Latest / Fractional Frequency / The Entrepreneurial Leap: Reinventing Investment Research with AI

Transcript

The Entrepreneurial Leap: Reinventing Investment Research with AI · Fractional Frequency

• Billie Tienda: Tonight, we are joined by John Smucker, partner and portfolio manager at Marietta, a boutique employee-owned investment firm with more than 25 years of helping clients grow and protect their wealth. What began as an internal AI tool to improve investment research has grown into Sherlock AI, a platform designed to help investment professionals quickly surface accurate insights from complex financial data. In this episode, we'll explore John's journey from a Investor to founder, the pivotal decisions and entrepreneurial leaps that brought Sherlock AI to life, and how AI is reshaping the future of investment research. John, welcome to Fractional Frequency. Thanks for having me. It's a pleasure. Yes, thank you so much. So I would love to get into the origin story of Sherlock AI. So building it, you know, for internal needs at Marietta because if it doesn't exist, I should build it. Right. Well, ⁓ I guess it was a it was a little over three years ago when the first Chat GPT was released to the wild, ⁓ to the public, and people started playing around with it and I had a ⁓ a actually a business school friend of mine come into town and I took him out to dinner and he said, ⁓ my gosh, are you using this this thing? And I was like, ⁓ I've I've kinda heard about it. I don't even know. Like I googled it a little looked like a bunch of scammy things were out there. I just d I and I don't know how to use it. And so he was he's like, ⁓ my gosh, John, he's like, go here's here's the website. Here like sign up, use it for a week and then tell me like tell me that you don't think is going to change the world. And so I I did. ⁓ thank you so mil. And I use it in immediately, you know, it's just like it's doing stuff for you, little tasks, things like that, writing an email, you know, things like that. So I was like, ⁓ my gosh, this is amazing ⁓ and I'm sure it's a hundred percent accurate all the time and I'm gonna use it for for all this stuff and then you know, then you play around with it for fifteen minutes and it starts just messing up left and right, right? So I think I was like, Wow, well it it like let me learn more about it. So I dove into like the actual ⁓ the research papers that they were they were writing about like all of the math behind it, all of the things behind it, how they built it. And I was like, this should this should work ⁓ for deeper research. And I was I really wanted to use it for my my work for you know, my work is very heavy in in researching individual companies. We we invested essentially like stocks and bonds. So municipal corporate bonds and then, you know, the stock stock market, but individual stocks, not mutual funds. So we dive really deep into research on individual companies. And that just takes a lot of time. And I have, you know, as as part of a boutique firm, you know, and I'm a partner there. So we're we're on top of portfolio management. We're doing the investment research. We're doing ⁓ the client servicing. We're doing client communications, ⁓ relationship management, and, you know, marketing and all that stuff when we have enough time, which is never, right? So we're always I'm always kind of looking for an efficiency. I'm rambling a little bit here, but basically I was like, okay, this should work for investment research. It doesn't, on like the the open large language models. So no one if this this is brand new, no one's done it yet. Maybe I can build it for myself and figure out how to do that. I do not have a tech background. I'm not I am not savvy. I don't I've never done software before, so I just needed to find people. So I was actually I went into a coffee shop one day, and there was a guy that he's like a guy in Milwaukee that knows everyone. And so I was like, ⁓ my gosh, Ian. I was like, Hey, do you like, do you know I'm doing this thing? And I I actually had like the open AI like papers in front of me with and he was like, That's over my head, but I do know a few people doing like AI in Milwaukee, you know. He and he listed a few people and then he like looked across the room and he's like, And actually that guy and pointed at him across the in in the coffee shop with me. And I was just like, seriously? He's Yeah, he runs like this kind of this funny company. I think his name's Michael. And I was like, Okay. And I just marched right up to Michael, ⁓ my now business partner. I explained him what I was doing, and he was like, Hey, come over, come over tonight. I will talk about it. So I went over to his house. He was like, Hey, a lot of people come to yeah you know, he's done software before. He did some some crypto things. He's done he's a pro ⁓ you know, is ⁓ on the board of Marquette University here in Milwaukee too, so He's like he's pretty well connected and he's done some startups before. So he had that sort of experience and that whole side of things. But he said, People come to me all the time with these ideas. I usually just say, you know, help ⁓ connect ⁓ or whatever. But I really wanna do I really wanna do this. I think it's I think it's a cool idea and I wanna do it with you. So the rest was history. So then we started trying to build the ⁓ V P and go from there. But yeah, totally serendipitous. I mean, but it was really just like getting off your seat and going to talk to someone that Actually two different people at the same coffee shop that l led me to the right direction. I mean, honestly, this reminds me of proposal story that you could put on your website that is saying like how we met. ⁓ yeah. I feel like people love that, yeah. It is. It's very serendipitous, like you just said, right? Like you're in a coffee shop and you just happen to be there and your worlds collide. It's very cool. So, now that you're on your way. Welcome to the Founder Journey. What about the journey has surprised you or felt different from what you expected? Well, it's my first startup, so everything is new. I think ⁓ Maybe what I wasn't expecting was like the highs are so high. When you make a breakthrough or you get like the result that you've been like grinding away at, it is like it is like no other feeling I've ever felt. ⁓ it's it's incredible. Like the adrenaline you get, like it's amazing. The lows are so low. When it totally misses or you're miscommunicating with the software team, or they're just like, We can't the like this can't be done, or whatever it is, and like This whole project is a waste of time, waste of money, ⁓ you know, it's not gonna work and everything. I guess so it's the the highs and the lows definitely are are pretty ⁓ pretty insane. That's relatable. I feel like previously you're constantly on the go or constantly going from meeting to meeting. But I think when you're in the position now that we are, it's it's slower. So then when you feel something, you really feel it 'cause you don't have the next thing to really occupy you, right? So yeah, I resonate with that. Yeah, there's always more I mean, there's always more to do and build and everything, but it is very much I mean, we're a really small team, so we're picking our picking our spots and when one of those spots doesn't work, ⁓ my gosh. It's it's just like a disaster. But then when Then you keep keep pushing, keep pushing and you get a breakthrough and like it's the best thing that happened to you all year. So Yeah. So John, asking for a friend, how do you try to mitigate like going too low, right? Because I you know, I think this is like Jenna said, something that resonates sometimes when you c start to get too low, you can go down to a dark place and it can be hard to pick yourself back up. So any like thoughts or tips on how you manage that in yourself? I mean, stay ⁓ staying busy and keep talking to people. I think there's have been a few times where I was like, This is this is just crap. It's never gonna be good and then I had to schedule like someone to see it, a friend or acquaintance, and then they see it and they're like, Whoa, this is really impressive. Yeah. It is. We're building something cool. Like maybe this you know, maybe I'm disappointed. It's never gonna be good enough for me. I know I know it won't. Is ⁓ Yeah, I think just keep talking to people and and fresh eyes are always good too. Especially in software. It's like someone that hasn't seen it, they're like, Whoa, what you're doing here is I can't believe you're you're pull the pulling this off and then it is uplifting. Yeah, it's a kind of a reminder. And that makes a lot of sense. I think a lot of founders are absolutely their own worst critic and so their lens is so much more detailed and the expectations they have for themselves and for their their team are maybe so high that anybody else is still impressed. Even with the words as you're kind of working through it. That's great advice. No, sorry Jenna, I know you've got some more questions, but I just had to jump in there 'cause it sounds so familiar. No, and something you didn't say, but what I I think I mean how I'm interpreting it is that Like one conversation or one bad thing shouldn't basically ruin the day or make you think that what you are trying to build is like wrong, right? It's like, okay, that was one person. So it's like on to the next, right? So being able to just remember like what you are doing and at the end of the day, like no one completes everyone, right? So you just keep pushing what you're wanting to do. I know you didn't say that, but that's what I was interpreting. I mean, yeah, keep I mean definitely keep keep pushing like rejection's part of it, but it is ⁓ Yeah, that like the rejection lows aren't quite as low. I always get a little miffed. This this could really be helping them, but ⁓ but then it's yeah. For me it was more on like the development side when it was just like something just wasn't happening. correctly or whatever, or you know, just blow past a deadline and you're way late on something and you're just like, we're never gonna get this done. That that kind of stuff really that's where I think I hit the lower points. Well, in terms of your, you know, let's say funding philosophy, are you funded or bootstrapped? We're bootstrapped. And this was a decision with me and my business partner. And I don't feel super strongly one way or the other though. I know you said this is maybe a a hot topic and people dig in on it, but like I mean I think for me like bootstrapping is nice, like we're that we can go at our own pace. But that's a double edged sword, right? Where it would probably be going a lot faster if we had if we had funders and board members and things like that. We can ⁓ we can really build it the way that we want. Because at the end of the day I'm building a tool for me to use for myself, for my work. So that's where I'm the expert that I'm building it for. So if we had a lot of other people involved or other investors, they'd be like, ⁓ well, you should do this, you should do this, you should do this, and like some of that, like maybe that would be good advice, but if it's like it this allowed us to be very focused, I think. And so now I've built a really cool tool. I'm just trying to find other people like me, you know, portfolio managers that are investing in individual stocks and looking, you know, especially if they have like a small research team, small analyst team, or maybe no analyst team, ⁓ you know, that type of person. really built a great tool very specifically aimed at them. So I think ⁓ would I love a direct sales team? Yeah. Do I have one? No. We're bootstrap, you know, so I think that's where you the the trade offs come. Would you say that you knew that right away or was it something that evolved into that opinion? Well definitely evolved. I think that's out of out of experience for sure. And you know I I'm not you know, I'm I'm also not good at like pitching the product because I never had a pitch to investors. I just made ⁓ I just think there are definitely trade offs. We're in a situation here where we've been able to pull off like building a very focused, very useful product that I use every day. So it which is awesome. So we're kind of like we're kind of gearing up for that next step now and I'm kind of realizing that we don't have that personnel staffing to do this direct outreach that we need where we certainly would if we had gone the investment route. But I don't know if we would have built like this specific ⁓ of a tool kind of on our own timeline. I think that would have been a lot more stressful. So do you think it plays if you were to have ⁓ let's say the funding route, would it I guess you kinda said this, but would it impact the speed and the way at which you'd wanted it to go? Or would it just given the, you know, ⁓ intensity but also speed at which AI is constantly evolving too? Yeah, given it I mean you'll never know, but guess would be ⁓ it would have happened a lot faster and maybe the tool would not have been as good. And I can say that because as a portfolio manager, I get pitched AI tools all the time that are supposed to help us do research. I've yet to find one that I can't find a mistake in within the first five minutes of using it. And so there's a lot of like highly priced stuff out there that's just AI slop. And I'm sure it's because they're like they raised hundreds of millions of dollars and they need to like get something out there and get some people paying for it and they're just throwing stuff at the wall. ⁓ and that to me would be that's not what that's not how we built it. I don't know if that would have been the way I wanted to go at first. Yeah. So let's pick your brain about that, if you don't mind, John. You mentioned a ton of other tools coming to market. I think this is sort of like the beauty and also the risk of AI being so accessible. It is so the barriers to entry for entrepreneurs to build something without a lot of information or domain knowledge or engineering expertise is lower than it's ever been. So you've got all of this noise around you. Explain to us what makes your tool different. What makes the accuracy and the impact and value so different from all these others we built an AI tool but we hate AI. So Our development team does not use it at it is forbidden to use on the development side at all. So it's all all done with ⁓ old school, I guess, now. Like humans punching in code. And ⁓ there's several reasons for that. One is just ⁓ safety, sloppiness. ⁓ I know I'm not a coder, I don't know all of the things there. There's a lot of very good reasons to to have it be that way. I think also the most important thing for me in doing investment research is the accuracy. And every tool I've seen just like quickly falls apart. I think there's a lot of reasons why. One is just if you're using like a general language model, it's meant to answer general questions and you really like you can you can get specific. That's like when I back to when I was reading the research papers, I was like, this should work. But I think when you have all these different inputs all over the place and you have people prompting different prompts and you're trying to accommodate for all of that, it becomes really difficult. Right? Your the number of potential answers is just like infinite. instead of what are the EBITDA margins over the last five years. Like that's a very specific question, right? So I think what what we did with Sherlock is we we took out prompting. So we just have buttons and you click in buttons and then so you get the same answer every single time. We've also mapped every single variable that when you click a question, it knows exactly what data it needs to answer that. And then we've like mapped to that very specific data pulling the the correct data and then having context around the variables that it's talking about. So like just really guiding the LLM to be so very, very specific. And I think that's where we're getting just better outputs than you could ever do with the general language model. Even if you write a 10 page prompt in Claude or whatever. Honestly, like they've gotten a lot better in the last three years. ⁓ I'll be the first to admit that. Still not perfect and you kind of need perfect ⁓ in the in the industry I'm in. So that's a little bit of a of of how we're how we're getting there. It's like it is it is truly like mapping behind the scenes and then just taking away the prompting, I think does a ton to get like that specific ⁓ information that you need. That it sounds really familiar and and makes a lot of sense. And I've spent my whole career in finance and fintechs. And so that's probably partly why I've been around people who are making all of these decisions on on behalf of customers. There's a lot of gravity to that. But maybe for the benefit of of listeners who have not been in those environments, tell us a little bit about what's at stake. If a portfolio manager or an investment advisor is making decisions with incomplete or inaccurate information, what could that mean for their end customers? I mean, it it'll cost you money. It could potentially cost you money, right? ⁓ if you're if you're making decisions based on inaccurate information. And you can't go after Claude, like you were the one being lazy using, you know, using the LLMs or whatever. So I'm I mean, or maybe not lazy, just trying to be efficient, right? ⁓ with your time and resources. But if you're not careful enough doing that, it can cost real dollars. ⁓ we're in the trust business with our clients. And when you do something like that because you're taking a shortcut, you lose credibility with your client and you're not very likely not gonna be a client very for very long much longer. So I think that's that's what's at stake. Even if you're like an analyst and you're you're pitching to just a portfolio manager and they're ultimately going to make the decisions for the client. And you have some of this these interac inaccuracies that come through just because you're trying to do stuff really ⁓ quickly or whatever. I mean that hopefully there's a catch it like a there's some feedback that you can catch it. But if not, like that can that can cost you credibility with your your investment team and then ultimately that it's hard to recover from that. Yeah. ⁓ that makes a lot of sense. And so you mentioned that, you know, your target audience is really busy. They've got a lot going on. And I'm not surprised. I'm actually reminded one of the first ⁓ groups that I supported in my career was an investment management group and there was a portfolio manager. And I could not get a hold of that guy to save my life. And I actually at one point I was trying to hire for him. He had this open position and ⁓ he was he wouldn't respond to my emails or my outreach on Instant Messenger, I went and sat at his desk and worked from his desk until he got back and said, I need you to give me feedback on these candidates so that we can move this forward. And so I think ⁓ you know you're just pulled in in a number of different directions. It's been a role where there's a lot of plates spinning and it's been that way for a long time. With the entry of tools like Sherlock AI, the ones that actually work and add value, how do you think the role of a portfolio manager or an investment advisor changes. What does that look like in the future? ⁓ well first of all, I I knew you were tenacious. That was well done just sitting sitting at his desk and just waiting. I love that. That's a great move. ⁓ it's really hard to say. I think portfolio managers on one hand can do even more, right? So I can be even more informed. With Sherlock I have five thousand stock write ups at the click of a button. right now. So it's just like click boom, like here it is. So I can do more than that. I don't I don't have to, you know, wait for a research analyst or something or an investment team meeting to like cover it. So I have that at my fingertips. ⁓ but with that, I think and a lot of these tools, they kind of they stop exercising starts parts of your brain that are very important to exercise. So I think the best portfolio managers, the best investment firms have a have a disciplined process and they follow that process. We're always looking for technology to kind of improve and make that more efficient. We're not trying to take away from the magic or the art or whatever it is, those muscles in your brain that like that make the process meaty and and work. I'm not sure if that answers your question, but I think the more and more tools are going to come out and make it more efficient. And I think you just have to be very intentional with the tools that you're using to make sure it's not just hitting a pain point for you, but also hitting one that isn't essential to what Replacing an essential like brain function. Like the way that you describe that. I think, you know, as I consider even parts of my day-to-day where you know it feels like you're challenged always about should you be spending your time doing this manually or should you be leveraging some sort of a tool to get it done for you? And ⁓ there's almost kind of like an internal tug of like actually no that like there are certain things where there's a little bit more. imagination or magic that happens that I don't I don't know if it can be replaced, at least not today. Yeah, and say go going down a little bit of a rabbit hole. I think my my biggest concern with AI and everything i is not replacing jobs or anything like that, but it is it's you know, we live in a time that we we skilled up and became experts in our own fields without AI. So we know what we're talking about. So when this AI slop comes back, we're like, no, that's that's wrong and we can fight with it. I worry that if if you know, if you're growing up and you're just like, okay, just take it for what it is and then move on, there's a very that's a very like low conflict, faster way of doing things, but it's it's worse. So I d and I think that's where like at what point do you lose like the pushback? And you know, I think they've done some studies already, I forget what profession it was. I'll just throw doctors under the bus. But basically they said like doctors plus AI did have slightly better like diagnosed outcomes. But then when they took the AI away, the doctors had a way worse rate of success rate or accuracy rate than they did before the AI, right? So I think that's it's it is it does become a crutch. It starts we start to lean on it a lot. And I think that's that's what I think worries me a little bit more than than the other stuff. Yeah. I could probably get lost in a whole conversation about this. We just welcomed little pack of high school aged interns who are gonna support some, you know, some kind of things that we have going on in the background, some social ⁓ work and and things like that. And I you know, I have to say in my first interaction with them, the impact of just the world that they've grown up in, think about like the, you know, pandemic era high school experience plus all of all of the entry of AI, it's it it makes for a very different eighteen year old than I have ever seen before. And I think ⁓ yeah there's there're certainly some things to consider there for the world at large. Well yeah I mean I've got a two year old at home and she will not know she'll not she won't know a world without without AI of some sort. Yeah, it's wild. I remember before the internet. Yes. Yeah, like going to the library and having I I am not old enough to have like had to, you know, go through the encyclopedia at the college level, but I remember there was like this closed system research database. You you didn't just go on the internet. Like you had to go to the library and like use some sort of a login key and get into the get into the ⁓ the information. So much Yeah Yeah. Well, John, ⁓ we usually end with one piece of quick hit advice. And so I would love to ask you as you consider founders in finance or financial services who are building products that are either AI centered or that leverage AI, what would you ask for them to consider? Can we just stop trying to repackage chat GPT and chat GPTs and can we come up with like I I think that like the the tools are so cool that like the capabilities are really awesome. I would say that like we don't need another thing where I can ⁓ a sandbox where I can type in any prompt. Like we already have that. I'd say the one thing is like do not do not forget about how important customer privacy is in in the financial sector. If if our client data gets out there, we you know, again, like we don't have a client. It's a trust business, right? Remember the business that you're in is a trust business. And so that that is one really important one for me is the second that, you know, an AI tool 'cause we we look at them all the time from at Marietta, the second they're like, ⁓ yeah, it'll live in your system and like do da blah we're like, Nope, you're out. Like I'm not looking for another vendor to be close to our c clients' private information. So I'd say that's if you're a founder k that has to be top of mind, I would say. to to get adoption. I think really wise words. And so John, if anybody's interested in learning more about you or Sherlock AI, where can people find you? Find me on LinkedIn, John Spocker, J O N. Also you can ⁓ yeah. The website is Meet Sherlock ⁓ E E T meetsherlock dot AI and if you go on there and you click ⁓ contact and fill out your info and say you heard me on the podcast, I'd be very happy to give you a a demo week and show you around. So it'll be great to meet people. Awesome. Well thank you so much. John, it's been lovely to chat with you again. Thank you so much for taking some time with us and I'm sure we'll chat soon. You as well. All right. Take care. All right, bye.