First Time Founders with Ed Elson – Could this AI Founder Replace Investment Bankers?

0:26Race the rudders! I'm NFL linebacker TJ Watt, and this is my personal best. YPB by Abercrombie is the activewear I'm always wearing. That's why I reached out to co-design their latest drop. I worked with designers to create high-performance activewear that holds up to my toughest workouts. Shop YPB by Abercrombie in-store, online, and in the app. Because your personal best is greater than anything.

0:58Scott, you started an e-commerce company in the dot-com era. What would be your advice to founders building in the hottest sector right now, specifically AI? The market's frothy. If you have a good idea and it has anything to do with AI, you're going to be able to raise a ton of capital. When capital is cheap, take advantage of it. Raise a shit ton of capital, but don't make the mistake of believing that because you raised a lot of capital and that you're supposed to your company is worth a lot, that it's actually worth that. Take a real disciplined approach to spending. Whenever I've raised a ton of capital cheaply, that company oftentimes struggled.

1:30Your ability to raise capital says something about your business, but it's not the full story. So if the market's right, access the market, raise as much as you can, over-raise, but at the same time, call on a different side of your brain and throw nickels around like they're manhole covers. And, you know, spend money, but be really judicious because just because you can raise a lot of money doesn't mean your company deserves it. What was your approach to spending when you were building Red Envelope? I've always been very scrappy. We were very penny-wise and probably pound-foolish. and then we raised, I don't know, 20, 30 and then 60 million from Sequoia and Western Presidio and all these brand name VCs and we started spending money like drunk sailors.

2:14I don't think, you know, there's a cycle here. You're in something hot. Every VC is going to try and encourage you to go faster until you run out of money at which point they'll come in and cram you down. So I would resist the temptation to spend aggressively. It's really nice to have a lot of money in the bank. if you find a customer acquisition channel that is really paying off, then yeah, put the hammer down and hire people and, you know, make some investments. But I just can't tell you how many times I have seen people who, because they access a ton of capital, they think they should spend it and they think that they're going to be able to get a good return on it.

2:48Because keep in mind, when you're raising a lot of capital, it means other entrepreneurs are raising a lot of capital, which means customer acquisition costs are going up. The price for employees is going up. So I think it's just better to err on the side of not spending enough as opposed to spending too much. A lot of people disagree with me when capital is cheap. You know, make some mistakes, spill some capital. That's fine. I'm on the board of a couple of these companies right now. They're just spending too much money. And everyone's in consensual hallucination that because the category is hot, it means that, you know, we should spend aggressively.

3:21I'm like, spend aggressively when there's metrics that justify it. So anyways, raise as much as you can. but spend as if you're probably not going to be able to raise again.

3:34Welcome to First Time Founders. It's been 18 months since ChatGPT launched, and the market's still going crazy for AI. At the same time, though, workers are anxious. From lawyers, to accountants, to writers, to artists, AI threatens to replace them. In fact, one-third of American professionals fear that AI companies will make their job obsolete. My next guest might have started one of those companies. After working 100-hour weeks in investment banking, he realized that much of the work he was doing could be done by AI. So, three years ago, he built one. His AI model can analyze earnings, assess transcripts, create decks.

4:19In essence, it's your own personal analyst. And even more remarkable, some of the biggest banks in the world are already using it. This is my conversation with Gabe Stengel, founder and CEO of Rogo. Welcome. How are you feeling? Good. Thank you for having me. This is your first podcast, right? It's my first podcast ever. And you recently had a business insider profile. Do you feel kind of famous now? You know, it's getting to me. It's getting to me. We'll see. We should be clear. I've been friends with you for about, I think, six, maybe seven years. And I've been looking for the right moment to do this.

5:03And I think now is the right time. I mean, you just announced a$7 million seed round. You've officially launched commercially. You're now serving some of the biggest hedge funds and some of the biggest banks in the world. And your company is now doing what I think a lot of bankers speculated and feared it might do, and that is it's doing their job. Just as an example, I use Rogo, and I'll ask it to compare the price-to-earnings multiples of the biggest social media companies, and it pulls it right up. You could ask it, what are CEOs saying about AI and their implementation of AI at XYZ Company?

5:43It pulls it up, and it will deliver the answers with footnotes from sec filings from uh investor presentations transcripts etc so i just want to start with the question that i think most of wall street would want to ask you which is is ai going to replace bankers no um i think bankers would actually be happy if if it could replace a lot of the powerpoint and excel work that they do but the reality is that we're a long way away from full automation. And we're a helpful tool. We make people smarter. We make gathering materials quicker. We make putting together PowerPoints and research memos a little bit more efficient.

6:22We're not replacing anyone. Are you sure that a banker should believe you when you make that statement? I mean, tell me more about why you're not replacing anyone. Yeah. Because it feels, I guess I don't believe you. We're not replacing people, right? I mean, even in recent history, five years ago before COVID, I mean, you probably had investment banking analysts after staying up all night creating a pitch deck, they would go print it out, buy in the books, deliver the books to their managing director and partner's apartments. That would take a few hours. No one's doing that anymore because everything's virtual.

6:52It's not like these bankers are working less, right? They're just filling up their time doing smarter, more interesting things. When I was a banker, I spent a lot of time doing very interesting, thoughtful work at Lazard, and that's why I loved it. And then there was occasionally work that was not so thoughtful and not so interesting, and we're helping get rid of that. I mean, there's a lot of examples throughout history of automation, you know, creating jobs, right? If you look at the ATM example that folks like Ezra Klein bring up often of when ATMs got invented, actually what happened was there were more bank tellers than ever over the coming three decades because banks became more efficient to operate commercial branches.

7:25And so they expanded. I mean, I think what you'll see in investment banking and in investing writ large is it's going to become more efficient to operate these businesses and they'll want to expand. Give us an example. I mean, you were a banker and you said that there are some services that you were doing that were meaningless. Here's a great example. You cover a certain subset of companies. When they release earnings, you just want to write a quick update for everyone on your team saying, you know, what did they hit? What did they miss? What are they saying about M &A? Are they interested in M &A?

7:55What are people in the space saying? What's the analyst reactions? That can take a few hours to write. That can get automated. And now what you can spend your time doing is being thoughtful about, okay, so what does this earnings mean for us? Should we engage? Should we present new client, new acquisition opportunities to them? Instead of spending a lot of time synthesizing publicly available information and putting it in a little email. Another example is just like benchmarking, right? Every growth equity VC investor needs to know, like, what does great net revenue retention look like at Snowflake, at Procore, at all sorts of businesses?

8:26What they're doing right now is having an analyst benchmark that on a quarterly basis so that they can reference it in podcasts and reference it in IC memos and so on. That shouldn't be done by hand anymore either. So what kind of work do you think bankers will be doing if they're not doing all the things that you just described? What kinds of opportunities would you say this opens up? Look, I mean, it's why do people use M &A bankers? And just to set the stage a little bit, what is an M &A banker, right? Like an M &A banker is not making investments for anyone. They're advising CEOs on when they want to sell their company or buy another company.

9:00They're offering real thoughtful advice. And that's the work they'll continue to do. And when you're offering really thoughtful advice, you have to have evidence to back it up, materials to back it up. Rogo helps make gathering that evidence a lot easier, prosecute more deals, advise more companies, offer smarter insights and offer maybe more products, right? If you were just offering M &A advisory as a service before, maybe now that's easier for you to do with your team of 100. And you can also offer some consultation style work that McKinsey does, too, or some geopolitical advisory work as well.

9:31And the opportunity to expand your service and enrich it is now kind of endless. Do you think that that's going to be the case across all industries? I mean, one thing that I was just reading about before this interview, one of the most popular books of the late 90s was this book called The End of Work. And it basically just predicted that robots would leave us with nothing productive to do. And this has happened sort of throughout history. I mean, there's this other stat. In 2014, half of tech workers said new technologies would be net job destroyers. I remember back in 2020, when I was listening to Andrew Yang on this podcast, he was saying that every job is going to be destroyed by robots.

10:14In reality, there are currently more jobs than ever. Unemployment's at a record low. It sounds like you think that this is going to continue in banking, but what do you think about AI as it affects the entire world? I mean, it's so interesting you bring that up. My brother, who you know as well, who's much smarter than both of us, brought this up to me this weekend. Much smarter, much smarter. Significantly smarter than mostly you. Mainly talking about you. He brought up the point that 100 years ago, economists said, hey, if we experience the type of economic growth we're expecting, the work week is going to go from 40 hours to 15 hours.

10:52You're not going to need all that incremental productivity. And what have we seen? The average work week in the U.S. has gone from like 44 hours to 41. You know, it's declined a little bit, but people are still working. They're still doing a lot. They're still coming to work every day, even if it's virtually now. I can't really imagine a world in which AI changes that so much. That seems like a cultural phenomenon rather than like a productivity phenomenon. And then the other thing that's really interesting about like AI automating away jobs is it's not just this thing where, hey, if they're capable of doing it, they'll do it.

11:24There's a lot of work that, you know, doctors do that nurses are capable of doing, but doctors have to do it legally. Doctors are the ones responsible for doing it. We will legislate what it has to be done by humans and what doesn't. Right now, autonomous driving could probably take all cab drivers jobs today. We're stopping it. And that's because we want it to reach, you know, a degree of safety and accuracy see that we're more comfortable with than we are with humans. And so what you're going to see is before these jobs get automated, they're going to be done a million times better than AI than by humans.

11:54And by that point, we're going to want them to do it. And there'll be more jobs created, but it's not going to be like that. So let's talk about the product itself. It's basically a chatbot that answers any questions that you have about finance. It's very similar to ChatGPT in that way. It's basically ChatGPT, but for finance. I think that goes for basically every AI startup that I'm looking at right now. Every AI startup is ChatGPT for XYZ. Why not just use ChatGPT? I think there's two parts of that question. One is, what is the interface for the work you want to do? Is it chat or is it something else?

12:32And the other is like the domain specificity of what we do. I'm going to start with the latter. ChatGPT in a lot of ways is a pretty smart 15-year-old kid who has generic accomplishments and can do basic math, can do basic reading, has reading comprehension skills, and so on. Would you rather have a generic 15-year-old or a 15-year-old who has maybe studied finance for the last four years and then also comes to work every day with a big textbook of all SEC filings, all earnings calls, and then reads Bloomberg on a daily basis, right? What is the context they have and what do they know how to do?

13:04Rogo is like ChatGPT in a lot of ways, and it's dissimilar in many more ways. We have so much data that ChatGPT doesn't have. And not only do we train on that data, but - How do you have data that ChatGPT doesn't have? Sorry. We pay for it and we integrate it, right? We have a, you've used the product. It's a different product than ChatGPT. In some ways, it's clunkier. We cite everything we allude to. We have millions and millions of documents that you can pick from when you query. And for a normal consumer who might just be looking up how long to bake chocolate chip cookies for, it's probably a worse experience.

13:35For an investment banking analyst, for a hedge fund PM, for, you know, the CEO of a private equity firm, it has everything they want and all the data they want. And we go out and we form data partnerships and we purchase data from other folks, whether that's market data, whether it's SEC filings directly from the federal government through Edgar, whether it's earnings transcripts through providers, you know, there are all sorts of providers, FACS, SP Global, Refinitiv. We have all this data that we integrate. And then on top of that, we integrate a firm's data themselves. And so we can have all your IC memos, all your historical pitch decks, all your historical Excel models too, within Rogo, you could ask questions about that and not just individual questions like, hey, summarize, you know, why we think that Chewy is going to grow 20 % next year.

14:18You can say, look over all the models we've had for Chewy over the last 10 years and tell me what assumptions we got right and which ones we got wrong. That's not something ChatGPT is even close to being able to do. And it's because we focus on ingesting financial data and citing it in all of our answers. Are you concerned at all that AI is becoming too powerful? I'm sure you get this question all the time, but this is the ultimate AI doomer question. What are your thoughts about AI doomerism versus accelerationism? Yeah, I was actually getting dinner with my dad a while ago, maybe a year and a half ago, and I fully panicked.

14:55I was like, Dad, AI is coming. You don't understand because you don't understand how technology works, but it's coming and it's scary. And I'm not sure why I'm scared, but I'm just, I'm just viscerally scared. It feels like a disrupting thing. Because of the work that you were doing or because the conversations people were having? Just all of the above, right? Like, I mean, it's something that lives in the collective consciousness of folks in our world is this idea of AI, right? People have read science fiction. They've gone to the movie. They've seen Terminator. It's a scary thing to think about.

15:25I am not that scared anymore. And I'm not that scared for a number of reasons. One is artificial general intelligence just feels farther off to me than I think it felt a year ago. And even if it wasn't that far off, I've adjusted the way I think about what AGI is actually going to provide. And I'm excited for it, but I mean, that's above my pay grade. I want to talk about your career. So you were working in investment banking for about two years, and then you decided to start a company. Why? Yeah, I always wanted to start a company. I think that's a result of overconfidence and not wanting to work for other people.

16:07But I mean, even back when I was, I think, a sophomore in college, a buddy of mine and me tried to create a pitch deck preparation tool for investment bankers back then. And we actually went into the Credit Suisse offices and met with like a, you know, someone who led North American investment banking. And obviously we were kind of laughed out of there. We couldn't do anything, but it's been on. Who all you? They're like, literally who let you in here? You're like, don't worry about it. But I've been thinking about it for a while. And I mean, I loved my time at Lazard and I learned so much and I had amazing mentors.

16:41And I thought that LLMs and generative AI was going to change everything. And I wanted to be able to work on it in the way that I thought was going to be most productive and interesting. And I love to build things. Like, I am interested in creating products every day and thinking about the buttons to put in and the interactions and what it should look like. And that's what I'm passionate about. I remember when you started this company, you barely used the word AI. I mean, it was, I mean, you told me it was a, it was a, what did you call it? A natural language interface for financial data. Which just doesn't, I mean, it's taken some— We were so scared of using AI as like a buzzword.

17:22And mind you, when we started Rogo, it was pre-ChatGPT. And actually, we prototyped Rogo using very anachronistic NLP, natural language processing techniques called like context-free grammars. But when we were starting Rogo, we didn't want to feel like we were just, you know, tying into some buzzword-y trend and we were, you know, hopping on the back of this big shift. I mean, we wanted to seem legit. I think that was a mistake. You should use these terms. They're powerful terms. And like, we are an AI tool. Why were we doing all this, you know, verbal gymnastics to not say AI? Because we were like kind of self-conscious and like thought we were like holier than thou.

17:59That was dumb. But it worked. I mean, you raised$2 million in your pre-seed round. And now here you are, you're at a million dollars in revenue. I mean, what do you think your investors liked about you? I'm sure it made you appear smart and specialized. I think the deck is worth a whole other episode of this just to look at how bad that deck is. It is one of the worst fundraising decks of all time. Kevin Ryan led our pre-seed. Kevin Ryan is an enormously successful entrepreneur. He's a titan of New York City VC and entrepreneurship. Started MongoDB, Gilt, Business Insider, Zola, and so on. I think Kevin just thought, these guys know finance.

18:36These guys know technology. I'm going to bet on them. I'm going to bet that they can build a tool that actually disrupts financial services. And I think that's what it was. And what do you think you did right that made him believe that you were the guys to do it? Because I'm sure there are thousands of startup guys around the country saying, I'm going to build the AI for finance, whatever it is. You kind of hit the jackpot. You're now the AI for finance guy. What do you think he actually saw in you? I mean, at Lazard, I had a really interesting role. I was working on both M &A work, financial services work, but also on their data groups team.

19:13This team called LDAC was our data analytics group led by David Wang, super smart. And so I got to see what it was like to work at the intersection of technology and financial services at a firm with very cutting edge technology and data science and AI capabilities. And I think that was an incredibly unique background. And that was only possible because, you know, I'd been I'd been I'd come there after having worked in software engineering and quantitative trading and writing my thesis at Princeton on AI for econometrics analysis and financial statistics. I think that the story made sense to Kevin.

19:46And when you're a guy like Kevin, you're looking for a story that makes sense. Right. You want to believe that Gabe Stengel is going to change financial services, that he's going to be the next Bloomberg. You want it. You want to get the pieces to that narrative. And I think the pieces were there. And I was just tremendously lucky. I mean, John, my co-founder, and I were just, and Tomás, were just incredibly lucky to have all been in the right place at the right time, interested in the right things, and, you know, having said the right words and meeting the right people. How have you convinced yourself of that story?

20:15Because the story that you describe is a pretty big one. And the expectations are high. And you mentioned this sense of overconfidence. Did you need to convince yourself that that story was true? Do you believe that that story is true? Early on, frankly, when we raised money, when we raised that first$2 million, I thought to myself, you know, this might not be the thing that makes me enormously successful, but it's going to set me up well, right? Like, I'm going to do this thing. I'm going to do it as well as I can. And then I'm going to have a lot of opportunities afterwards. That really shifted over the past year, shifted to the idea that, no, no, this is going to going to be my big thing.

20:56This is going to be my life's work. This is the best opportunity I have to create a massive company that's valuable, not just to the people that invest in us and myself, but to the people and the clients we serve. And it's going to be enormously valuable to our customers. And I think that thinking has changed in part because of what we've been building and what we've been able to accomplish and, you know, the trends in generative AI writ large. And I I think it's not that often you find yourself at the center of this changing dynamic system in tech where there's a new technology that's enabling so much rapid growth and you're situated selling to a clientele that has the most appetite for increases in efficiency and increases in intelligence than anyone on earth.

21:40And I think we have a tremendous opportunity and that did change. And frankly, it wasn't that I had to like sit down and convince myself, hey, Gabe, you're going to be able to do this. It was like as more pieces of evidence accumulated, my thinking changed. And I wish I could be really thoughtful about like how I thought about it at the time and how I thought about it now. But I think frankly, like as things ebb and flow and you're more successful and less successful, you make products that work better, you work that work worse. The thesis develops. Would you say that you're sort of throughout your entrepreneurial career so far, you've been shooting from the hip in a way?

22:14That's kind of the way it sounds. is you go out and you just try as many things as possible. I mean, it sounds like you haven't actually overthought the journey too much. Look, part of the reason that original pre-seed deck is so bad is because it's very overthought. One page has a million graphs and charts and a million things on it. And it's not that we weren't thoughtful, it's that all that thoughtfulness was wrong or it was on the wrong thesis or using the wrong assumptions. We try to be enormously thoughtful and then we try to change our minds very often. I mean, you have to try a lot of things and see what sticks and run with it and know how to capitalize on opportunity.

22:52I mean, we've tried all sorts of things, right? We've iterated on all sorts of technologies, whether it's querying structured data, unstructured data, internal data, you know, presenting it in one way or the other, the way that we actually synthesize insights. From a product and tech perspective, you have to be incredibly nimble and dynamic. And then from a selling perspective, it's the same thing, too. I mean, figuring out what your message is when you go to market and sell to a hedge fund, that's just as much IP as the technology. Figuring out what words resonate and what convinces someone to open their checkbook, that's IP too.

23:24And the only way to figure it out is to be really dynamic. And, you know, oh, that hedge fund didn't like this language, but this use case resonated. Let's tweak. Oh, you know, this use case didn't resonate this time. Why is that? Let's tweak it and so on. And so it's not firing from the hip, but it's just being very nimble.