Latest / The Longest View with Dez Fleming / Mitchell Jones on Building Lava: Helping SaaS Companies Price for the Future

Transcript

Mitchell Jones on Building Lava: Helping SaaS Companies Price for the Future · The Longest View with Dez Fleming

• [Desmond Fleming] (0:00 - 0:03) From Dayton to Yale, what was that journey? [Mitchell Jones] (0:04 - 0:31) Oh, man. So, you know, me, I grew up in a, you know, lower middle income background. My parents always told me, kind of the playbook was very simple, save your money, get a good education. And for me, I was like, all right, well, I'm going to do my odd jobs and stuff like that. I, you know, would do landscaping. Then when I was finally old enough to actually work somewhere, I worked at a place called Skyline Chili. It's a Cincinnati chili, so if you've ever heard of it, it's like the... [Desmond Fleming] (0:31 - 0:35) I've heard it's big in Cincinnati. [Mitchell Jones] (0:36 - 1:04) And Dayton is about an hour away, so we're like Cincinnati's little brother. And so that was my first job, so I was doing that. But the real thing was get a good education. So I just, from an early age, have been very competitive with myself and just also just understood very... I just always listened to my parents in that way. And so just studied hard, went to Yale. You know, I didn't have a dream school in that way. I wasn't able to actually go to Yale and visit it until I got in, because I just couldn't afford to do the trips where you go to, you know, you lie them all up. [Desmond Fleming] (1:04 - 1:05) What are the 10 you want to do? [Mitchell Jones] (1:05 - 1:09) I wasn't doing that. So it was like, you know... Sight unseen. [Desmond Fleming] (1:09 - 1:10) I'm going to Yale. [Mitchell Jones] (1:10 - 1:18) And then, you know, once I got in, right, I didn't apply to any of those types of places early decision. I was like, man, that's a lot of commitment right now. [Desmond Fleming] (1:19 - 1:20) I don't even know. [Mitchell Jones] (1:20 - 1:32) So then it was a thing of just making sure, you know, when I got there to kind of look at places and I fell in love. I met my best friends very early on. I met one of them at Admitted Students Weekend. [Desmond Fleming] (1:32 - 1:33) Oh, fantastic. [Mitchell Jones] (1:34 - 2:03) Yale has like a Harry Potter house type system and we were both sorted into Gryffindor and life was good. Do you view yourself as a Gryffindor person? If you would have saw yourself in Harry Potter, would it also be Gryffindor? I've done the tests. I go every... I've done Pottermore. I've also done the Buzzfeed one. I'm a big Harry Potter fan. I think Dumbledore has some of the hardest quotes in the world. And I also love quotes. I think they're a good, like, pithy way to live by things. I am definitely a combination of Gryffindor and Ravenclaw. [Desmond Fleming] (2:03 - 2:04) Ah, okay. [Mitchell Jones] (2:04 - 2:38) I'm very nerdy. I love to learn. Yeah. I love to, like, I just... If there's something new I can learn, I prefer to learn it myself. Yes. So I can understand something. And I was like, the Ravenclaw on me. I think my leadership style is very, like, quintessential Gryffindor-y. You know, I played football growing up. It's, like, played sports growing up. So there's, like, a very kind of, like, archetypal thinking of, like, a Gryffindor of, like, we're a team. We win together as a team. We lose together as a team. Those things. But, you know, I got, you know, the ambition of a Slytherin, I guess. They're not bad, too. Yeah. And I think I have the warm, folksy, Midwesterner vibe of a Hufflepuff. [Desmond Fleming] (2:38 - 2:39) You're a little bit of everything. [Mitchell Jones] (2:39 - 3:16) I'm a little bit of this, a little bit of that, but definitely mostly Ravenclaw and Gryffindor. Yeah. When did entrepreneurship get on your radar? When I think about, when most people think about Silicon Valley environment, when you look at some of those charts of where entrepreneurs are produced, obviously Harvard, MIT, Stanford all get mentioned. I think Yale has a underrated, both entrepreneurial and venture ecosystem. When did entrepreneurship get on your radar? Was it at school? Was it after school? Yeah. So I'm going to answer this in two ways. When did I think it did? And then when now did I look back at my life, do I realize that it did? [Desmond Fleming] (3:16 - 3:16) Yeah. [Mitchell Jones] (3:16 - 4:04) Where I'm from, I didn't know what a software engineer was. I didn't know what an investment banker was. I didn't know what a consultant was. When you come from, like, Dayton, Ohio, you're like, the good jobs are doctors, lawyers, politicians. 100%. That's it. And so my playbook growing up was pretty simple. Work really hard in school. You have an opportunity and a shot to give yourself an education to do whatever you want in life. But to me, the opportunity set of what was in life until I went to college was doctor, lawyer, politician. I get to Yale. I'm like, okay, well, I don't really know what I want to do. And people suggested check out finance. I interned at Goldman Sachs as a sophomore, which was actually pretty valuable for me. I got to do two summers. But at that point, you know, Wealthfront, Betterment, Robinhood, Acorn, Stash, those businesses were starting to take off. So here's me sitting on a fixed income debt. [Desmond Fleming] (4:04 - 4:06) And is this, like, 2010 timeframe? [Mitchell Jones] (4:06 - 4:14) Yeah. I was there 2013. Yeah. 2013, 2014. And then Robinhood and Coinbase actually get founded right around there, too. [Desmond Fleming] (4:14 - 4:14) Correct. [Mitchell Jones] (4:15 - 4:22) So be me. I'm sophomore. I'm sitting on this desk. I'm so proud of that. You know, I'm like, I'm here. [Desmond Fleming] (4:22 - 4:22) I'm here. [Mitchell Jones] (4:22 - 6:10) I'm like, all right, I got a good internship. I'm learning about this stuff. I think this investing stuff is interesting. Okay, maybe I'll be an econ major, stuff like that. And, you know, I loved understanding how those concepts connected and how money moves and all those things. Then you're reading all this news every day and you're, in a point in time, people have to remember it's a low interest rate environment. Active management, we're getting their butt kicked. So active investing is just the idea of there's alpha to generate beyond just essentially putting your money in an ETF. And at that point in time, these companies like Robinhood, Wealthfront, Betterment, Acorn, Stash were saying beta investing is actually the way to do it. It's saying we're going to create you a very simple asset allocated portfolio and that's going to essentially be all you need is a diversification. And that inspired me, because I was like, oh my God, my parents who know, my parents saved everything in savings accounts. And I remember I had that realization after my first summer at Goldman's, like, oh my God, my parents are, I'm going to have to help them with retirement stuff because they just don't have all of that right. And they're smart people, but it's just, it's not approachable. And that really inspired me, like, man, these companies that are trying to solve this, they're not here in New York doing investment banking, they're like Silicon Valley startups that are saying we can use our, like, willpower, brainpower to build things that help average people. And that's what inspired me to get into tech. And so from then I started studying, you know, CS, I didn't have enough time to get the major at that point. This is like now my junior year. I double majored in economics and political science. Yale doesn't have minors, but I just kept taking CS. So I had effectively what was enough to, like, I could code, I could build apps, got a job at Dropbox after college, and Facebook had to save up money before I could become a founder. Saved up my money, funded my first business, and that was how that got started. [Desmond Fleming] (6:10 - 6:12) You graduated when? [Mitchell Jones] (6:12 - 6:53) 2016. Okay. I'll add, though, the really most important thing on this journey is I thought my entrepreneurship journey started there, but actually did not. The first entrepreneur in my life was my father, and my father was a builder, right? He recently passed away, so rest in peace. Sorry for your loss. Of course. And, you know, I honor him every day, because he's the guy who, every time I do a new business, actually, when I started Lava, first person I told about it. So I'm on this journey with him right now. But, you know, he was a builder from day one. And, you know, he had a little, small, little construction contracting business, so it's wonderful to be able to walk around Dayton and see little things that he's built and all these things. [Desmond Fleming] (6:53 - 6:54) Go back, go back, go back. [Mitchell Jones] (6:55 - 8:28) And so, shout out to David Jones. I love you so much. Rest in peace. And even though that wasn't technology, what it made me recognize, and to any founders who are listening to this, is like there's many ways to be an entrepreneur, and there's many ways to take the leap to say, why not try? And I got to see a father who tried, and I got to see a father who took those steps, and his brain was always thinking. Let's talk about your first business, then we'll talk more about Lava. But before we talk about your first business, when did you get the conviction moment? You know, because there's, I think, obviously, entrepreneurs have a habit of being entrepreneurs over their life. It's like, you become it, and then you're like, dude, I could never work for someone else ever again. But the hardest part, there's lots of people who want to be founders, and I find that the hardest part for people is always that first step, and for a lot of smart people who come from great backgrounds or so used to planning out, and if, you know, I do ABC, then XYZ will happen, and then XYZ, I do ABC again, and that's just not the case with entrepreneurship. So talk about that first moment of kind of diving in to your first business. Yeah, so the other thing I'll say is I love doing these because it's a time for me to get to talk to founders as well. I did not have money. I did not come from money. I was not connected in tech. As you mentioned earlier, Yale is not the biggest, frankly, tech hub on the East Coast. Harvard's much better, which I hate me to say. It's a little bit more of a connected network. Changing of the guard. Just terrible. [Desmond Fleming] (8:28 - 8:29) It's all right. We're coming up. We're coming up. [Mitchell Jones] (8:30 - 9:16) Obviously, MIT, et cetera. You know, and for me, I've always been good at being able to say, I know where I want to be, and I'm disciplined enough to say what is just one step. Most people don't understand that little steps are actually very big steps. Every single person that you think is great, it's just a bunch of little steps. And they're just very patient on that. So for me, you know, most entrepreneurship for me was like an iceberg. And it was an iceberg that actually didn't start very solidly. I was like, okay, I want to do a, my first idea, not my first company, was a personal financial assistant. Anybody who, if you're listening to this, personal financial assistants are what a lot of investors will call... Darp it. [Desmond Fleming] (9:16 - 9:17) Darp it idea. [Mitchell Jones] (9:18 - 9:21) Mine was an AI-powered travel itinerary. [Desmond Fleming] (9:21 - 9:22) There we go. [Mitchell Jones] (9:22 - 9:24) That's another one. They exist. It's okay. [Desmond Fleming] (9:26 - 9:28) You're all the real ones. You're going to try to figure it out. [Mitchell Jones] (9:29 - 11:58) Come talk to me if you want to learn a couple things. I'll give you some feedback. But anyway, so I wanted to do this idea. Be me. You're now sophomore going to junior year of Goldman. I have, I still have the notebook to this day. I was just writing ideas on how to, what I didn't think like the wealth front and the betterment of a world perfectly had right, how you could blend these things, the need for human touch. I had all of these ideas. I would just write them. I would write them every day and I'd just do research. I was in college. I didn't have, I didn't know, I was learning how to code and didn't know how to code yet. But that was my first step of saying, I'm just genuinely interested in this. And I remember at that point in time, people were like, what are you, what are you talking about? What do you mean? And then the next step was I was like, I need to get into tech. I've always been a believer of the shortest distance between two points in a straight line. I had offers in banking, consulting, and tech. And I was like, I need to go to tech. And that is where, if I want to learn how to build things, I need to get to California and I need to get into tech. And that was the first step. And every single time I got a step done, I just asked myself, what's the next step, right? When I wanted to make sure I had more fintech chops, I went to Facebook. And that was one of the major blessings in my life. Ginger Baker, who's the VP of payments now, took a bet on me to run, you know, she had been handed a bunch of teams and she's like, I need somebody to run mobile financial services, which was Facebook's digital wallet for Latin America and Asia. And I got to be the person to run out at a very early age and, you know, handle a team of a lot of engineers, a lot of moving parts. And I understood what it looked like to run something from the ground up. And then I became a founder, took that journey, saved all that money, started my personal financial assistant and it flopped. I was using my money. I was using my time and it didn't work. I felt like a fraud. I didn't know what to do. I was running out of time. I was about to move back to Dayton because I was running out of cash. And even in that moment, I asked myself, what's the next step to take? And I was like, I need to find a co-founder. And, you know, I got introduced from another friend from Dropbox to my co-founder of Lentable. And that's when that started. So the answer to your question is it's just little steps. Yes, it's gradual. How long did you work on the finance app? Eight months. And when did you know it wasn't working? And what was that like emotionally? I think this is actually one of the hard parts is like, there's this balance of it's very easy in the early days to give up too quickly. I think a lot of people have the idea that I'm going to find product market fit in a week. And that's not actually how it happens. I don't think I've ever heard of that. [Desmond Fleming] (12:00 - 12:01) It just doesn't happen. [Mitchell Jones] (12:01 - 12:06) I mean, we have a little bit right now of everyone's doing some fun stuff with some revenue. [Desmond Fleming] (12:06 - 12:08) You know what I mean? [Mitchell Jones] (12:08 - 14:15) Like it doesn't happen that fast, right? Like, you know, I know every, and founders, if you're hearing another company is now the fastest company to 100 million ARR, don't worry. They might be annualizing their hourly wherever. They might be annualizing their hourly. Focus on finding fit and find to solve real problems. Just find real problems and have enough conversations that make you feel confident that you're solving a real problem. I think the big thing for me is actually, there was a couple problems that were actually uncovered along the way. I recognize a lot of things around where people were hung up in the journey. One of the big things that actually led to, but in my first company, Lendtable, which Lendtable, all we did was we helped you get your full 401k match if you couldn't afford to get it on your own. So imagine you had a $5,000 match from your employer that you couldn't get because you needed to feed your kids. We would give you $5,000. Your employer would give you $5,000. We helped you earn $5,000. We'd take our money back. We'd take $1,000, what we helped you earn. It's that simple. You just made $4,000, no money down. And we were the only business in the world that could get paid out of retirement accounts. Because again, my background is in payments. So everything I do is payments right at the end of the day. But in that journey, right, there was a lot of things with the personal financial assistant that I could see. I would see at the end of the day, you know if someone likes your product or wants to use it if they will pay for it. No one pays for it. And if you do get them to pay for it, it costs far too much for you to even find them. Once you accept that truth, you understand you don't have a strong enough value to buy. You're not solving a big enough pain. But there were very clear things on that journey. It was like, wait a minute. We're making this assistant that's giving you suggestions on what you should do. We kept getting blocked in the same step. Number one, three months of savings. Number two, if you have free money in the form of a 401k match, that is the next step everyone should do in their finance. Get your free money. Every single person we were trying to help was like, yeah, it sounds great. You got $5,000 for me? I don't have it on me. Yeah, then my co-founder was like, actually, as a matter of fact, we could do something like that. [Desmond Fleming] (14:16 - 14:16) Wait, what? [Mitchell Jones] (14:17 - 19:13) Wait, we can just give people money? Yeah, where do we get the money from? Yeah, that's fascinating. I was having a conversation earlier today where I think people overweight downside risk in starting a company because they underweight the value of the experience and what you learn along the way. And obviously, there is a tough personal emotional decision because there is uncertainty with respect to your paycheck. Like people go from making money to making no money, and that's a hard thing to do. Let's transition the conversation a little bit and talk a little bit more about Lava. I think very briefly, what is Lava? Yeah, so in a very, very short, Lava enables companies to track their usage by LOM, by customer, and by pricing plan. By doing that, we solve a very core problem. Right now, most businesses are moving from needing to do seat-based pricing to some sort of hybrid pricing. The reason why is if you're building an AI company, you understand a very clear truth. Seat models are going to look very different because you're actually inherently building something that's automating the way, the need for as many seats. That is a truth that everyone is understanding, and now there is a trillion-dollar industry, which is called SAS, that is trying to figure that out. The problem is we're moving from seat-based pricing to what I'd call measurement-based pricing. Pay for the work that your AI service or agent is doing, paying for the outcomes, paying for these things. One of the biggest things I always tell people is payments is fun for people who just love punishment. It's plumbing, it's the plumbing. Startups should not be doing the plumbing. We make it easy for them to focus on building their core product and not having to build out payment solutions. Because they're moving to measurement-based pricing, the existing solutions would say, slap a fixed-cost subscription on it, just don't work. Because we're moving from a world of high-fixed-cost startups and low-variable-cost startups to a world of high-variable-cost startups and low-fixed-cost startups. We solve that problem at two points for founders. When you're building for the first time ever, we make it easy for you to access models and track them. Usually, that's the only thing you're thinking about when you first build a company is, I need to access models and track them. What we do that's a little bit different than everybody else is we understand that not only that is important, but actually enabling you to then say, I understand how I want a price in the future, and I want to actually associate that with my tracking. Once you understand and make that link, it then becomes very, very easy for you to then slap a price on it. You get back to being as easy as don't make me thick. The most important product principle is don't make me thick. When I'm working with the team, that's our job to do for startups. Do you all also own billing? If I'm a startup and I'm on top of Lava, do I use Lava as my billing engine as well? Correct. You can build with us and charge on top of us. What we understand and the way we think of the world is, the world used to be, you just process. You decide a number and then you process it. You decide $30 a month, you process it and you put a time window on it. Now, what a startup that wants to do any type of AI pricing, or even any company really, we have companies that come to us and say, we're trying to figure out and fix our AI tears, is you need a first measure, then you got to configure a pricing that understands the measurement, which is usually very three-dimensional. We're going from a two-dimensional world, amount and time window to three-dimensionals, type amount one, type amount two, do you want credits, do you not want credits, and then processing. The interesting work where founders right now are getting hung up is they're spending a lot of time in that middle layer. They're having to figure out how to build these systems that can track credits, track cost, understand these things with relation to their data, understand it with relation to their LLMs, and what we understood is any good solution we think can handle all three. To answer your question, we help you with measurement, where that makes it actually a little bit different than everyone else. We help you with pricing and then we also help you, of course, with processing. What's a good example of either an early user of Lava or an early customer of Lava? What are these applications that are building on top of it? A great example is an AI sales company, AI BDR company. This actually is a great lesson maybe for founders as well. If you shout into the void about, hey, is this a problem or not? Usually, if it's a problem, somebody will come banging on your door and say, yeah, actually, I was trying to do this and I need help with it. I'm in a group called Maker. Shout out to Matt Carley. He runs one of the best communities. [Desmond Fleming] (19:13 - 19:13) What is Maker? [Mitchell Jones] (19:13 - 23:58) Maker is a founder community that is actually, if you think about if you've been through YC, I've been through YC before, one of the best parts is the dinners, because it allows you to be vulnerable, get to know your batch, all those things. Maker, the way I think about it, is extending that concept. And it's extending it actually on post-YC. You can be in YC, but Maker is just like-minded individuals who are building startups who recognize that every once in a while it's good to be able to just talk about what you're going through. And so I shout out in the void of that group. And I got a good amount of reaching back out saying this is an issue. And one of the people was very explicit that they were building an AI sales company. And one of the big problems they had was like, Stripe couldn't do most of these things. And the reason why was they had, if you think about an AI sales agent, they need to be able to message you on LinkedIn. They need to be able to get your contact information. They need to be able to email you. They need to be able to follow up with you. They need to be able to schedule time with you. All of these different things are actions within a system that need to cost something. There's no way for Stripe to both handle that and also put that into a subscription and understand all of that. And what was curious to me is, I always think of myself as a doctor, not a salesperson. It's like my belief is that the existing solutions should already be able to solve this. Why can't they not? Can't you just do this? And the answer that came back very clearly was, yeah, there's tools to help me understand and measure my traffic. There are tools that help me process. There's nothing that helps me merge those two and say, okay, I need to understand traffic and how that relates into pricing configurations and then how that relates into processing and getting paid. And the existing solutions require so much heavy lifting. These early stage startups are now spending days and weeks building wallets and systems that, fine, you made the system, but then the problem is, what happens when the latest and greatest model comes out? And you gotta redo the whole thing. It doesn't flow through. And so that was very, very clear. And that was the aha moment. It's like, we can be the link that allows that system to flow very well. And that was actually a problem we solved for our first customer. It was like, there's all these actions. I'm switching them out all the time. There's back end action, which are one of the cheapest models available. And then there's these thinking actions I want to use expensive ones for. Those are changing all the time. I have no way to connect that, reflect it immediately in my pricing, adjust my pricing so I keep my profit margins, and then process it out. And when I heard that, I was like, okay, great. I'm flying to California. I'm gonna sit down with you and I'm gonna understand how we need to build this. So the before and after, when someone's not using Lava too, they are using Lava. Who do you rip out? Who do you replace? Yeah, so first of all, the before and the after for folks is just to understand the before. You're gonna spend weeks doing this. The afters, you're gonna spend hours or maybe a day or two doing this. Honestly, it's gonna be hours. And the reason why is because of how we think of our product. Let me go back to the example I gave before. I think there's measurement, understanding and measuring what has happened, pricing configuration, reflecting that into a tier or subscription, whatever you wanna call it, and then processing. That is how everyone has to do this now. And what we do is most, no one really handles that middle layer, right? We can also handle those other two layers. So in processing, there are folks, many different payment processors out there. They're very, very good at processing either A, subscriptions that don't have any usage, or B, doing one-time payments. Then there's also this measurement layer where there's companies like OpenRouter that route between all your different traffic, things like that. And those are great tools. But the problem is they still don't help you with that middle layer. What we've built our system to inherently do is we understand startup founders, what you're trying to do is you're trying to measure as a means to an end to understand how much you're gonna actually make on profit and what your pricing should be. Because we understand that, we understand those things need to flow together. So you can then actually just use us for all of that and get that through. We also can play with everybody else, right? You may get modular. Correct. It's modular, but the real magic moments, that's another thing, a product guy at heart, I love magic moments. One of the biggest ahas our customers have is when they first use us for measurement and they recognize, wait, it's like a couple button clicks for me now to get paid. [Desmond Fleming] (23:58 - 23:58) Yeah. [Mitchell Jones] (23:58 - 37:46) That is magic, especially when before us you were banging your head against the wall. Did that AI, BDR customer go from a fixed subscription to an outcome-based pricing with you all or no? They had a bunch of different plans. They were measuring it all themselves. They were essentially doing it all by hand. Oftentimes, when I first started talking about this with investors, they were like, who are you competing against? Are you trying to compete against Stripe and Audient and all those things? And I'm like, actually, now we're really competing against, for the most part, people doing this and stuff. So you're largely, you're like a developer tool and like an infrastructure platform generally. You touch payments. Correct. We believe at the end of the day what has changed in the world comes back to the same simple truth. We're moving from a seat-based world to a measurement-based world. The reason why is oftentimes you might only have one agent doing a lot of different things. That agent needs to measure what it's done for you and say, here's how much I've done in this scope of work. Here's how much it costs. Here's all the things I had to buy. Here's the compute I had to use, et cetera, et cetera. That is what we think about. Inherently, measurement is a part of payments now and specifically for services businesses. AI commerce is a whole other thing. I call AI commerce, hey, agent, go buy me a shirt. That's a completely different thing. But in our space, inherently what used to be just a developer tool, measurement, is now becoming a first-class priority for if you want to also help someone get paid. Because measurement is tied to your margin. It's measurement-based payment. Yeah. It's that simple. It's like if you want to be able to continue to improve your margins and price the right way, you've got to understand and you've got to be able to reflect usage. Is this a problem? There's the types of businesses that sell to net new cohorts and grow with them, thinking about selling into YC as a general cohort. And then there are the types of businesses that sell into existing established companies. Obviously, everyone is paying attention to AI these days. Are you leaning more toward the former or the latter as your ICP, meaning who has more of this acute pain today? So you said the right word there, and founders always remember, just the only truth is who has pain. That's the only truth in this whole thing. It's not what you want. It's not what you like. It's just who actually has pain. And if you don't find anyone with pain... No business. ...you don't have a business. It's that simple. Yeah. The pain right now is for whoever is having to become a payments engineer way too early. Who is doing the plumbing? And the answer to that ends up leading you pretty simply. It's the startups who are building agents, startups who are building AI services, the startups building AI workflows, the people who are saying, I used to be building a software business. I'm now still building a software business. It's a services outcome. It's a services agentic outcome. For those businesses, you are going to have this issue every single time. What we actually uncovered, though, is now businesses start to reach out to us and say, because again, the problem is still the problem. For example, there's a large ads-based business where they're transitioning their ads creative tools. They're trying to transition those over to how do we actually make those for all these different creators and things like that. And even those types of businesses, they're like, well, I'm doing essentially agentic pricing. And for building these agents that do these workflows, I need to figure out how to make this system work, make this pricing system work. So we know that, frankly, enterprises are going to have this problem too, larger businesses are going to have this problem too. We can handle those too. But the people day in and day out who we are serving is if you are a founder, if you are a startup, if you are a founding engineer, if you're the CTO, if you're trying to figure out how much time do we want to dedicate to building and maintaining payments in this credit system that allows us to do these plans and then we have to maintain it and then it breaks and someone needs refunds. Do you refund at the processor or do you refund at the credits? We're here to help you with that. And that's who we focus on. That is our ICP. Startups, building AI services now. How does this differ from existing usage-based billing platforms? Yeah, so the great question. Existing, so there's a lot of really great companies that are helping with usage-based payment already. AI companies have a very interesting and very specifically different type of usage-based need. A lot of the way it worked in the past is you would effectively say, hey, you tell me what to listen to but you have to do the work to tell me everything I need to listen to. And then I'll listen to here and listen there, listen there, listen there, listen there, listen there, listen there, listen there, and then you tell me what you also want to price it on and then I will spit out what this needs to cost. So it's almost more like in some ways a consultant for you. It's like a consultant saying, hey, what are all the things you need to track and listen to and then do that. The way we see the world is actually, that's not how most startups think about things. Most startups just say, I'm doing this stuff, can the tools that I'm doing it through tell me what I need to do, inform me of what I need to do and make it easy for me to do that. So in that world, in that previous world where these companies that were built before kind of this, they thought about everything. Frankly, it's also because of who they were serving. They were serving some of these large data warehouse companies and they were serving a lot of those where it's just a lot of intricate data. Some of it lives in sheets, some of it lives in, it's actually like measuring compute. It's measuring a bunch of different things and so they're just doing listeners and then they're using that to inform some configuration of a plan and that configuration of a plan does what it needs to do. The other kind of difference here is for us is we understand that there's really twofold to this. One is that's still too much work, number one. And so how do we make that simpler for AI startups? Number two is there's still a bigger issue at play here, which is they're not necessarily opinionated on how AI is evolving from being right now merchant to human to then agent on behalf of merchant to human to then at some point agent on behalf of merchant, agent on behalf of human working together. That's gonna require some system for those concepts to speak the same language, use the same currency and say we've prepaid and so this can be settled immediately. And that does not exist right now. And that is the bigger observation that we have, which is why for us, we think it's very important and the way we do everything is not only do we help you solve your first order problem, which is how do you get paid now? We make sure every single one of our merchants, we understand how their credits and cost structures work together. So that we see this actually as a larger network plan in the long run, which is that we think that there's gonna be an open system where agents have some form of payments or credits that they use to do things on the AI internet. So in the same way. Yeah, and you need to understand what is this credit worth for my business. Correct. If I'm going to accept it. Exactly. And so they don't really have the legacy usage based billing companies that they don't really have an opinion on that. That's not really their like, they're like, hey, we need to get enterprises that can do, how do we get these enterprises that are struggling with updating their frameworks? Yeah, I viewed as I'm just less familiar, but it's basically like out of my usage of this database, like what should I charge my customer? And I need to have a continuous understanding of that. And then it's like some of this lives in databases, some of this lives in APIs, and you're having to like share your keys with them or even actually just create custom endpoints with them. This is all these things you need to do in order to do it. Where we see it is, it's like, hey, if you're building the app for the first time, you will have not even thought that you were setting up payments, but by building your product, you set up your payments. Yeah, and they're always gonna be focused on solving a pain for their customers. Payments, measurement, tracking usage, all that's an afterthought. And one of the things that we think is very important, if you want to understand where AI is going, we have a sub-concept of you pay through this idea of a wallet. It actually does not matter for customers at all. You don't even really need to think about it that way. Kind of similar to how NewBank is actually very much powered by, for example, Stablecoins. But everything and most of the customers don't actually even recognize that NewBank has created this very elegant system between Fiat and Stables. How much of their business is powered by Stablecoins? I think a lot of it is their underlying rails. So it's like when they're moving money across countries, things like that is usually when they're doing money movements and things like that, that is where that's happening. So in terms of like how much of NewBank's business is that? I don't know, but I mean, NewBank is crushing it. They're absolutely crushing it. Bringing it back to what we think about is there needs to be at some point, we're in stage one of AI, which is right now, sometimes AI workflows are engaging with humans, right? Agent AI workflows are going to become more autonomous. An agent, all that means is an autonomous AI workflow. And that autonomous AI workflow can be structured or unstructured, right? We're thinking, if you're thinking in your head of like Jarvis from Iron Man, that is an unstructured AI workflow that's autonomous. We're not there yet. But even as we get there, it's going to still need to be able to understand how much things are costing, how much work it's doing, and have some form of credits or something or wallet or some capacity to be able to pay for the things that it needs to do. Yeah, you could see a really interesting world where if you have a Jarvis-like product and you're a business, right? Like take procurement. You've got a Jarvis-like product for procurement. And on the other side, your suppliers also have some sort of agent workflow that can complete the work for you. You would imagine there's a world where it's the same thing that happens today. Your Jarvis is going to ask for a quote. And in order to win that business, the other agent on the other side, the supplier agent, is going to need to have some frame of reference of understanding of saying, that quote's going to cost this amount. And in order for them to do that profitably for their business, they would need to have the understanding of, okay, like, what has all of our prior history and usage and cost for our inference, what does that entail, and what should I price this customer at this point in time? You could get to really, and this is one of the other questions that I wanted to ask you is, how do you think about, I guess we'll just use the term of price discrimination in the age of AI. We were talking about this earlier before we hopped on, but different token usage, in my opinion, should have different costs. Today, everything's generally priced the same from an end-user perspective. So, there's two ways we feel about this, and this gets a little bit into that idea of, one of the hard things that we do that's going to matter in stage two is making everyone's business models foundationally map back to some understanding of what their credits inherently internally cost, not with their charging customers. Number one, we want business. This has been around for forever. Founders always remember this. For the most part, in most cases, the customer is always right. If the customer wants to price how they want to price, your job is to inform them as much as you can on how much could you make in this pricing, like what are the ways you could do this pricing, etc. You're pointing out, though, that a lot of businesses right now are not mapping the idea. They're doing a lot of sins. They're mapping this idea of credits, but then credits mean something different if you go to a higher tier. The credits cost different things. Then some businesses are saying, okay, well, when you get to an overage state, the credit doesn't now cost $2. It costs $8. So, there's all these different things people are doing in order to try to make their business models work, which is inherently interesting because they're not always attaching it to the underlying value of a token. And what we do, that is actually why the analogy, the way we think about it, is kind of similar to there needs to be some layer that foundationally understands what was used, and that is us. And so what we do is because we actually have all that visibility, internally we can understand things in terms of how much they actually cost, and then also reference and have a transformation layer that says, here's what you want to communicate to your customers. Yeah, here's what it's going to cost you because you're pinged anthropic and open eye and deep mind, whatever it is. So, here is your reference cost. You decide what you want. Do you want to charge 20% on top, 50% on top, 75% on top, whatever it is. Here's where this gets very interesting. If you believe at some point there's going to be an AI agent, essentially economy of agents doing all these different things. Who do you believe? I think there will be. Yeah. I think that it's going to take some time. The first lever of that then is they're doing work on everybody's behalf and they're doing all these things on everybody's behalf. You have to be able to do what? You need to be able to measure how much they're doing autonomously. And so what that means is then also you need to be able to measure how much things cost. [Desmond Fleming] (37:47 - 37:47) Mm-hmm. [Mitchell Jones] (37:47 - 38:03) And what that means inherently is subscriptions in a way actually block autonomy because they're such high price points, it almost always requires an agent to come back to the user and say, hey, are you sure you want to buy the $500 subscription? [Desmond Fleming] (38:03 - 38:03) Mm-hmm. [Mitchell Jones] (38:04 - 41:06) Whereas if an agent's able to say, I would need to do a purchase that's for 50 cents or an action that's for a dollar, it's going to be able to do that much easier. Yep. If you believe that, then that means that there needs to be some form of understanding of prepaid relationship payments. There also needs to be some sort of measurement of all of the things that are happening in that system and how many credits are moving. As in I prepay my agent. Correct. Preauthorize them to make payments on my behalf. The analogy here is saying actually the internet. We all prepay to use the internet. And there's a standardizing layer called Verizon AT&T that tell you how much data or credits you have used. Data is a representation of gigabytes. Credits are a representation of dollars of compute. And so if you really look at where we're going in the long run, as we get more autonomous, we need to understand no matter how people, how every merchant, how every website, they can talk about it in whatever language they want to. At the end of the day, YouTube can call a stream whatever they want. It fundamentally comes back to data and gigabytes. It's going to be a similar thing in my opinion with the agentic economy. It's going to all come back to dollars of credit, which is dollars of compute. And if you believe that then we need to, whoever learns how to measure that across is going to be in a good space. And so we think that the very important thing in the long run is actually it all comes back to the same thing I've been saying over and over again. Step one is measure. Do you think, let's say we have two startups. One startup is trying to, is helping oncology doctors with parts of their workflow. They're very high value, high stakes use cases because it's related to cancer. Another startup is helping, I don't know, bakers. Right. You know, create coffee for their new case. You know, they're both still, unless something changes in the world, still going to be routing their inference calls to one of the foundation models or through another third party that still routes to those foundation models. Do you think those inference calls should be priced the same or are you saying, hey, we'll leave it up to those providers to figure out the pricing? We'll lead it to the providers to figure out the pricing. There's an inherent cost of those things. Yeah. What we're doing is making sure that that cost is understood across whoever the provider is. The way to think about it is if you even go back, another good analogy is in payments, in store credit in the form of like a card or thing you swipe. The people tried that before like Visa and MasterCard. But what they did was they said everyone needs to be communicating and talking on the same thing. Most of the times, individual businesses are not inherently incentivized to make an interoperable system. Right. Right. So I don't think that the compute level players are inherently incentivized to make an interoperable system where it's easy for that to happen. [Desmond Fleming] (41:06 - 41:08) Yeah. Because they want to drive lock into themselves. [Mitchell Jones] (41:08 - 41:13) To themselves. Yeah. But what we also know is that in the internet, it was great that it got standardized. [Desmond Fleming] (41:13 - 41:13) Right. [Mitchell Jones] (41:14 - 41:29) Thank God. We're not on dial up anymore where everyone was trying to figure out how am I going to monetize and how am I going to pay for this traffic and all of that. Another good example is I'm really glad cards work on card networks. Right. Thank goodness that everyone's not doing their own system. You can now settle a transaction very, very quickly. [Desmond Fleming] (41:30 - 41:30) Mm hmm. [Mitchell Jones] (41:30 - 42:45) When I was, you know, running digital lots of Facebook, the first thing I would do was I would go to the regulator in the Philippines or Peru just because the whole thing was everyone was on the same page of or when I was talking to the bank CEOs or, you know, the startups. Can we do interoperability? Right. That is the key. And none of those groups are ever inherently incentivized to do it themselves, which is why there always ends up being someone who recognizes, wait a minute. Yeah. The real pie here is if all agents are able to move and talk to each other. But then that's not a big issue just yet. Mm hmm. What is a big issue right now is the people building them need to be able to measure this agent's usage. Yep. And reflected in their cost. And then when you have that moment, you're like, okay, there's stage one, which is help agent help AI companies get paid. Yeah. Even if the model providers price their compute usage differently, you're still giving everyone visibility into what your aggregate price is. And at the end of the day, it's like, you know, I think it's pretty clear at this point, you know, like there's a term people use is called being model agnostic. Mm hmm. It's pretty important. I think that became a bigger thing after the quote, quote deep seek moment. Yeah. Where like. [Desmond Fleming] (42:45 - 42:47) It's gonna be way cheaper. Yeah. [Mitchell Jones] (42:47 - 42:50) Like Tony Stark building from a cave with scraps. [Desmond Fleming] (42:50 - 42:51) Yeah. Like, wait, how do you do that? Right. [Mitchell Jones] (42:52 - 43:00) And so then everyone's like, oh man, we need to switch that. But then it's like, do you want to switch all of your API integrations, all your calls, all those things? No. [Desmond Fleming] (43:00 - 43:01) Yeah. [Mitchell Jones] (43:01 - 43:05) Which means that inherently people feel a push to be model agnostic. [Desmond Fleming] (43:05 - 43:05) Yeah. [Mitchell Jones] (43:06 - 43:10) And by definition, what that means is then that means that there needs to be some understanding of these things across. [Desmond Fleming] (43:10 - 43:10) Yeah. [Mitchell Jones] (43:10 - 43:16) Right. That's how I think things are going to end up. Who knows? It could end up being a world where actually one player. [Desmond Fleming] (43:16 - 43:18) I don't think. I don't think. [Mitchell Jones] (43:18 - 44:49) I don't think anyone thinks that or anyone wants that. Another thing I want to talk to you about because I think you have a unique vantage point relative to most people. The type of business you're building is a little bit of an index. Like in success, you would aggregate a lot of these companies that figure out a way to sell the work. Yeah. I'd be curious to get your view in terms of, you know, where are we in that transition? You know, a lot of capital is being invested behind vertical AI startups in every which industry you can name it, you know, in manufacturing, in real estate, in legal, in financial services, in health care. And businesses are starting to work. But I'm curious about, you know, from your view of being at the bottoms up version of seeing that growth. Like what is your intuition in terms of where we are in that transition? Let's just say if we broke it down into four stages. Stage one being there's business use and you can make business models here. Stage two being like this is a clearly accepted as the way to do things. Stage three is businesses that were legacy are like truly waking up to like this is the way I need to like transition my organization. Stage four it's like we're now into that future state of like fully agentic and business is just foundationally different. I'd say we're still in stage one. And that's what makes it very exciting. [Desmond Fleming] (44:49 - 44:49) Yeah. [Mitchell Jones] (44:49 - 45:07) Is there is, you know, founders if you want to if you want to hop in you're not too late. I know sometimes it can sound and feel like it feels it feels late and like you're not, you know, but to get back on like there's always a feeling of am I too late am I too late and the answer is if you're solving a real problem you're never too late. [Desmond Fleming] (45:07 - 45:08) Yeah. [Mitchell Jones] (45:08 - 53:02) You just got to always be willing to just only focus on the problem. But I think we're in stage one and the reason why is right now I think about stages by kind of like sales. You can have a very good product that a lot of people will want to use, but you always got to sell to super users first. The way it was taught to me was like the broken down bus. If you tried to sell Uber before you had all of the bells and whistles and you just thought like would people use it? You would have probably gotten the wrong signal. You would have said most people don't want to use this. But if you had only went to people who had been working all day standing on their feet, were super tired, were raining outside and it was hard to get a cap and you only focus on those people over and over and over again would they use it. And the answer is they probably would. They're willing to do things that are a little bit different than they would have before. And right now the people who are willing to do that, the power users are these like hardcore user startups. Right. Those are the people who understand where we're going and how much more business can transform. But you know I talked to some of my my buddies and friends that are like even in tech companies the amount of AI they're using is still way lower than what it's happening in startups. Right. A lot of companies aren't like yeah understanding the shift and they're like we're trying to understand to make sure like someone said it to me it's like well it's hard habits habits have to change and I'm a big believer that there's a last mile problem within AI. So depending on your job within a tech company let's say let's leave software engineers out of it. But let's say you are a controller or working finance or work you know to a certain extent on the sales team or a marketing team like you don't have purpose built systems the systems of action for your role. Correct. Yeah. Yeah. And that's why I think what actually there's like levels of these things. It's like actually the first thing is is like the way I kind of described it is like at some point in everyone's job you had to learn how to do email at some point in everyone's job many people you had to learn how to do Excel at some point in someone's job you had to learn how to do PowerPoint like we're now at that point where for your job you got to learn how to do these systems. Right. They just they will make you faster. Yep. That being said there's actually another interesting problem which is that most companies don't a have an ability to choose which tools are the ones that makes the most sense for them and B are very worried about tool lock-in right. Stuff is changing so much. It's like do we buy subscriptions to X or Y or Z which is actually another argument for actually things moving towards a more like fluid credit style model. But we'll save that for a bit later. But you know the big observation is there's phase one right like most companies are still most people don't even know how they can get the most out of Claude or open AI. Like there's so much you can get out of those tools even before you move to like in eight and or or any of those things. Yes. So yeah I think we're I think we're very early. Yeah. I think that you know there's there's some things though like some fields you just feel like engineering is a very different field now than where we were before. Yeah because I mean people have 24 seven you know agents working in parallel to them. Yeah. And honestly the thing that we even see is it's actually the real value is the context. Yeah. You now have. How do you guys use. We so actually one of the best ways we found is you know we're more of a we use all of the different you know LLM providers. I have not enforced that like anybody has to use certain tools yet like you need to use codecs and use cloud code etc etc. The biggest thing that you know actually is really interesting is one assessing how good an engineer is has slightly changed. Right like lead coding and things like that is I mean you know an LLM can can do that and dip pretty quickly. I think it's now you have to just number one assess systems design. How good is someone understanding how systems should be built and in number two it's like which is the interesting thing to see is. Are there assignments and things you can give someone where actually you could only get it done with an LLM. Yeah. But you can only get it done right with an LLM if you're a good engineer. Yeah. And so like that is actually that I think it's going to be an interesting next step. Internally one of the most powerful uses that we what we do is number one every time we're having a systems architecture system design discussion we do it first without like tools. We record the whole thing then we actually run it through an LLM that also has all the context of how our product works. It's wild. It's so fun hearing how different companies different startups are adopting these tools because it's like you've got I oversimplify and I'm like you've literally got a genius that will answer any question for you as long as you provide it, him, her with information. literal O.G. You have like yo this guy's been at the company since day one so he knows all the systems. That is the thing if you if you give it the right context is what you have. Yeah. Now the problem still is like scoping right like you still need to make sure you're kind of pointing it to which part of the context you want to reference like which part is important for understanding this there's all of those pieces. Yeah. But really what you have is like I foundationally like the the core issue and I remember this all the time especially being at like Dropbox and Facebook people leave you're like you're handed systems you're like okay how was this built and then we do this trying to get up to speed and now you can understand the decisions they sit within your like we use a company called Quill for our meetings it's a note taker for our meetings. Okay. So now we have these home bases where the context of how we got to the decision why we made the decision exists there also it lives in code. You feed all that programmatically? Yes. Okay. So you almost have like to use the Jarvis example from earlier you almost have a home grown Jarvis like operating system. Correct. And one thing I do want to come back to because I think it's like a piece of interesting tactical advice for building today how do you tactically set up a engineering candidate eval to push the candidates to see oh you can only solve this with an LLM but you still need to have good engineering and systems thinking in order to do that. How do you tactically do that? It's hard right now. like filter mechanism. It's still hard because really usually the way things work is major innovation happens and then we're all trying to catch up to figure out how to actually do that well. The way I've done that is I just go to everyone I trust who I think is a very very smart very good engineer and I'm like help me understand something that if you just ask a model it's going to screw it up but help me understand something that you also can't also get done and that you could code within one hour sitting without an LLM and then I help try to understand that and relate to our business. So for example one of the things that we actually did which was really interesting was trying to understand it like going a layer deeper. Like most of the times LLMs actually sit on like a top layer where they can understand things very well but like for example when you get into like lower level coding or like lower level understanding of things so for example if we want to be able to measure real time APIs that's an interesting problem. Right like how do you build a system and what I mean by that is imagine you're trying to track usage for somebody who's on a call. Right we have like Doctor Scribe, Medical Scribe. Doctor Medical Scribe is on this phone call and all of that usage is being logged in the call. Do you turn the call off? Do you keep it going? What if there's overages? It's just hard. There's a bunch of this stuff that no one has had to think about. They only paid for $10. [Desmond Fleming] (53:02 - 53:03) They only paid for $10. [Mitchell Jones] (53:03 - 54:23) Keep going. It's like who's putting that? What's the protocol for that? But even then using the real time APIs the other issue for an issue like that is latency. Right I can't say oh one second load the call. Like that's like milliseconds of latency on a call is very weird. It's like the worst call you've ever had if there's like seconds of lack. So those are issues where it's like okay you have to foundationally understand the systems design of that first. You could ask an LLM how to do that but then if you're trying to ask that and understand that within the framework of the type of company we're trying to build you can try to give it all of that context for how to solve that type of thing but it just inherently requires creativity and it also inherently requires a lower level understanding of like very important concepts that are engineering concepts. So I think that's like the thing right now but even like doing that what I did is I just went to like the smartest people I know both at Lava and then like also outside and I just like that's the other thing. It's understanding where the limit is today. 100%. And like the other thing founder's big suggestion and I wish I would have done this more with my first company Lin Table. One of your superpowers is recognizing what you don't know and then just immediately saying who's the expert. Let me go find them. Yeah. [Desmond Fleming] (54:23 - 54:23) Right. [Mitchell Jones] (54:23 - 57:02) That is literally it. That's also but another tactical founder building moment but hard to do easy to say and practice hard to do because like let's say in the abstract you're building something hot right now you're building in defense and you're like hey I want to build a defensive missile system and I need to find the one expert in this type of radar like or whatever it is. Again it's one thing to say like yeah I want to go meet that person then you got to go figure out who they are. You got to have enough context to validate okay this is actually the expert. Work whatever you can figure that out maybe a day or two. But then to like go meet them and like get on their radar in time like it's kind of like that's tough tough in my opinion. What I've found it is for sure and that's the job of founders tough and you know the athlete brain in me is like well that's the part of you figure it out you chose the job you know what I mean? You wanted to be the boss. Yeah exactly I had a cost to be the boss as my grandfather would say. And like you know with these types of things right the thing I will say is along my founder journey I have found that one of the reasons why being a founder is one of the best fraternities to ever be in is because people understand how hard it is and they genuinely usually want to help you whether or not they think they're going to make money on that. Yes. That was one of the things that drew me to entrepreneurship was you know there have been people who I've advised who have now done crazy awesome things. There are people who advised me and they're like whoa look what you've done and all that like and it's it's a beautiful thing right? It's like it is a very much entrepreneurship is an abundance mindset and the reason why is that most people are going after these ideas of blue oceans right? They're not going just after of course there's red oceans as well and people there's a concept of red oceans and blue oceans blue oceans are new things that are untapped markets that could be very very big red oceans are very big legacy markets need to get fixed but there's inherently this idea of there's these both oceans and because the set of blue oceans is in theory infinite you know there's this abundance mindset I think that that brings that which is usually things like experts and things like that are very very valuable for example I was at a dinner last night and someone was asking me about lint table and they're like dude I've thought about this idea and I'm like oh man if you ever want to talk about it I can tell you all of the intricacies all of the things I don't you know what I mean it's like I want and that's the other thing founders you want you want to fix the problem yeah you want whether it's you or somewhere else you're like great [Desmond Fleming] (57:02 - 57:03) someone to fix the problem [Mitchell Jones] (57:03 - 59:06) I totally get that okay um Mitchell this was great we went a little over but I do want to end on a question that I love that is actually very related to the theme of basically giving back within the founder ecosystem if you were speaking to a founder who's about to start their company today what is the one thing you'd want them to know or be aware of as they're about to go through that journey honestly it's being a founder is a struggle and you will be better for it it is one of the most important things is it is a literal look in the mirror reflection of yourself every day and there are a lot of hard moments tons and it's a reflection of how many times are you willing to get back up and tell yourself kind of like we talked about originally I truly believe I made a step today I'm proud of myself for that step it has to be a mental belief of I'm proud of that step I made and be able to look at yourself inward and be proud of the steps you're making so founders the most important thing I'll tell you is like take small steps for many small steps you will get to the top of the ladder but you have to be proud of those steps that is greatness right, greatness is not you know when you get the press release and you make the splashy Apple style keynote video that's not the greatness the greatness is when no one else is thinking about what you're building except you that every single day you wake up early you start building you go to sleep late and then you wake up and you do the same thing again whether or not anyone else thinks it's cool that is greatness and even if you feel like no one is watching right now at the end of the day the thing that matters is that you will struggle but that struggle is going to make you a great founder love it love it well I know that thanks for coming back appreciate it man love it