Venture Europe: Entrepreneurship | Technology | Venture Capital | Eu/Acc
00:21:31 · Calin Fabri, James Cadwallader
[00:21:31] S1: They don't fit in a box. Like, they come from the box. I'm not saying that. Could do well, like, maybe with this. [00:21:36] S0: Yeah. I'm not saying that in a kind of, like, a, you know, I'm so I'm so random way. Like, I'm saying it more just like that when I was asked that question, it was like... It was a weird question because I like, there's no... If I... If this one doesn't work, I'll just start again. Yeah. Yeah. That's the good... That's the thing. If ProFound didn't... If ProFound failed, and frankly, even now, if ProFound fails Yeah. I'll just start another one. [00:21:58] S1: Yeah. I I I like the way you say it. And also, think, like, a lot of people were kinda taken aback when Dario Amodei said, look, we either reach $1,000,000,000,000 in revenue or we go bankrupt. And a lot of people are like, no. No. No. But you're doing well now. You're doing well now. I was wondering if you can walk us back to South Park Commons and how you met Dylan and all the ideas that you guys went through until you kinda landed on ProFound?
00:22:27 · James Cadwallader, Calin Fabri
[00:22:27] S0: Yeah. So I think, you know, having started businesses in the past, one thing, it was, like, very obvious to me early, early on. And also the interesting thing, do know the thing that sucked that I did before where I lost all my money? I was doing that solo. And I... I'm also... You know, I I could... You could say I'm, like, pseudo technical. Like, I've worked with engineers most of my career, so I can definitely interface with engineers, but I don't I don't write code. And I was trying to start a tech company as a solo founder that doesn't write code. And I just came out of that, like, I need a cofounder. Yep. And it's so obvious. I mean, Paul Graham has, like, a bunch of very good essays on on, like, you know, the necessity of a cofounder. And I think it's very, important. So... Yeah. It's weird at South Park Commons, like, you kick off in a cohort, and everyone goes and starts building things. Whereas I was just, like, looking around, like, dating. Like, I like, who am I gonna date here? And, yeah, I circled in on Dylan, my cofounder, who was previously a software engineer at Uber, very product focused, like very... A true product talent. And, yeah, we we, you know, I I kind of... I... We shook hands and decided to go into business before we knew what we would build. And I think that's better. Like, I think, you know, if I come in and I'm like, hey. I really wanna build, I don't know, AI for [00:24:04] S1: Gentic payments, which is one thing.
00:24:04 · Calin Fabri, James Cadwallader
[00:24:04] S1: Gentic payments, which is one thing. [00:24:05] S0: Like, if I come in with a thesis, then you're, like, limiting the pool of founders that you can work with. Because, like, you know, maybe Dylan hates the idea of fintech. Yeah. So it's better to go in as, a blank slate and just be like, hey, I'm just looking for a cofounder. Maybe I can just can work on things that you find interesting, or you could work on things that I find interesting. So, yeah, the long story short is we ran through a bunch of ideas. We were looking at lithium mining. We were looking at [00:24:29] S1: Lithium mining. [00:24:30] S0: Yeah. Because Dylan has a background at Uber. He was doing a lot of geospatial work. Right. So we were looking at, like, what applications, like, you know, commercial application of good geospatial product. Yep. And we started exploring lithium mining, but never really landed anywhere there. We got quite deep on this idea of agentic payments Yeah. Which now there's actually a few, like Sounds like a great idea. Yeah. Yeah. There's like... But I think that was a timing thing. Right. The the place where we've rubbed up against was like, hang on a second. How many people... This is 2024. Mhmm. How many people are actually using agents? Right. You could argue what we were trying to build for or the idea in 2024 was basically built for what's happening right now Yeah. Just at this moment, only over the last eight weeks. Like a bank [00:25:16] S1: like a bank for agents.
00:25:17 · James Cadwallader
[00:25:17] S0: Yeah. Instinct, Muse, Grokbat, Dots. These things are are just coming to life now. Yeah. And I think, yeah, this timing thing is so important. Like, we were basically too early. And I don't think we had the perfect founder market fit. I don't have a fintech background, neither does Dylan. So, yeah, we landed on ProFound. Yep. The idea being like, hey. In the future, every business on the planet will care deeply about how AI talks about their brand, products, services. So agents become the new customer. Yep. So if you're the CMO of Nike Yep. How does ChatGPT, how does Gemini, how does Grok talk about your running sneaker? And then the second idea, which is maybe sort of less potent, but is just as important is like, okay. Cool. Well, we can show you what ChatGPT says about the Nike running sneaker, but what what do you do about it? Right. So we had this idea that you wouldn't need human beings sat at a computer looking at analytics. You'd have an agent that we call AI marketer AIM that goes through the data, surfaces opportunities, can build the marketing for you. So now ProFound, skip forward two years. Yeah. We've raised, like, $350,000,000 from Sequoia, Kleiner Perkins, Coastal Ventures, Lightspeed. We're, like, 400 people across. We're headquartered in New York, offices in London, San Francisco, Buenos Aires. We're opening an office in SF. We work with a third of the Fortune 100. So, yeah, some of the biggest brands on the planet. Right. It's hard to name them because they don't always have logo rights, but, like, Comcast, Estee Lauder, Ram, Walmart. What was, like, the the first product? The first product was just like, hey. How do you show up?
00:27:04 · Calin Fabri, James Cadwallader
[00:27:04] S1: How do you show up? Yeah. And then, like, like, how do you go from that? [00:27:08] S0: What does chat... If I ask a thousand questions to chat and Gemini and Grok about the new Nike AlphaFly Yep. And I say, how much does it cost? Or is it better than the Adidas version? Or what's the best running sneaker? Just open ended. Right. What gets outputted by the models? [00:27:25] S1: And then, like, the action would be, let's say, okay, you need to create more content that gets labeled, so then you kinda change how the product is seen, or was the action [00:27:38] S0: Well, the action is, like, once you understand where the answer is coming from Right. You know, are you producing enough content? Because, basically, what's happening in the background is I ask ChatGPT a question, and it's sending an agent to the Nike website, to a third party blog, to a review site, to Reddit, to get that information. ChatGPT hasn't worn the the alpha fly. Right? It doesn't know Yep. Intrinsically. And it uses the web to to build its responses. So your new job as a marketer at Nike is to build a 100 x more marketing, and it can't be swapped. It has to be high quality, you know, non commodity Yep. Content that the model hasn't been trained on already. Yep. So you have to create, like, a lots of original content that is aimed entirely for an agent to consume Mhmm. And then publish, distribute that across the Internet, and then monitor it and see how it performs. If it does well, do more of it. If it doesn't do well, do less of it.
00:27:38 · James Cadwallader, Calin Fabri
[00:27:38] S0: Well, the action is, like, once you understand where the answer is coming from Right. You know, are you producing enough content? Because, basically, what's happening in the background is I ask ChatGPT a question, and it's sending an agent to the Nike website, to a third party blog, to a review site, to Reddit, to get that information. ChatGPT hasn't worn the the alpha fly. Right? It doesn't know Yep. Intrinsically. And it uses the web to to build its responses. So your new job as a marketer at Nike is to build a 100 x more marketing, and it can't be swapped. It has to be high quality, you know, non commodity Yep. Content that the model hasn't been trained on already. Yep. So you have to create, like, a lots of original content that is aimed entirely for an agent to consume Mhmm. And then publish, distribute that across the Internet, and then monitor it and see how it performs. If it does well, do more of it. If it doesn't do well, do less of it. [00:28:34] S1: So so how does marketing look like, I don't know, in five years? Maybe ten years is too much, and based on what Anthropic said, like, we are at risk of not surviving. But what does what does marketing look like like in five years?