The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
01:05:20 · Harry Stebbings, Luke Harries
[01:05:20] S0: it hard. They did kind of change their whole AI team, I think, as a result of What have you changed your mind on in the last twelve months? [01:05:28] S1: The biggest thing I changed my mind on in the last twelve months is whether voice AI agents or conversational AI agents actually work. And so when I joined ElevenLabs, we had this fantastic box where you could type in text, generate life like speech. And Matti was saying, like, oh, soon you'll be able to have full AI agents that you can have natural conversations, and it'll sound really human like, and you could use that for sales and for customer support and for relationship coaching. And I was like, that will never work. The latency will just be too high. You won't be able to have these sub two hundred millisecond interactions. I was completely wrong. And so on about the twentieth customer that us at ElevenLabs had helped build these conversational AI agents, we were like, okay. Let's actually turn this into a proper platform and build conversational AI as a feature. And now it's... I actually prefer speaking to an AI customer support agent than a real person because they're more knowledgeable. They know exactly how to solve your problems. They can escalate. They know all the policies.
01:05:28 · Luke Harries, Harry Stebbings
[01:05:28] S1: The biggest thing I changed my mind on in the last twelve months is whether voice AI agents or conversational AI agents actually work. And so when I joined ElevenLabs, we had this fantastic box where you could type in text, generate life like speech. And Matti was saying, like, oh, soon you'll be able to have full AI agents that you can have natural conversations, and it'll sound really human like, and you could use that for sales and for customer support and for relationship coaching. And I was like, that will never work. The latency will just be too high. You won't be able to have these sub two hundred millisecond interactions. I was completely wrong. And so on about the twentieth customer that us at ElevenLabs had helped build these conversational AI agents, we were like, okay. Let's actually turn this into a proper platform and build conversational AI as a feature. And now it's... I actually prefer speaking to an AI customer support agent than a real person because they're more knowledgeable. They know exactly how to solve your problems. They can escalate. They know all the policies. [01:06:27] S0: Also You don't have to be fake polite. You know what I mean? You're like, I know that you don't have feelings. It's like, what's this? What's that? Yeah. Yeah. Totally agree. Percent. When does the infrastructure layer become the application layer? And what I mean by that specifically is, like, you could be Synthesia. You could be, hey, Jen. Yes. Why does the infrastructure layer not become the application layer?
01:06:27 · Harry Stebbings, Luke Harries
[01:06:27] S0: Also You don't have to be fake polite. You know what I mean? You're like, I know that you don't have feelings. It's like, what's this? What's that? Yeah. Yeah. Totally agree. Percent. When does the infrastructure layer become the application layer? And what I mean by that specifically is, like, you could be Synthesia. You could be, hey, Jen. Yes. Why does the infrastructure layer not become the application layer? [01:06:46] S1: I think it does, but you as the infrastructure player really need to partner with companies and be very explicit around which areas you want to go deep on. We want to build the best platform to make these end to end conversational AI agents, but we explicitly don't want to be building full digital avatar platform. So that way, we're able to be trusted partners to companies like HeyGen, Synthesia, Captions, who rule customers, but we're able to build these fantastic infrastructure layers on top. You know, I invested in Captions. Really? [01:07:18] S0: Yeah. They're doing great. So I I invested in Captions. This is... This shows how theory driven I am as an investor. And... Invested when it was basically a BeReal competitor. Wow. And then a year later, a friend of mine texted me saying, like, we've got to invest in this company, Captions. And so I found the founder on LinkedIn. I messaged him. Like, oh my god. We'd love to find a way to invest. We love Captions. And he's like, dude, you already invested. And I'm like, say what? And he's like, we're a pivot. You remember me? You remember me? I'm Gaurav. Wow. And I was like, oh, shit.
01:07:18 · Harry Stebbings, Luke Harries
[01:07:18] S0: Yeah. They're doing great. So I I invested in Captions. This is... This shows how theory driven I am as an investor. And... Invested when it was basically a BeReal competitor. Wow. And then a year later, a friend of mine texted me saying, like, we've got to invest in this company, Captions. And so I found the founder on LinkedIn. I messaged him. Like, oh my god. We'd love to find a way to invest. We love Captions. And he's like, dude, you already invested. And I'm like, say what? And he's like, we're a pivot. You remember me? You remember me? I'm Gaurav. Wow. And I was like, oh, shit. [01:07:53] S1: Anyway... But I'm sure you really helped him through that pivot, and it was very useful having your advice. [01:07:59] S0: Listen, I don't wanna say I'm like a cofounder, but [01:08:01] S1: Yeah. [01:08:01] S0: I was it was up there. Why is inbound SDR as a role dead? If you think about
01:08:01 · Harry Stebbings, Luke Harries
[01:08:01] S0: I was it was up there. Why is inbound SDR as a role dead? If you think about [01:08:08] S1: the actual function of most inbound SDRs... So with an enterprise sales cycle, someone fills in a form. They then need to speak to an inbound SDR who basically goes through band screening. So they work out, do you have the budget? Do you have the authority? Do you have the need? Do you have the timing? If yes, they then book you into another call with the AE, which actually closes the deal. And so the function of an inbound SDR, I'm sure the best ones are persuasive and they're friendly, but, really, the function is data collection. And so instead, one, why don't you just add that to the form in the first place? But the reason is people don't really want to go through this back and forth and go into more detailed questions and, you know, trying to design these complex flowcharts or forms. And so instead, you can switch out to a conversational AI agent, which can completely replace that role and then get the people who should be booked in with AEs much quicker. You could even clone the voice of the AEs, so it's like them almost prescreening for their own deal flow. [01:09:06] S0: I love that as an idea. That'd be very funny to do. You could do a lot of very funny things with your promotional videos, by way. [01:09:12] S1: Yes. What should we do? [01:09:14] S0: Well, I I think you might also get in the realm of, like, copyrighted problems. [01:09:18] S1: Oh,
01:09:18 · Luke Harries, Harry Stebbings
[01:09:18] S1: Oh, [01:09:18] S0: yes. But, like, I I don't know if you're allowed to, like, use, like, Trump's voice and Obama's voice and, like, just rip. [01:09:24] S1: We're not. And, actually, we do a whole bunch of different safety work to make sure, like, it's explicitly used where people give permission for their voice on the platform. [01:09:32] S0: Don't do that, kids. Yeah. Yeah. I remember when I met Victor from Synthesia the first time, like, five, six years ago. He was like, imagine, we could have, like, Trump declaring this and doing this. And I remember being like, no no no. That's a bad idea. And I didn't invest in the seed. Oh, wow. I know. [01:09:47] S1: No. Well, he's done great, and I know they've... They take safety very seriously as well. [01:09:53] S0: No. They do. Absolutely. Sorry. That sounds terrible. He's so good. I like him and Stefan a lot. Final one. What's the best growth strategy that you've just been really impressed by in the last, like, twelve to eighteen months?