Investment Bankers Built A $2B AI To Replace Themselves | Term Sheet

0:01Allie Garfinkle:Ready to soundtrack your summer? With Red Bull Summer All Day Play, you choose a playlist that fits your summer vibe the best. Are you a festival fanatic, a deep end DJ, a road dog, or a trail mixer? Just add a song to your chosen playlist and put your summer on track. Red Bull Summer All Day Play. Red Bull gives you wings. Visit redbull.com slash bright summer ahead to learn more. See you this summer. so good so good so good everything you want for summer is at nordstrom rack stores now and up to 60 off stock up and save on the brands you love like vince sam edelman frame and free people join the nordic club to unlock exclusive discounts shop new arrivals first and more plus buy online and pick up at your favorite rack store for free great brands great prices that's why you There are certain types of workflows that will never need to be done by a human again, and I think that's a good thing.

0:58Allie Garfinkle:Gabe Stengel built a$2 billion AI tool to automate his old investment banking job. But for two years, no one was buying it. The early days were pretty terrible. I probably had 30 passes that round. Now more than 35 ,000 bankers are using his tool Rogo, which does in minutes what used to take them a week. And Lazard, the bank that Gabe quit, is now a paying Rogo customer. I can do all of these workflows in one-tenth the time with one-tenth the resources. Last week, two of the world's largest banks signaled major AI-driven job cuts. Investment bankers are especially in the crosshairs as the job is indisputably changing.

1:35Allie Garfinkle:And the work your banker did last year could now be an email. Welcome to Termsheet. I'm Ali Garfinkel.

1:44Allie Garfinkle:And on to this week's news. The SpaceX IPO is upon us. After more than 20 years as a private company, SpaceX is officially going public over the next few weeks. Now, regardless of how you feel about Elon Musk, about SpaceX, about Tesla, this will be a genuinely historic event. It will be, by every account so far we've heard, the largest IPO of all time by a substantial margin. Last week when the S1 dropped, it revealed some genuinely shocking things. I would go so far as to say it was a banger of an S1. Three things that you should know from it. Number one, revenue is up at SpaceX, but so are losses to the tune of billions.

2:26Allie Garfinkle:Number two, Elon Musk pay package is tied to and among other things establishing a colony on Mars of at least a million inhabitants. And number three, Anthropic and SpaceX have an existing deal to the tune of billions, which is genuinely surprising given that Musk has previously described Anthropic as, and I quote, evil. SpaceX is widely considered to be going public in the vicinity of June 12th. And this is also happening at a time when open AI and Anthropic are also widely rumored to be considering going public. Now, I'll believe that when I see it. But what do you think? Are we headed for a red hot IPO summer?

3:03Allie Garfinkle:Let me know in the comments. And now for a dispatch from a very different corner of the AI boom. Here's Gabe. Gabe Stengel, thank you so much for being here. Thanks for having me. Rogo is bringing AI to Wall Street. But what does that actually mean? Like walk me through a day in the life of an analyst, for example, using this tool. I actually think in some ways it's easier to think about the market and the firm perspective versus just the analyst. It's pretty easy to imagine a junior banker, you know, working in Excel more effectively, working PowerPoint more effectively. for us, bringing AI to Wall Street means transforming what high finance looks like and thinking about reinvention of M &A and capital markets and making them radically more accessible, you know, radically more efficient.

3:46And to me, it's helpful to think about the macro, i.e. if it takes three months right now for a banker to go to a company, maybe it's Glossier or someone else and say, hey, how do you get ready for a process that might take three months and then another three months to consummate an acquisition. And I think five years from now, that sort of M &A process can take 48 hours.

4:07Allie Garfinkle:Now walk me through a couple of use cases. Because one of those things is I think there are going to be a lot of people watching this who are going to be like, well, I've never worked on Wall Street. Or even if you have, they may not be able to necessarily imagine this. What a banker does is very different from what an investor does. Some of the underlying workflows are the same. but what a banker is doing is helping a company sell itself or helping a company buy another company or raise equity or raise debt. It's less about, you know, investing your own personal capital in a business. But if you're going to help a business sell, you need to help that business value itself.

4:44Think about what are similar businesses that have been bought and sold so that when you're presenting it to the market, you have reference points for how to value that business. And that kind of valuation work, positioning work, strategic work to understand, the value of a business is very similar to what investors do. But for a banker, the kind of end to end process is the value add, not just that analytics work. And so if you're a great investment banker, it's about finding clients, i.e. how do you go into market and find companies that want to buy other companies or companies that want to sell or companies that want to go public and raise equity?

5:14Great. Once you've done that, it's how do you help value them? How do you help actually think about what the business is worth? And then how do you find requisite buyers who are prepared and agree with your valuation, either to buy that equity in a public offering or to buy the whole business. And that's still, you know, abstracting away a lot of the complexity. But if you think about that full life cycle, what a junior banker is doing is less often speaking with clients, finding deals, speaking with potential buyers. It's more often the blocking and tackling around saying, you know, what does this business work?

5:44How do I get their financials, get a full picture of who they are, and then value, you know, the price per share. And so that happens in Excel, and then it's presented in PowerPoint.

5:53Allie Garfinkle:Now walk me through your life as an investment banker before this, because you actually do have experience with this. Yeah. So I began my career at a kind of storied independent bank, Lazard, that is known for M &A advisory. Did you always want to be a banker? No. What did you want to be when you were a kid? Probably astronaut is the cliche answer or MBA player. But I studied computer science. I loved engineering. I interned at a quant training firm, interned at a big software engineering firm. And then I kind of serendipitously met an MD at Lazard who pitched me on, hey, why don't you come in and serve as a data translator between the M &A teams and between the data science teams and help us think about reinvention of this business.

6:35And when I was an M &A banker, I kind of loved it. I covered healthcare companies. So we worked with large healthcare companies to think about what businesses they should buy. One engagement, we were speaking with the CEO of a$100 billion healthcare firm, and he wanted to buy an oncology company. I did not know what oncology was when I began at Lazard, but it's

6:53Allie Garfinkle:basically - You were what, 22? 22, something like that. But you got to sit in the rooms with these business executives who shaped markets, who shaped companies and industries, and think about real business strategy tied to financials. And so I spent a lot of time doing the valuation for the companies they were interested in buying. But then you also got to overhear your sophisticated banking MDs speaking with these CEOs about the approach, negotiation, how to value the business, why it fed into a strategic mandate. And that was fascinating. What was the most surprising thing you ever heard as a listener in one of those rooms?

7:25We advised a company that was trying to buy another business, and that business was worth a few billion dollars in public markets, and they offered double that. And the CEO of the business that they were trying to acquire just said no. And it was shocking to me that someone could offer twice what the business was worth, but this founder was so mission-driven, believed so much in what they were doing, that even though they would have had a billion and a half dollars wired to them directly the day that acquisition closed, they could say no. And to me, it made it clear, well, M &A is a lot more than just valuing businesses.

7:56It's personal. It's emotional. There's negotiation.

7:59Allie Garfinkle:I started my career covering consumer M &A. I did a lot of cosmetics M &A sort of at the height of the cosmetics M &A boom a few years ago. I guess, gosh, the better part of a decade. But to your point, a deal is never just a deal. It's never just about money changing hands. It's emotional. It sounds like from the way you talk about the story, you actually admire it even now. A deal is always emotional. I mean, even think about buying a house, right? It's just a transaction at the end of the day. But will people do that without thinking it through, without feeling it, without calling their parents?

8:27I mean, the whole thing can be emotional.

8:29Allie Garfinkle:So you quit Lazard when you were 23? 23, yeah. 23? Why? Rogo, my business now, AI platform for high finance, was actually based on my thesis in undergrad of AI assistants for finance. And while I was at Lazard, while I was doing the banking job, it became so clear that what I had worked on as an undergrad would be relevant in my life as a banker. And GBD3 came out. And I went to my co-founder, John, and said, hey, John, that kind of crummy tool that we made could now actually be pretty good. And so we left to start this business. How did you know the technology had gotten there? So the original versions of our tool and these kind of NLP techniques for parsing natural language utterances and turning them into instructions for a machine was really a you know called semantic parsing and it was based on context-free grammars and these kind of like almost rule-based systems where if you said you know run a model on this you picked up the word model and you said great let's run this python code gbd3 was a radical reinvention of of that kind of technique and it was completely different and it was more about you know the scaling of deep learning and these algorithms that were clearly going to work at kind of massive scale versus these kind of janky rule-based systems.

9:43Allie Garfinkle:Now, the early days, what were they like? Because I can't imagine it was easy getting some of these folks to talk to you. Yeah, the early days were pretty terrible. How terrible. We both didn't know how to sell. We didn't know how to build. We didn't know how to hire. We didn't know how to talk to investors. John and I used to joke that it was the most expensive form of business school. And we also used to joke that we should get back in contact with the associate VC who did diligence on us and asked who he spoke to that wanted the product because we couldn't find anybody. And so it was pretty miserable.

10:14And the reality was that the market wasn't ready for it, right? ChatGBT actually hadn't come out. It was just GBD3. And so that kind of wave of AI hadn't washed over the industry and made it clear that it was going to be transformative. And so you just had John and myself, who really had just only spent a few years in investment banking, trying to evangelize what AI was going to do to Wall Street.

10:34Allie Garfinkle:What were those early conversations like, or were there no conversations at all? No, there were early conversations, but it was just, you know, that magic moment, it was hard to show people. Like, the technology wasn't quite there, so unless you had a deep kind of empathy for how different LLMs were from prior techniques and how well they were starting to work, it wasn't clear that you could extrapolate towards everything that was going to happen, right? And even back then, our vision was kind of more blinkered, right? It was, how do you make junior bankers a little bit more productive, a little bit smarter?

11:05It wasn't, how do I radically transform what M &A and capital markets look like? I mean, even we had a hard time thinking about, you know, what does this actually mean? What are the second order effects? And it was very hard for us to kind of communicate that to, you know, bankers, investors as well, too.

11:19Allie Garfinkle:Well, it sounds like part of the challenge was just that you couldn't show that aha moment, but it also sounds like the conversation around AI, I mean, we forget now what the world was like before chat GPT but I remember when chat GPT came out and everyone was like what is this thing and it sounds like Wall Street had that moment too after you'd been trying to have the conversation for a while oh for sure I mean we John and I never used the word chatbot or AI because there this been this prior wave of technology where chatbots were kind of a letdown and so if you came and said hey we're you know this AI chatbot tool people were saying well we tried something like that four years ago and they kind of sucked and so we you know we did all this mental gymnastics to like not say it was AI because, you know, didn't feel like people - What did you say instead?

12:01Natural language interface. It was a terrible marketing term.

12:05Allie Garfinkle:So that went over great. Yeah, yeah. It was pretty good. So when did it change? When was the first customer? About two years ago. And it was really, you know, kind of the 4.0 model, GBD 4.0, where the quality of the LLMs was strong enough to do tasks that were credible and you could see and you could demo live. Before that, the models weren't quite smart enough. Now walk me through the sequence of no's and how that felt and what it actually meant to finally get a yes. I would say it was, you know, the kind of delta between momentum and lack of momentum to me. It is very clear when people say yes, why they are saying yes.

12:52And when you have the wind at your back, why there is wind at your back, right? Is it the product? Is it the moment in time? Is it the zeitgeist? Is it a specific buyer? When things are not working, it's very hard to piece out why, right? Is it because I'm bad at selling? Is it because the product isn't good? Is it because I've misidentified the problem? And so it's very demoralizing because you actually don't know what's going wrong. Is it me? Is it the market? Am I an idiot? Like, you know, there's all these things. And so that two-year period of us constantly pitching, changing the pitch, building, being unable to convince people, you know, it's hard to identify what the failure is.

13:24And so it's feels like you're constantly doing this breath first exploration exploration they just get lucky right like how do you have a lucky break that then you can double down on um and that lucky break didn't come until we started demoing and positioning a very specific version of the product

13:40Allie Garfinkle:um yeah about two years ago it's funny i remember covering a very early round for you guys um the market was completely different then and it was kind of like oh this is going to be a really helpful tool for some bankers on Wall Street, maybe. It was before the partner and compete model in AI had fully emerged. But when you sort of think about the momentum you've gained since then, when did you know that the momentum had shifted? And when, how did your relationship with venture capital evolve through that process? Which was the round you covered, the A or the C? I think it was the A. The A was when we were starting to get kind of branded.

14:18Allie Garfinkle:It was clear you were going somewhere, but if I'm honest, I was a little surprised by how quickly everything picked up. I don't know. Is that all steel? No, no, no, no. I mean, it was, look, Keith Raboi from Coastal Ventures led our A. I probably had 30 passes that round and Keith was actually the only yes. It wasn't like, oh, great. We had a bunch of term sheets and we picked Keith because he's a legendary investor. I love Keith. He is a legendary investor. He is the only person who said yes. I actually remember sitting down with Keith because I could only schedule with him, you know, about a month and a half after I started my process because he was so busy that I couldn't get time with him.

14:52But because it was so late, everyone had already said no, right? Like I had already gone to IC with a bunch of firms and they had all said no. And I sat down with Keith and Keith just goes, all right, Gabe, I get it. AI for Wall Street. It's consensus. It makes sense. I don't like consensus bets. Like I like to be contrarian. And I was like, Keith, well, if it's so consensus, why did everyone just say no to it? And he kind of perked up. And that was kind of the moment that changed in my conversation with Keith. And he saw what we were doing and he understood it in a way that other folks didn't.

15:23And I actually think the kind of mental tar pit that other investors fell into is they said, we're investors, we work in finance, we should be able to use this product today and get the value out of it. And they didn't have vision for where it was all going. And so they ran one question or one workflow and said, oh, this doesn't perfectly address all of my needs as an investor today. It's not going to work. Whereas Keith was just like, oh, this is obviously a huge market. There's really no one else situated well to address it who has the finance background and the technical background. I'll bet on that.

15:55Allie Garfinkle:To your point, there is also a lot of investors, particularly in venture, have at some point been bankers. So there's probably the idea of, well, I know this space. I know this product. I know what it can be. And I'm going to bring back something you referenced towards the beginning, because I think this is an interesting market when it comes to articulating the future. You sort of draw a distinction between where finance has been and where it's going. Where do you think it's going? I mean, I think high finance will kind of emulate what has happened to other parts of the fintech financial services ecosystem.

16:26And I think M &A and capital markets will become more and more efficient and automated. And I think if you, you know, work in private markets and you want to think about selling a business or you own a private market business, it's going to be much easier for you to tap into capital markets and transact. And what happened to public equities over the last 20 years, where it become easier and easier to just buy a share of a business. And so the second order effects are now you can buy an ETF for 50 basis points that bundles all sorts of different products. Those sorts of things are going to happen in private markets as well, too.

16:58And I think people perennially underestimate both the need for kind of human in the loop in fintech, but also underestimate the size of these markets. And to circle back to the kind of buying a home example, 20 years ago, you would not have ever thought about buying a home without speaking to someone, right? Like, I'm about to get a mortgage. This is a huge decision. I need to speak to my local bank branch about that mortgage and what it's going to look like. Now, you know, you can go to Rocket Mortgage and get that in 20 minutes. And actually, there's just as many people that deal with human real estate brokers because they're still emotional parts of the transaction.

17:33But so much of the actual financial services can be provided, you know, programmatically.

17:38Allie Garfinkle:This is actually a great segue into sort of the questions around humans, jobs, AI. I think we should start with Felix. Now, you named Felix after a very famous banker, Felix Roatan. Why? Felix Roatan was famously both a great investment banker, but also served the public good and was a great New York citizen. He saved New York City from bankruptcy. Save New York City from bankruptcy. You know, Felix kind of typifies that prior era of investment banking that was about apprenticeship advice, working with clients, forming relationships versus just the kind of like factory model that sometimes capital markets becomes of just churning out deals.

18:24And as we think about the next decade of high finance, we think there's going to be a return to this apprenticeship model where you invest in people and relationships and differentiated insight and advice. and the type of work that bankers are doing has changed radically over the last 20 years and as it changes what stays consistent is that the best bankers are relationship driven have high trust etc and you know one of the things Felix uh would say if you could if you could find him on the floor at Lazard and I was not around to do this but uh Rahul Recky on our team what was was that you know originally a lot of bankers used to call up their clients at 9 a.m on a Monday and say, this is where your share price is trading.

19:05And that was the kind of novel data that they provided because it was very hard to access that. That is obviously not part of the value proposition of being a banker today because anyone can access that on Google Finance. And yet you still need investment bankers to help you deal with an order of magnitude of harder problems than just finding where you're trading.

19:22Allie Garfinkle:You said something really interesting, that the job of an investment banker has changed over the last 20 years. Walk us through exactly how. You used to have to print out the kind of PowerPoint pages into a book and deliver those books to people before meetings. Now you can send it over email in a PDF. Crazy. Well, but seriously, it's like, imagine if 10 years ago, someone said, hey, guys, we're going to automate the printing out and delivering of these books. People might have said, well, what are the analysts going to do? It's like, well, I don't think analysts are doing that today. And I don't think they're any less busy as a result.