doubts because, of course, I also have my own software, as you say, and my own agents , but sometimes there are agents who, when an entity comes to us, we are always very honest with ourselves, we say," Hey, okay? Is it worth maintaining our little legacy and charging a license or is it worth starting from scratch? " Because if you cheat yourself there, that's when you're going to end up screwing up. In other words, you have to be thinking at all times whether it's worth going from scratch or going with our own license. How about that? I think that those who insist on it, and I see this with all SAS companies, especially the big ones, will obviously insist on continuing to charge a license fee and continue with their model. And in a way, since it's not going to die because it's going to have a lot of clients who are still captive and even new clients, I don't think it's going to die, so maybe they'll go that way too. But my philosophy is that I don't know what's going to happen and therefore I have to be on top of

everything and move on. It's very complicated to think when you start developing, it's very complicated to think that it's not going to be all cloud and forward developers and engineers, right? That's what you think too. What is important is the craft, that these forward developers are verticalized, specialized, and know about a sector. I do think that's important, but it's very complicated to think that software development isn't going to be all that. But hey, there are companies that believe in it, right? For example, Sierra, companies that are so big that they go with their agents or Palantir itself, the Palantir model itself. Uh, of course, when someone explains it to me, I can't even talk about Palantir, right? Nobody argues with that. I have a hard time not being skeptical about it. Hm. I think we're a bit in the same boat as you, I guess. Yes. I actually think that the whole debate about the Sasap Apocalypse or whatever we want to call it suffered a bit from being a very Twitter-based debate, right? In which it seemed like you had very extreme opinions about it and you gave a big headline like" Zasca. or whatever, "right? It didn't have much visibility, right? But come on, a scenario in which, look, as you explain it, I mean, I think at the same time, a mature company that has a SAS integrated inside with everything and that everything revolves around that SAS, the starting price is very high. That is to say, it is not that simple and it also has to do with what you said. And in the end we have many systems where compliance, right? Complying with the standard, is part of the price that you pay to the SAP in question, right? It gives you security and peace of mind and a fairly interesting assumption of risk, but at the same time I see that the company that is born in 2026, uh, can have different visibility on, right? The cost of profitability changes a lot, right? I mean, SAS is almost always cheaper, right? Than doing it yourself, right? And the CTO who was more or less a purchasing manager, right? You see how the manager changes, such and such. For me, the role of the CTO, which was before, has become partly someone who knows what software to buy and how to integrate everything, more of a buyer and more of an integrator. The CTO likes it more and this is almost a return to basics, right? The C, no, I want to do everything, control everything and be like this, done everything. Now you have opportunities with AI, right? Anyway, what I wanted to ask you about is that, well, you have a very financial product, the GPT Advisor, and that, well, I've been looking at it a little bit, but I haven't been able to try it yet, which basically, uh, allows the financial institution to have an MCP and I can, let's say, work from my CH GPT, from my cloud, with that information, and I find it super interesting. I mean, there are two questions here. One is, how do you think this is going to affect savings, investments, and the financial world of individuals? And who is gaining power with this? I mean, uh, look, I'm going to give you an example and forgive me for going on a bit, right? I recently gave a talk, a conference to people from the world of customer support, okay? All companies that created technology for customer support and companies that used, well,

technological solutions for customer support. So with AI, a whole new world opens up, but there was a blind spot that they hadn't identified and it surprised me a lot because it was, of course, you're all saying, how am I going to use such and such agents? I, well, hey, the blind spot here is that as a user, I'm no longer going to talk to your system, my agent is going to talk. Yes. And maybe my agent is better than yours . Hm. It could be. In other words, there's a reconfiguration of, not of the strength or the negotiating power that changes. So, uh, what's your vision right now, a bit of how all this is readjusted when you get into investment, funds, etc.? Of course, entities, for example, have to decide what their model is. In other words, these decisions, apart from the technological ones, are distribution decisions about how I'm going to distribute my product. And there are some that can afford to do certain things and others that can't . Almost all the projects I do are internal for the bankers themselves, and there they have their commercial arguments, they have everything closed, they have a track, they have access to all the calculations and so on. There is a project that you mentioned that is interesting, which is almost the only one of these that we have done. In fact , there are some other things with index to capital that are very interesting, but this is My Investor, uh, with My Investor it is the first MCP that connects the My Investor database with AI, so to speak, so that the client, through Chat GPT or through their Perplexity or their cloud, can talk about the products that My Investor has, ask openly if they are good or bad, etc. So, of course, only My Investor can do this because My Investor is a platform that has everything and that also does not advise, that it is also a platform where the client does not receive advice, there is no advice from Made and My Investo. So, as it is a transactional platform, it is super well positioned to be able to do this. Okay. One question, Sala, because I don't want to forget to ask you this. So, who is responsible for the Fund's Board? Because here I see a chain of actors, right? My Investor, uh, GPT Advisor, which is a technological facilitator, integrator, CH, because CHGPT has never told me that it wants to be a facilitator or an advisor, and financial advice is highly regulated. I'm very interested in the whole responsibility thing, how does that come about? Yes. I mean, the clear thing is that we've mentioned a very, very critical word, which is advice. In Europe, in the world in general, but in Europe there is a regulation called MiFID, which states that you can't receive advice if you don't have a series of tests done, saved in a database, etc. In reality, strictly speaking, if you ask, I mean, the person responsible for what is said there, the person responsible for the data is My Investor and the person responsible for the connection of the data is GPT Advisor. So, if you start asking, we could do the test, uh, and if you start asking a closed GPT advisor system and, for example, a client cannot receive advice because they do not have the suitability test, you start asking:" Advise me ", and it will tell you:" I cannot advise you ".

It has one of these guardrails. I cannot advise you because such and such ." And as bankers always start, hey, it's a matter of life or death, it depends, my grandmother might die tonight. My grandmother is going to die tonight and sometimes they try to trick you into that. If I don't buy this, if Amundi doesn't advise me on this, then in the case of the My Investor project, which is more open, right? Because it's directly the client. I ask Cloud, the person responsible for what he says should be Cloud or should be responsible for intelligence there, it's more like Cloud Chat GPT. We try to manage it, but it's very difficult to manage it because in fact, there has been a situation in some conversations that I've seen where he recommended a product that is not from Mayo Investor, that is from the competition, for example. In other words, he recommended a transfer of power from Claro, people are on my side, right? Sure, sure. That's why you have to be my investor and have all the products so that that happens very rarely, right ? Or you have a lot of self-confidence if you have it. That's why in the end what's going to change is the offer. What's going to change, if you notice, is the catalog. Entities are going to be forced to have everything, not just their products, which many times, if an entity, I'm not going to mention names, but if an entity were to put its product catalog and its product catalog is unreadable, it's better not to put it, it's better not to do this project, right? It's better not to do this project. But then I think this is where the person responsible is. If you ask Cloud or Chat GPT or Perplexity and ask for advice, they don't give you advice. They tell you, "I can't give it to you, I mean, they know those catchphrases well." And they tell you, "I can't give you advice, you have to go to your advisor, I don't know what, or whatever ." Even so, of course, if you tell them , "Hey, of course, it's no longer advice, it's, uh, what fund is best for me." There it will tell you this, that is, there are ways in which it is advice in some way. Yes, no, you don't buy this fund, but this suits you if such and such, this one, if such and such and more or less lower commission. Then if you go to quantitative things it is not advice. If you say, this fund has a lower commission and therefore it has a lower commission. You think that the big problem that has arisen here, I mentioned to you before that we founded Invertia, Invertia, we set up Invertia in 98 or 99, the first dot-com revolution, right? Back then, when we were counting on Invertia, which we were, super, we were just kids, we already thought that thanks to the Internet and bad or expensive financial products or they were going to disappear, because a kind of perfect competition was going to arrive, because, since all the information is there, you see it and say: "Ah, look, this is better than this one and this one." The reality is that, I don't know , 25 years later or 27 years later, that hasn't happened. Entities continue to have very bad products that customers continue to buy massively, and you say, "Well, how is that possible?""Uh, I mean, we thought that ending information symmetry, which we used to say when we were leaving university, wasn't enough, but it turns out that it hasn't been enough. And the question I have now is, are you

going to end not information symmetry, but process symmetry? I mean, is it really going to end, you know? Is it really going to force a purge of bad products? I think it will happen because I think I perfectly understand that a normal person wouldn't look at a fund if it had a fee that wasn't very well published, if it had volatility. I understand that a normal person wouldn't go into depth about a lot of things to take into account, and it's true that there's a complexity. It happened to me, didn't it? I mean, before buying anything, I read Bogel, I read Bstain, but it's a process that I just wanted to set for myself, but I think it's not the norm, but that people make decisions with very little information, right? And that has been part of the problem with our poor financial education, and of course, it's true that AI can do research, right? It will do it for you, and it does it in a way that, well, that's another virtue that the regulator really likes. There's something that the regulator really likes. I'm learning words with the regulator. For example, they tell me," This is a tool, "when they see it every day, they like it," This is a financial inclusion tool. "That's a nice word, right? And it's true. It's true. And then there's another one that they say," No, no, what I like about it is that it increases explainability. "" Why? Because people don't understand anything, man, Antonio. People are given a brochure and, but what, where do they have to sign? But nobody understands because they're not prepared for it either. That's the reality. And people aren't prepared either. That's probably an interesting debate to have with the minister or vice president friend because the great European project, that is, is to get Europeans to invest much more, let's say, outside of real estate and savings accounts and that would flourish in a greater amount of capital available to European companies. But perhaps what happens is that, and I think it's already happening, where there's a lot of transfer, is to the SP500, right? Which is another debate, to say that the state itself doesn't have the entire offer and it could also say, damn, AI could raise a problem for my future European financing plan, which is going down a path, right? Yes, but look, but don't just think about the capital part for companies, but for individuals. Right now in Spain , as the 40%today, 40%of savings are in deposits and checking accounts. In funds, if I add funds that are not subject to inflation because they are in cash, in very, very conservative funds and such, I think it is 60 to 70% of Spain's financial savings. It's a tragedy. It's a tragedy because you think it doesn't beat the CPI. I mean, 70%of people who have managed to save with all the effort it takes to save, with all the effort it takes to earn money, all the effort it takes to save, are throwing it away because they think they won't have the CPI, but the CPI is the least of it, think about asset prices, I mean, think about housing, gold, Bitcoin, the stock market, the Standard and it's a tragedy . So, if AI can improve that, yes, I'm sure it can, in that sense it will be good. Okay, last question because we're running out of time. Of course, I would

have you here for a couple of hours, Salvador, but I hope you come back because, of course, No, whenever you want, I've managed to get 20%of the script, but there's one that I don't want to leave out, which is the issue of Moats. My impression and I'm seeing very aggressive, for example, GPT is very much in your business. In the United States they've started to integrate with everything, well, like when they integrate bank accounts with another bank, right? They have that type of integration like for the finance discussion at the HRGPT. I see Cloud as super expansive in science, in design, right? Suddenly it's like those who were going to be the platforms enter the services and applications layer very strongly, right? So, of course, that was a dilemma that was the dilemma that we had in another generation like between Microsoft, right? I made Excel for you and integrated it into Windows, you screwed up. And then it was like between Google, right? And Google releases the maps, your maps website goes to hell. Well, the thing is that it's a constant. We're wondering where they're going. to enter the big AI labs, how easy is the experience of these generalist artificial intelligence systems going to be and how can startups find their niche, their place, their MOAT, right? Now that the word is also back in fashion, how do you see this being reorganized? How do you have it planned , Salvador? I think one must adopt the perspective of a startup or whatever; you have to assume the worst-case scenario, and the worst is that there is no mod, which I find to be a reasonable state of affairs. It is obvious, it is obvious, because if you have a night with Cludes to do something you say, "Well, but how did you get this done, right? This used to cost me 3 months and 25 paychecks." So I think that's what we have to assume is that there is no MOAT. What does exist are modes that we don't perceive, but they do exist. And there, for example, I've seen the movements, for example, that Open AI and Anthropic have made, which are creating a large fund. I think Anthropic is doing it with Helman and Friedman, with large venture capitalists and they are hiring that, yes, forward developer engineers are needed to change organizations. So, I think that from the startup perspective you have to think that there is no mode and that despite that there can be a startup because if there is not one, there is a phrase that I always like that I always say in this that is very nice by Ronald Coast that says that a company exists if there is a gap of information between two parties, right? So there are many parties. Of course. So, in other words, it's not going to be distributed all at once, as they say, it's not distributed, right? I mean, it's going to come like this. So, uh, I think that There will be nothing left but this and then you just have to go there and enjoy it. The bad thing is to believe yourself, the bad thing is to believe that you do have a mod that you don't. That's horrible, because then you're fooling yourself and, furthermore, it's easy to believe it because, these days, investors want to invest in anything; There is plenty of venture capital money and they will help you with that. But I think we have to say, "Look, there's no way, let's go all out here and see what happens."

Look, he reminded me and with that I'll end my business as a speaker. The business of speaking is that since I'm a sick person, I try everything, I read everything, I'm superstitious, I'm going to tell people who aren't sick and who have a normal life what will be common sense and everyone will know in 6 months or 2 years. So I give them something in advance, right? Like news from the future. But of course, I'm always saying this about explaining. AI has a certain value because now there are no people going to explain the internet to the people, right? But it is true that there is always something, there is always something and there are people explaining the digital transformation and there are people explaining, right?,a lot of things. Uh, there is always something that the one who is at the cutting edge has to keep up the fire that the gods have created and tell others, right? There is always that gap in reality. Yes. And we have to keep going. Because , for example, when the disruption came that really allowed us to make programs , and I tell you, for us it was in January or December, of course, we already had two guys always trying to do things differently, trying to do, I mean, we were like, let's...when we were going to do a project, I would say : "Come on, let's do it with our license on the one hand and, on the other hand, we are going to do it from scratch, and there will come a time when from scratch will be even faster than our license." So, that has not happened yet. in everything, yet there are projects that can be so. So, if you're not doing that innovation, the day comes it won't catch you, you're not there. Sure. I mean, it's like you have to be there, like they have to catch you working, right? So, if you're not on the wave, it's a very paradoxical issue because on the one hand you say there's no way, but on the other hand you say, but there's no other choice but to be there. Yes. This is a continuous and endless reissue of Only the Paranoid Survivors. Yes, yes, yes. Thank you very much, Salva, for your time, for your knowledge. Uh, good luck with the minister, which I hope goes well, and thank you very much for the time with Monos. I hope it's the first and that we can see you. Yes, no, I'm delighted. The truth is that I also listen to you, I also follow you, I mean, and I've admired you for many years and I love Monos. It's also like it's very funny. I don't like it anymore. I listen to what I learn from AI. I don't know if I learn from...I don't know if I learn. Man, Salva, I think you already learned ...No, no, no. Because with you. You know what happens? If something isn't distracting, I don't either. You know that I really like fiction. I'll read a novel or a movie because if something isn't distracting, on the other hand, I just laugh. I go in the car and I laugh. Well, we're going to take the claims, right? When Carlos Bollero says that this movie is good, well, we're going to put you on the podcast, Alba. Very good, very good. Thank you very much, and well, to you, to you. Yeah.