ElevenLabs Rebuffs Takeover Bids to Build Voice-First AI Platform

ElevenLabs 拒绝收购要约，打造语音优先的 AI 平台 联合创始人 Mati Staniszewski 拒绝了多份收购要约以保持这家初创公司的独立性，他押注随着模型引入刻意的人类瑕疵，计算界面将永久性地从键盘转向语音。 Mati Staniszewski 详细阐述了 ElevenLabs 的演变历程：从受波兰旁白配音（lektor）传统启发的一个想法，发展成为一家将纯粹的音频研究与应用产品部署相结合的前沿 AI 语音公司。他解释了为何公司效仿 Honda R&D 模式和 Palantir 的前线部署工程架构，采取“引擎优先”的策略，选择完全专注于语音领域，而非将资源分散至视频虚拟形象或通用 LLM。Staniszewski 探讨了公司在金融科技、电信和无障碍领域增长最快的企业级用例，并重点介绍了刻意的声音瑕疵与细腻的情感表达如何增进人类联结。他强调了放大人类潜能而非将其取代的理念，最后分享了让语音取代键盘成为主导计算界面的长期愿景，并解释了他与联合创始人为何多次拒绝收购要约以维持完全独立。 ElevenLabs 拒绝了多份收购要约以保持创始人的独立性，优先推进将语音转变为主要人机计算界面的长期愿景。 (56:39) ElevenLabs 采用了受 Honda 启发的运营架构，将内部前沿音频研究与不到十人的小型自主产品小组相结合，而不是简单封装第三方 API。 (11:36) 借鉴 Palantir 的做法，ElevenLabs 将前线部署工程师直接派驻到企业客户的工作流中，并直接向产品部门汇报，以便将客户解决方案提炼为可扩展的平台功能。 (27:36) ElevenLabs 发现，技术上完美无瑕的合成语音会引发用户的不安，而停顿、犹豫和语气词等刻意的不完美之处对于用户自然地接受这项技术至关重要。 (52:46)

ElevenLabs co-founder Mati Staniszewski argues that natural voice will replace keyboards and screens as the primary interface for artificial intelligence, requiring a specialized operating model that unites frontier audio research with direct enterprise deployment. Audio Specialization and the Engine-First Philosophy (11:36) Drawing a direct parallel to Soichiro Honda treating his enterprise strictly as an engine laboratory rather than a conventional automaker, Staniszewski explains that ElevenLabs confines its research exclusively to proprietary audio models rather than building general-purpose foundation models or coding assistants. The company tests prospective initiatives against a strict internal filter: whether proprietary audio grants an unfair, defensible advantage. For example, when evaluating digital video avatars, ElevenLabs paused the project after recognizing that video rendering—not speech synthesis—was the core bottleneck limiting quality. Staniszewski maintains that high-fidelity speech synthesis relies heavily on novel neural architecture design rather than raw compute scaling alone. By concentrating researchers solely on audio, transcription, and real-time voice orchestration, ElevenLabs aims to build an enduring moat in human-level vocal expression while relying on external partners for complementary components like large language models and video generation. Evolving from Media Localization to Enterprise Infrastructure (2:59) The company's initial catalyst was Staniszewski and co-founder Piotr Dąbkowski's frustration with Poland's traditional "lektor" broadcast style, where a single monotone male voice reads over foreign dialogue. While the founders originally set out to solve automated media dubbing—inspired by the universal "Babel fish" translator from The Hitchhiker's Guide to the Galaxy—direct creator feedback forced an immediate pivot. Creators explained that everyday post-production tasks, pre-narration pacing, and clean voiceovers were far more urgent bottlenecks than cross-language translation. ElevenLabs paused the translation component to concentrate on generating natural speech, which subsequently expanded into consumer applications, a marketplace of over 20,000 authenticated voices, and enterprise communication infrastructure. Conversational AI agents and creative production now generate the bulk of corporate revenue, led by rapid adoption across fintech, healthcare, and telecommunications. In one deployment, Deutsche Telekom expanded an initial marketing audio trial into call center voice support and in-network telecommunications features integrated via CRM and telephony pipelines. Product-Embedded Deployment and Autonomous Pods (31:16) Drawing from Staniszewski's background at Palantir, ElevenLabs places Forward Deployed Engineers (FDEs) inside the product organization rather than the go-to-market division. While these engineers deploy on-site to customize agent logic, simulate workflows, and audit client systems, their core objective is abstracting operational discoveries back into generalized platform features. Staniszewski notes that without this feedback loop, enterprise deployments quickly degenerate into low-margin professional services consultancies. Internally, the organization limits structural hierarchy to roughly five reporting layers, maintains broad managerial spans of control, and organizes around autonomous pods capped at fewer than ten people. Software engineers are embedded across operational departments—including legal and revenue operations—to automate repetitive administration. For instance, internal revenue engineers replaced conventional web registration forms with interactive conversational voice agents, which dramatically increased prospective client satisfaction while capturing richer unstructured sales data. Natural Voice Dynamics and Human Accessibility (49:09) Staniszewski views voice as artificial intelligence's ultimate interface, arguing that future conversational systems will blend cognitive processing with emotional intelligence by adapting cadence, listening, and interjecting naturally. In developing conversational agents, ElevenLabs discovered that synthetically perfect speech unnerved listeners; intentionally injecting human flaws, such as pauses, hesitations, and conversational filler sounds, significantly raised user comfort and engagement. Beyond commercial automation, ElevenLabs treats voice as an essential component of personal identity. The company collaborates with approximately 800 partner organizations to restore synthetic vocal models for more than 10,000 individuals suffering from conditions such as ALS and throat cancer. These restored voices have supported diverse public milestones, from brides reciting wedding vows to an address on the floor of the United States Congress and an Emmy-winning podcast hosted by former NFL player Tim Green. Staniszewski affirms that ElevenLabs has turned down three or four formal acquisition offers to remain independent, believing the current platform transition offers an unprecedented opportunity to redefine human-machine communication. However, the company faces the long-term challenge of defending its specialized audio lead as frontier foundation model labs continue to invest in multimodal capabilities.

[0:02] Michael Dell: this time is different. [2:59] Mati Staniszewski: In The Hitchhiker's Guide to the Galaxy, there's this idea of a Babel fish that you put it next to the ear, and you can understand everything around you regardless of the language they speak. So, we knew that we will get there. [11:36] David Senra: a research lab for engines [14:46] David Senra: Does that product have an engine? [16:15] Mati Staniszewski: Do we think we have a unique advantage by applying our audio models in that product experience? [20:22] David Senra: No one knows about ElevenReader. You have to do a better job. [20:22] Mati Staniszewski: We think how you communicate as a company, how you think about content communicates, all of that is changing, and we want to be at the intersection of that. We build the best models to help you do that, and then build a platform that effectively delivers that content, delivers the knowledge, delivers the conversation in a completely new way. [41:48] David Senra: The single most powerful pattern I have noticed is that successful people find value in unexpected places. And they do this by thinking about business from first principles instead of formulas. [44:38] David Senra: Taste is as rare as a unicorn. [47:14] Mati Staniszewski: amplify human potential rather than replace it. [60:08] David Senra: If you love what you do, they couldn't pay you to stop. [66:49] David Senra: stay away from the circus. [66:49] Mati Staniszewski: I got the first invite for one of those conferences, and it's like, 'Oh, this is super cool, super fun.' And then, you go. I'm European, so I drank alcohol at the time as well. And then I felt tired. I didn't think I did anything, and it seemed fun, but it wasn't very productive... You go there, you have the sessions and some of them are, of course, good, but most of them are not relevant to you or your business, or they are the circus part that you mentioned, that, like, you feel you are doing something important. It's completely unimportant.

[0:02] - We ran into each other at Michael Dell's event a few months ago, and you're like [0:06] "Oh, I'm very curious how building a company today is different [0:10] from building one in the past." [0:12] And Michael has 40-plus years of experience; [0:14] - Yeah. - he's been dominating for decades. [0:15] And when I saw him, we had a 30-minute conversation about this. [0:19] You read all these biographies of history's greatest entrepreneurs, [0:21] and you just see this over and over again. [0:23] It's like, "Oh, this time is different." [0:24] Turns out no, this time is not different. [0:26] And Michael's like, "No, no, I actually think this time is different." [0:29] So, I want to start with how you think about building [0:32] your company AI native from today. [0:36] - Yeah, Michael is a legend. [0:37] He must have so many interesting perspectives, especially like [0:40] he's still running the company. [0:41] So, it will be interesting whether [0:43] he applies some of the difference to the current running of the show. [0:48] But for us, in some ways, given we are first-time founders, me and Piotr, [0:53] my co-founder and my best friend of 15 years. [0:56] ElevenLabs is the first venture we started. [0:59] - That's insane. You started in 2022. [1:02] You launched before the first version of ChatGPT, right? [1:06] - We did. - Okay. [1:06] - It was still a year when the topics of the day were crypto and metaverse. [1:11] So 2022 was still a year where everybody was obsessed about those two. [1:15] So it was a perfect time because [1:16] we could actually focus and build a lot on the AI side. [1:20] - What were you doing before you founded this company? [1:23] - I was at Palantir, so I was helping build optimization models [1:26] and bring optimization models to the customers. [1:27] So, NHS during the COVID response on [1:30] how you distribute vaccines across the UK, or working with oil and gas industry [1:36] and figuring out how to optimize their energy work. [1:40] So, a lot of optimization models, [1:42] and then working with customers on actually figuring out [1:45] how you bring that into their production. [1:48] And my co-founder was at Google, [1:49] and he did a lot of the text models for knowledge graph. [1:54] Before that, did research at university around image visual models. [1:59] So, incredible brains. [2:01] The smartest person I know for developing a lot of the research work, [2:04] and I was happily on that intersection of building the product [2:08] and bringing that to the customers. [2:09] So that was before Palantir's BlackRock, building risk models, [2:12] bringing that to customers, and, by background, study mathematics. [2:15] So, good intersection of the things I liked. [2:16] And now at ElevenLabs, that's also what I do. [2:19] And that's also, from the company lens, [2:22] when we started, that was the whole goal of: [2:23] can we combine research and product deployment under one hood? [2:27] On the research side, built all the audio models, [2:30] starting with model to produce speech, text-to-speech model. [2:34] And that was 2022, the first model [2:36] that could finally cross that human-like quality. [2:39] And then over time, now it's entirety of audio models: [2:42] transcription models, localization models, [2:44] orchestration for voice conversations with agents. [2:47] And then on the product side, can we unify that as one platform [2:51] that helps people, businesses transform [2:54] how they would communicate with their audience? [2:55] - Wait, so the first idea, you started with audio. [2:58] - We started with audio. - Why? [2:59] - The actual trigger point comes from where you're from, from Poland. [3:02] Very peculiar thing. [3:03] If you watch a movie in Polish, all the voices, [3:07] whether it's a male voice or whether it's a female voice, [3:10] get narrated with one single character. [3:12] So you have one voice narrating the whole movie. [3:16] All the emotional intonation disappears. [3:19] So we grew up with this. Everybody in Poland grows up with it. [3:22] It's like one person dubbing. [3:24] And then in 2021, we retested or re-experienced [3:30] this still happening in that content. [3:32] And then second thing happened, which was... [3:34] - Wait. It was still happening in movies in 2021 that [3:36] - Yes. - you'd watch in Poland? [3:37] - Yes. [3:38] - Same thing that was happening when you were a kid. [3:39] - Exactly. - Okay. [3:40] - And it's crazy because, of course, it's cheaper, easier to do, quicker timeline, [3:44] but the quality is poor. [3:48] As you can imagine, it's a pretty terrible experience. [3:51] And that was like a trigger point. [3:52] In the future, that experience will be completely different. [3:56] You will have original voice, original emotions, original intonation, [4:01] and actually be able to experience that in that incredible way. [4:04] Which is very timely because just a week ago, [4:07] we released finally a model that is able to do that [4:10] extremely well and finally bring the content from one language to another. [4:13] But maybe more broadly, [4:15] that was also our eye-opening of how we interact with technology, [4:18] how you can interact through technology will change, [4:21] and that will apply the language barrier that will break. [4:23] But also just the general conversations with devices with digital world [4:27] will happen differently across different modalities, across different channels. [4:31] And we wanted to solve it. [4:32] - This is what you thought back in 2022, or what you think today? [4:34] - In 2022, we knew that you will need to unlock the stories, [4:38] the content across the modalities and channels. [4:40] We didn't yet know how quickly the kind of the shift from static [4:44] to interactive will happen. [4:45] We hope this will happen, but we didn't know how quickly. [4:48] - So the first idea was - First idea was dubbing. [4:51] - dub static content that - Exactly. [4:53] - I was watching. I'm watching a movie in Poland. [4:55] I want this to actually feel like I'm watching in my native language. [4:58] - Exactly. [4:59] - That had to be appear like a tiny business, though, right, at the time? [5:02] Like, if that's your initial idea, you weren't going into it thinking, [5:07] "This is going to be a giant business," [5:08] like you're one of the fastest-growing startups today. [5:11] - We actually thought it was a huge business at the time, too. [5:13] Because what we thought is like, if we think about all the content, [5:16] all the stories out there, [5:18] how incredible if they were available in audio, [5:20] and whether it's some of the biggest streaming companies, [5:23] whether that's TV, whether that's the conversations, [5:26] all of those could be actually delivered in the local language. [5:30] So we thought it's actually huge, - Okay. [5:32] - and we still think it's huge. [5:33] And then, if you shift this to even the conversation we are having now, could [5:36] this be, in the future, this version [5:38] where I speak Polish, and you understand me in English, [5:40] or I speak English and you understand me in any language that you want? [5:43] In "The Hitchhiker's Guide to the Galaxy," there's this idea of a Babel fish [5:46] that you put it next to the ear, and you can understand everything [5:48] around you regardless of the language they speak. [5:50] So, we knew that we will get there. But initially, it was like dubbing. [5:54] So I'll give you like a full way of how kind of it progressed. [5:58] Initially, it was dubbing, and then, as we started diving into what we need to do [6:02] to solve dubbing, realized there are three steps in dubbing process. [6:05] There is transcription step, then it's translation step to another language, [6:09] and then you need to regenerate that in another language. [6:14] But the research that existed at the time for each of those steps wasn't very good. [6:18] - Which companies were doing that, the research? [6:21] - There were some good NVIDIA models, open-source models. [6:24] - But wait, it's just an independent researcher. [6:25] It's not any of the big companies? [6:26] - No. I mean, everything was poor. Everything was robotic. [6:29] - Okay. - Everything was pretty robotic. [6:30] It was still like you could immediately tell it was a robotic voice. [6:38] And nothing really crossed that uncanny valley yet. [6:42] There was one good open-source paper, an open-source repo that I'm trying to recall [6:47] that was pretty good, but very unstable and still took so much time to generate. [6:51] That was the best, and NVIDIA had some good research on some of [6:54] those components, especially on the speech-to-text side. [6:57] Translation was okay-ish. [6:59] I mean, DeepL at the time was doing incredibly well, [7:04] and then you had the kind of Google Translate version of that. [7:07] But all of the components to do dub were not good enough yet. [7:12] So that's part one: the research wasn't there. [7:15] And part two: [7:15] when we started testing the dubbing idea with a lot of initially creators, [7:20] the message we are getting back from the creators is like, [7:23] "Great, I would love to get dubbing one day, [7:25] but today I have different problems. [7:26] My problems are: [7:27] I want to post-produce and change the line that was recorded in the wrong way." [7:30] Or, "Before I record my video, I want to be able to narrate the script [7:36] and see how it sounds." [7:37] Or, "Instead of me speaking over a video, could I just have AI speak over that? [7:41] So before I even think about dubbing, can you give me that?" [7:44] And that for us, it was like, okay, before we can solve dubbing, [7:48] let's solve the research component to generate speech [7:51] and make it sound great. [7:53] And on the product side, [7:54] let's actually deliver for people to be able to just narrate content. [7:57] So let's pin down and ignore for a second the language shift. [8:02] Let's just help bringing content alive with audio, with a high quality. [8:07] - I want to tell you about the presenting sponsor of this podcast, ramp. [8:09] I have been reading a lot about SpaceX lately. [8:13] SpaceX is one of the most valuable businesses in the world, [8:15] and one of the main themes in the history of [8:16] SpaceX is constantly attacking and questioning your cost. [8:20] ramp helps many of the most innovative businesses in the world do exactly that. [8:24] The median company running on ramp cuts their expenses by 5%, and one thing [8:29] SpaceX has demonstrated is that a religious dedication to controlling costs [8:33] can help actually increase revenue, because you can pursue opportunities [8:37] you couldn't otherwise. And we see that in the ramp data too. [8:41] The median company running on ramp also grows their revenue by 16%. [8:46] So when you're running your business on ramp, [8:48] and your competitors are not, you have a massive competitive advantage [8:51] that compounds over time. [8:53] ramp is the only platform designed to make your finance team faster and happier. [8:57] Many of the top founders and CEOs I know run their business on ramp. [9:01] I run my business on ramp, and you should too. [9:03] Go to ramp.com to learn how they can help your business save time, [9:07] save money, and grow revenue. [9:09] That is ramp.com. [9:11] Deel is how the best founders turn the world into their talent pool. [9:15] I've been studying how history's greatest founders operate for a decade, [9:18] and one thing they all have in common is they understand [9:20] that recruiting and hiring the very best talent is your most important priority. [9:24] A players recognize other A players, which is why top companies like [9:28] Ramp, Shopify, ElevenLabs, Uber, and DoorDash all use Deel. [9:32] Many of the top founders I know have personally invested in [9:35] Deel after using their product. [9:37] And what they discovered is that [9:38] Deel is the best company in the world at building infrastructure for global hiring. [9:41] Deel will help your business hire, pay, and manage any worker anywhere in [9:45] the world so you can retain the best talent anywhere, [9:48] and spend the rest of your time focusing on what you do best: [9:51] delivering value to your customers. [9:53] The founder of ElevenLabs has a great description of [9:55] the value Deel can give your company. [9:57] He said, [9:58] "We built ElevenLabs to break down language and communication barriers. [10:01] With Deel enabling us to hire and support exceptional talent anywhere, [10:06] we can accelerate our innovation and bring more voices, stories, [10:09] and ideas to every corner of the world." [10:12] Deel is trusted by over 40,000 businesses. [10:15] Learn how they can help your business today by going to deel.com/senra. [10:20] That is deel.com/senra. [10:23] - Let me make sure I'm understanding this correctly. [10:25] You're like, "We're going to dub. [10:26] We're going to reach out to creators first. [10:28] So I'm reaching a couple million people in English. [10:30] Let me see if I can get this guy to say, 'Hey, can I translate it? [10:33] Can we dub this for Polish or in Spanish," and all these other things. [10:36] And then the feedback you're getting is, [10:39] "Yeah, it's kind of nice, but I have many more immediate concerns, [10:42] and can you help me with XYZ"? [10:43] - Exactly. And it kind of was two things at the same time were great, [10:45] because that was also true on the research side, [10:48] where you have these three steps you need to solve in dubbing. [10:50] Nothing in their space is good. [10:52] We need to solve one of those steps ourselves. [10:54] That's where Piotr comes in and is able to create and assemble [10:57] the best research team to solve it, [10:59] and he himself is an incredible researcher to actually bring that to life. [11:02] - Yeah, because I'm friends with MrBeast, and I remember talking to him about [11:06] this a few years ago, [11:07] where he was the largest creator in the world, [11:10] and he used to run separate YouTube channels. [11:12] - Yeah. - And he told me, he's just like, [11:13] "Yeah, well, for the Japanese versions of my videos," [11:17] he doesn't just dub them. [11:19] He'll hire a voice actor that's like the voice is famous in that country. [11:23] - Yeah. One day you should, whether with us or any other company, [11:26] think you should dub your podcast. [11:27] It's so much knowledge that could be brought [11:29] into so many different entrepreneurs worldwide, [11:32] like in Japan, if those conversations were there. [11:36] - It's funny you said Japan, because I was thinking... [11:38] I knew we were going to talk today, [11:40] and I've been thinking about you the last few days, [11:42] and I'm working on this episode with the founder of Honda [11:45] - Mm-hmm. - for my other podcast, Founders. [11:47] And what's surprising about that is it's not Honda the cars at this point, [11:53] when this guy's alive. [11:54] He created the most successful mass-produced motor vehicle in history, [11:58] which is the Honda Super Cub. [12:00] And the way they describe their company, he's like, [12:03] "We're just a research lab for engines." [12:05] Literally. That's what he's like, "All I do is think about engines all day." [12:09] He actually separated out and made the Honda R&D a separate company [12:13] because he thought it was so important, [12:14] and then they were just funded by a percentage of sales, [12:16] but then they turned over all the research to the manufacturing division. [12:19] And he's just like, [12:19] "If you just have a research division, and you only tell them to experiment, [12:23] and you have staff with engineers and researchers," he's like, [12:25] "They're going to constantly invent new things, [12:27] and then if you're a manufacturer, [12:29] - Mm-hmm. [12:29] - you can figure out how to apply those to products." [12:31] - Hmm. - I would never think... [12:33] You would think, "Oh, they're a car manufacturer." [12:34] He says, "No, this is a research team." [12:36] - To be honest, and you started with this question, this is not too dissimilar from [12:40] how I think about how we run ElevenLabs today. [12:42] - I know. That's why I'm telling you this. [12:44] - It's very much like there is a focused research lab; [12:48] there's research engineering, and bringing [12:50] that to the product work, and then we have a lot of [12:53] small teams running after specific product problems that we can solve, [12:57] and then, of course, the way to go to market [12:59] and deployment of how you bring that to the customers worldwide. [13:02] But the general philosophy is like a lot of small teams, [13:05] usually less than 10 people, having flexibility, autonomy to just run ahead [13:10] and apply their best judgment in what we can solve for the customer. [13:14] - Okay, you need to say more about this, though, [13:15] because this is where I'm still a little confused by this. [13:19] Somebody doesn't know who you are; they just met you for the first time. [13:22] They don't know what ElevenLabs is. [13:24] Describe to that person how you view your own company. [13:27] It's like a research lab. [13:28] Is that the word you're going to use? How would you describe this? [13:30] - I would introduce consistently the company as a combination of research [13:35] and product deployment. [13:37] Research is frontier audio models, [13:40] text-to-speech, speech-to-text, orchestration. [13:42] Product is one platform that helps companies transform [13:46] how they communicate with the world around them, [13:48] and that can be marketing with helping them tell a story, [13:51] like Ramp creating their Super Bowl ad. [13:55] It can be support, working with Deutsche Telekom on creating voice agents [13:59] for their call center, or it can be sales, helping on inbound sales qualification [14:03] to make sure that streams through. [14:06] Or even wider operations and working with the government of Poland [14:09] to help create an agent that can take a healthcare appointment, [14:12] follow up with a patient to remind them about the appointment, [14:15] ask how they are feeling. [14:16] And kind of one combination between those is building the research, [14:20] building the frontier of all voice, of all audio, [14:23] and on the product, doing the part of applying [14:26] that audio, combining that with knowledge, [14:28] combining that with the creative work to then allow companies [14:33] and people to change how they communicate. [14:34] - And you are solely focused on audio? [14:37] - On the research side, solely focused on audio. [14:39] Product: we combine the best of audio with integrations, knowledge, [14:43] LLMs to help deliver for us all. [14:46] - Explain why you think it's so important [14:48] to be solely focused on audio on the research side. [14:51] - We think we have incredible talent to be able to go after that. [14:55] The focus is so important and effectively building [14:58] the best architecture for those audio models. [15:01] We think audio... [15:02] There is a good combination of not only science, [15:04] but also the art that you need to solve. [15:06] It's a little bit subjective. [15:08] The voices that you produce will be subjective. [15:10] So you really need to get it right. [15:11] But ultimately, as you think about AI models, [15:14] there's data, compute, architecture that you need. [15:18] On audio, we think a lot of the problems [15:20] that still exist are on the architecture side. [15:23] We want to be solely focused on solving those architecture problems, [15:27] so you can actually get the best quality out there. [15:30] A lot of the team, a lot of the people that work on our side are [15:33] the best audio researchers in the world. [15:36] They're also excited by the premise of being able to continue deploying [15:40] the best frontier audio models. [15:42] - And you feel you have technology on [15:44] the audio side that no one else in the world has? [15:46] - We think so, yes. [15:47] - This is a weird analogy. [15:49] I'm so glad I'm reading this book at the same time we're talking, [15:50] because, like for Honda, they kept trying to get him to diversify. [15:54] He's got other products. [15:55] Obviously, he builds a bunch of things with engines, [15:58] but he's like, "Does that product have an engine"? [16:00] And they're like, "No." [16:00] He's like, "Then I'm not building that product." [16:02] He's like, "I focus on engines, [16:05] and sometimes it's on two wheels and four wheels and everything else," [16:07] but he's like, "Just engines." [16:08] It's very similar to what you're saying about audio. [16:10] And he also said that he felt he had the best engine technology in the world [16:12] and no one else could replicate what they did. [16:15] - There is a common question as we think internally of, like, it's 100% true of, [16:20] well, how you describe it now, too. [16:22] And when we think about a new product, the big question is: [16:25] do we think we have a unique advantage by [16:27] applying our audio models in that product experience? [16:30] If the product experience doesn't have a big bottleneck in that audio [16:35] and the voice communication side, then it's not our forte. [16:39] - Explain a situation where you realized that [16:42] you shouldn't go after that opportunity. [16:43] Could you give me an example of what you're describing? [16:46] - The example that I'll give you, two slightly different spaces. [16:51] One is, for a long time, [16:57] and I think it's still true, [16:59] the ideal version in the audio space, [17:02] if you've created a marketing campaign, is how that combines with a lot of the image [17:06] and video work to deliver a better content. [17:08] Two years ago, we would have first tried to see whether we can help people create [17:12] effectively lip-dubbing or avatars for their content. [17:17] So, let's say you switch from one image to another; [17:19] you need to move the lips, or let's say you want to narrate something; [17:22] you create an avatar. [17:23] That was roughly two years ago. [17:26] A lot of the models that existed on that site just weren't good enough, [17:30] and deploying a product in this space would be such a defocus [17:33] and such a shift for the company that it didn't have the audio as a superpower. [17:38] It still had a bottleneck of the quality of avatars, quality of a lot of the... [17:42] - Because the problem to solve there is video, not audio. [17:44] - Exactly. [17:44] - Okay. - So this was still poor. [17:47] So even if you had the best audio applied into that work, [17:50] still the experience wouldn't be very good. [17:52] So we decided to effectively pause any of that effort. [17:57] Fast forward, today is, of course, shifting now. [17:59] I think there's a good set of open-source models now that exist in [18:01] that space where the combination of audio, image, [18:04] and video can actually deliver that result. [18:06] So we are revisiting that today, and the recent London ads engine of bringing [18:10] and bringing that internationally [18:12] and shifting things in the video, perfect use case of that. [18:15] But still, a conscious decision [18:17] we are taking is not creating the model from scratch, [18:20] so like true text-to-video or VO-free type models that exist. [18:25] We want to be at that intersection [18:26] where we know the audio can give you the unique advantage, [18:29] which is usually you already have an existing asset, [18:31] you need to modify that asset, add that audio component, and bring it home. [18:36] - Do you feel this intense focus or relentless focus on audio is [18:41] kind of like defense against the larger labs? [18:45] - For sure. - Do you talk about this in the company? [18:47] - We do. The big part is, our focus is audio on the research side, and that's [18:51] where we want to win. [18:53] I think that we also talk about this: in the long, long term, [18:57] that advantage that we have today will hopefully still be there, [19:01] but it might not be as big on just the pure research. [19:04] That's why the other components that's why the product is so important, [19:07] and that's why the ecosystem [19:08] is so important that we build around that product. [19:11] And maybe just to explain what I mean by the ecosystem: [19:13] of course, there's the brand and trust that you built, [19:16] but there are also other parts that you can build. [19:18] And in our case, this was investing into effectively, a marketplace model [19:22] where people can create an asset, we authenticate it, and then you can share it [19:26] and earn compensation as a result. [19:28] So, like trying to create a completely different model for how that works. [19:32] - Give me details about that because I don't know what this is. [19:34] - Yeah. - Like an example is voices. [19:36] - Okay. [19:36] - So we did it with voices where people can create their voice; we authenticate it [19:40] then you can share your voice, your AI voice, and passively, as your voice is [19:44] being used, you earn compensation. [19:46] We have 20,000 voices today on the marketplace. [19:50] And new people that come in now have a selection of [19:53] different accents, different styles, different languages, of course, [19:58] different age, gender that you can pick. [20:01] Every time you use it, that other person gets paid. [20:04] - Do you, as a company, make these voices, [20:06] any of these voices, or is it just third parties? [20:08] We, of course, created a system for people to do that. [20:11] And some of them, when we see pockets that are missing, [20:14] we'll ourselves try to go after [20:16] and find people to fill those pockets that are missing [20:19] and create those voices for the wider ecosystem to benefit. [20:22] - No one knows about ElevenReader. [20:24] You have to do a better job. [20:25] I was listening to your episode with John Collison, [20:27] who I love, and I don't think he even knew. [20:30] He kept saying, he's like, "Why doesn't this exist"? [20:32] - Yeah. [20:32] - "There's an app." [20:33] And then the follow-up, I still talk to him and I was like, [20:35] "No, dude, I use that app. [20:36] What you're asking for, John, he already has it." [20:39] You need to show it to him on your phone [20:40] when you're recording the podcast with him. [20:42] So I use George; that's the voice I... [20:45] In fact, we do all these research reports for the people that come on. [20:48] - Thanks for being a user. That's amazing. [20:49] - No, of course. [20:50] - It's a fantastic product, and it turns every document into [20:54] essentially a podcast that I can listen to when my eyes are busy. [20:57] But the funny thing is, [20:59] we do these research dossiers about everybody that comes on, [21:01] and so I put the one about you into ElevenReader, [21:04] so I'm listening to you about you in your own product. [21:08] But in George's case, is that coming from you guys? [21:12] Is somebody else getting paid when I'm [21:14] - Yeah. - listening to George? [21:15] Somebody else is getting paid. - Okay. [21:16] - It's a great voice actor that worked with us now for a long time, [21:20] and every time you listen, he gets paid. [21:23] - Do you understand? [21:24] When you sat down, I was like, one, I keep running into you everywhere, [21:29] and so that's why I was like, [21:30] "Dude, we got to just do a podcast together." [21:31] But you're so confusing to me, in a great way. [21:32] This is not like a negative thing, because I was like, [21:36] I don't even know how to describe your company. [21:38] You're doing all this research, [21:39] but then you have all these other different products. [21:41] This is very unusual, especially for somebody that's so super focused. [21:44] - But it is very fair, and I think we are seeing that everywhere across AI [21:49] companies of like the model becomes a platform, becomes application. [21:53] But for us, the unifying theme across all of them [21:55] is that angle of communication. [21:57] We think how you communicate as a company, how you think about content communicates, [22:02] all of that is changing, and we want to be at the intersection of that. [22:05] We build the best models to help you do that, [22:07] and then build a platform that effectively delivers that content, [22:10] delivers the knowledge, delivers the conversation in a completely new way. [22:13] Of course, there's just so many different applications. [22:15] There's ElevenReader that lets you enjoy content a completely new way. [22:19] - So, research, platform, application's on that platform. [22:22] - Exactly. - Okay. [22:24] - And we explicitly are not planning to touch [22:28] any of the intelligence or knowledge work or coding. [22:31] Not our strength, not our domain. [22:33] But we are... - That's a vicious battle, too. [22:36] - A very vicious battle. [22:37] Lots of great companies in there. [22:39] But what we would love to be, is if you think about the next three years, [22:43] or however many years, there will be probably three platforms [22:46] that you set all the interactions on those platforms. [22:49] We want to be the leading platform for those interactions, [22:52] for that communication. [22:53] - Wait, explain that. [22:54] - The whole way... [22:56] And that kind of builds on that theme. [22:57] How you engage with content, [22:59] how you engage with the company, will change. [23:01] We want to help people, companies have one place, one platform [23:08] where you can set it up, you can set up your integrations, [23:10] you can bring your knowledge, you can bring your brand, [23:12] you can bring your assets from the company [23:16] and then deploy that for the entirety of customer journey. [23:18] Whether that's in the marketing example, where you tell that story through content, [23:23] capture that knowledge back about the user; whether that's in sales, [23:26] when you try to engage and bring your product to the customers; [23:30] whether you are supporting that across the customer journey. [23:33] Or maybe in a simpler way: [23:34] as you are trying to understand the customer journey, [23:38] every on-brand interaction that that journey has, we would love to be [23:42] able to capture and help you make that better [23:44] - Okay. - for the customer. [23:45] - So let's make this concrete because [23:46] I still want to be able to wrap my head around this. [23:49] Let's take your most deeply integrated customer, [23:53] - Yes. - right? [23:54] And say it's Goldman Sachs; [23:55] I'm making this up, and this is how they use [23:58] this product, all the different products they use. [24:00] Who is the company, or one of the companies, [24:02] that is most deeply integrated with your company, [24:04] and explain how all the different suite of products that [24:06] they're using are powered by your technology. [24:09] - Deutsche Telekom is a great example. [24:11] They will use a lot of the audio work [24:14] to create a podcast in their app, in the Magenta. [24:16] They will do that for the ads, [24:17] so they can create ads and distribute that to their audience. [24:21] They will use a lot of our ElevenCreative work to be able to do that. [24:25] Then they will use our ElevenAgents work for in-call center, [24:31] where you call in, you want the help, you get that help through the voice agent [24:35] integrated with the knowledge from Deutsche Telekom. [24:38] So you can make sure that if somebody's calling in to learn about the product, [24:42] they get that information very quickly. [24:44] If they are calling to get support on [24:46] how to get a refund or the recent billing request, they can get that help. [24:51] And then, more recently, [24:53] they even deployed an agent inside of the network. [24:55] So, if you are a T-Mobile subscriber, [24:58] if you call, you can ask an agent to join the call and help you out, [25:01] schedule a booking or real-time translate your conversation to the other person. [25:05] And they'll use our effectively combination of agents' work and bringing [25:11] that real-time dubbing to be able to communicate in another language. [25:15] And across all of that, all of that information is captured back, [25:18] so you understand how customers engage, [25:21] in this case, with the full spectrum of the journey, marketing, [25:27] the support, and then the wider operations, even sales, [25:31] and we help them do all of that. [25:32] - Okay. So, in a situation like that, because I think you mentioned, [25:36] obviously a huge increase in your revenue. [25:38] It's growing really fast. [25:40] It's like you're targeting bigger companies, right? [25:42] What was the first product Deutsche Bank used from you? [25:45] - Deutsche Telekom. - Deutsche Telekom, sorry. [25:47] - Yeah. The first was marketing. [25:49] So it was two years ago. [25:56] Two years ago, still most of the agents weren't really very good. [25:59] They weren't very reliable; they weren't very quick. [26:02] So marketing was the most obvious one. [26:04] So they started with that. They deployed that. [26:06] The quality of the content was great. [26:11] So people could engage with the podcast. [26:12] Effectively, the use case that you mentioned of, like, [26:15] bringing your notes and then being able to read them out loud. [26:21] - Mm-hmm. [26:21] - They were just creating that daily for all the customers, [26:23] so they could read about [26:24] what's happening in the world through the podcast with ElevenLabs' voices. [26:28] So that was the first use case. [26:30] Then, of course, [26:31] support is a combination of voice but also the integrations. [26:34] - Let's talk about how you expanded them from one product to the next. [26:38] Like, how do you actually do that? [26:39] - In that case, the main... [26:42] Are you asking mostly like, [26:43] how do we partner with them to help them bring that product alive or? [26:46] - Yeah, they start with one, [26:47] but you have 10 different products you could sell them. [26:50] - Yeah. [26:51] - So how do you go from one to two, and then two to five, and so on and so forth? [26:53] - Got it. Yeah, I think the main thing is one, [26:59] you, of course, deploy the first one. [27:01] You make sure that there is value behind that. [27:03] So across any customer engagement, [27:06] you try to make sure that we not only prove concepts, [27:08] we prove the impact, prove the value, [27:10] and only after that we try to get the companies working with us [27:12] at the broader scale. [27:14] So, we proved the impact on the marketing side. [27:16] To get the support going, [27:18] what you actually need to do is not only the audio; [27:21] you need to build the integrations. [27:22] So you need to connect it with all the CRM systems [27:26] that Deutsche Telekom works with. [27:28] You need to work on integrations with the output. [27:31] So how do you connect it to the phone system, [27:33] whether it's SIP trunking or Twilio? [27:34] How do you make sure that that connection exists? [27:36] So you would spend a lot of time on integrations, [27:38] making sure that their logic is respected on [27:40] how, when you do pick up the phone call, [27:43] the agent behaves the way you want it to behave. [27:45] So here, our forward deployed engineers would partner with the team, [27:48] spend the time, and born in this case in Germany, [27:52] working through side by side together on the integration, [27:55] on just being able, making sure that the agent follows [27:58] the flow that you want, has the knowledge that it should have. [28:01] And then, [28:02] in that step, the hardest thing in any voice agent work is actual deployment of [28:08] how you actually test that it works. [28:09] So then you need to trial that with... [28:12] We initially simulated that. [28:15] You verify the calls, then you do it at a smaller scale. [28:18] - And you're doing this with FDEs? [28:19] - Yes. We do all that with FDEs. [28:22] Then we deploy, and then of course, we scale over time. [28:24] - When did you realize that a huge path to greatly increasing [28:28] your revenue was doing FDEs? [28:29] - And before I answer this, and of course, the last part is, as you deploy, [28:33] the job isn't done. [28:34] You still want to continue, evaluate, monitor, [28:36] and refine the behavior over time. [28:39] And that applies across all of the use cases. [28:42] And FDEs for that are great too. [28:44] So, I used to be at Palantir, so it always felt like a big thing of [28:49] how we want to work with customers. [28:51] And it's like, almost I don't like the word customers because, in many ways, [28:54] I feel like all of them are partners, [28:56] where you are working together on the same problem and trying to solve that. [29:00] So a lot of the philosophy of the FDEs was there from the start, [29:04] to make it specific: working with the partners side by side [29:08] on their problem in their office. [29:09] - Hold on. This is actually really interesting. [29:11] So, tell me what you observed that was working at Palantir, [29:14] where you're like, "Ooh, if I start my company, [29:17] I want to do that too," and why. [29:19] - I think the most incredible thing was that [29:22] you were meant to be on the same side of the customer, [29:26] obsessed with their problem, and try to understand them [29:30] deeply from the beginning. [29:32] So, I used to work at BlackRock before Palantir. [29:35] In BlackRock, when I joined, given the compliance security, [29:39] the first month or two, [29:41] when you send an email, it gets verified by your team. [29:45] It gets trained before you send it externally. [29:48] So it's like a pretty detailed flow of making sure that it's good. [29:53] In Palantir, when I joined, after the onboarding in the first month, [29:57] I'm like, "Okay, you now need to work and understand the customer. [30:00] You're flying to Aberdeen, to North Sea, to work side by side with them, [30:05] understand what's happening, and then bring it back." [30:07] And that was a complete shock for me, [30:09] and the kind of the approach and culture of like, [30:11] I have never almost interacted with a customer on the BlackRock side. [30:15] And now here I am in the first weeks. [30:17] I'm meant to go there and be next to them, which was crazy. [30:21] I thought it was great. [30:21] I thought it was that mentality, and every time, kind of, [30:26] you are going to be there on the, I'll call it front line, [30:29] loosely on the front line, to work with them and bring that knowledge back, [30:35] understand what is actually the problem, [30:36] what's a fixable problem, and then actually fix it. [30:40] And that mentality is so true now at ElevenLabs too, where all of our FDEs, [30:45] our go-to-market team too, is going to obsess over [30:47] how can I actually be there with the customer, [30:51] understand the problem, and work with that. [30:53] There were other companies that I think were great. [30:54] The small teams, they had usually small deployment teams that [30:57] worked on a lot of that. [30:59] The best idea wins concept was very true, where in that small team, [31:03] you very quickly assembled the best knowledge of [31:06] what you think should be done for that customer, [31:08] and you had power to then actually enact on it. [31:10] - Define small. [31:10] - About five people was considered in Palantir big, [31:14] and in ElevenLabs it's going to be also relatively big. [31:16] I'll tell you something more, though, on the FDE side, because, [31:20] of course, now you probably see every company doing FDEs. [31:25] And in our case, FDEs are part of the product team. [31:27] They are not part of the go-to-market team; [31:29] they are part of the product team. [31:31] They are deeply embedded on understanding what's the roadmap of the product. [31:34] But there's a second reason, which is: [31:36] one, you want all the FDEs to actually [31:39] solve the customer problem and stretch your product in that direction. [31:42] But the second thing you want them to do is [31:44] bring any of that knowledge back to the product, [31:47] so the product becomes better for the next generation of companies [31:51] building on top of it. [31:53] And I feel like the second part is frequently missed, [31:58] where it's like, first, okay, FDEs are doing effectively [32:02] a lot of the hard integration work to solve the problem. [32:06] But if you don't do the second step of [32:08] how you actually learn the expertise and bring it back to the product, [32:11] then it's effectively just services, or just a... [32:14] - Wait, say more about that second part. [32:16] So you're gleaning information through trial and error, [32:20] working closely with them. [32:21] Then you can take that back and spread [32:23] that knowledge across the other tens of thousands of customers that you have? [32:26] - Yes. - Is that what you're talking about? [32:27] - Exactly. - Okay. [32:28] - Exactly. And it's a product knowledge of how you operate with... [32:31] You know, there are simple things. [32:32] Let's say you build an integration. [32:34] How do I now have that integration available to everyone? [32:37] That's a simpler version of that. [32:39] But let's say you are working in healthcare space. [32:42] In healthcare space, if you deploy the work, [32:46] then you want to optimize the product experience for the right... [32:53] If you are now starting as a second customer, [32:55] you want to make sure that you have... [32:57] - So wait a minute. [32:58] You're using your customer base as almost like R&D, [33:01] another form of R&D. [33:03] - You are definitely accelerating your R&D through understanding the domain that [33:07] you're working with. [33:08] And it's kind of... - Because their problems are not... [33:10] Very rarely a company's problems are unique to that company. [33:13] - That's right. [33:14] And the beauty of that is like, kind of everybody benefits, because [33:17] you work with one customer, you learn, [33:19] you bring that into the product experience. [33:20] You learn with another one, you learn, [33:22] you bring it back to the product experience. [33:24] Both of those customers benefited from having [33:26] the product optimized for the better work. [33:29] The domain expertise they have is still theirs. [33:31] They can win based on the domain expertise. [33:33] But the product experience of [33:34] how you build based on that domain expertise [33:37] can be abstracted, can be reusable. [33:39] - What else did you learn working at Palantir? [33:41] - They also were very much a no-title organization, [33:46] which we carry over at ElevenLabs, [33:48] which does help in that best idea wins approach, [33:52] where people do feel... [33:53] I feel like I'm mimicking you on that chair. [33:56] - I thought the same thing. [33:57] I was like, "You don't have to, if you don't want to." [33:59] - No, it's a complete coincidence. [34:02] - We're leaving that part in. [34:05] - No titles, relatively flat organization, very few layers. [34:08] There's like, between me and the C-suite, [34:13] I worked there before the DPO, it was four, five steps or less. [34:18] So you always felt like a proximity between... [34:21] Yeah, I think it was like four or three, actually. [34:22] So it was a good proximity of being able to work together. [34:26] At ElevenLabs, we have a couple five today, maximum depth of [34:33] how many layers there should be. [34:35] And hopefully, over time, actually, there will be fewer, not more layers, [34:39] if AI helps you run the organization the way we would like to run. [34:42] - Well, say more about that. [34:43] What can you not do today in the way you run your organization [34:47] that you hope AI can change in the future? [34:50] - I think there is, like, always you want the information from [34:57] the kind of person working as closely to the problem. [34:59] So, let's say you are developing a product. [35:02] You want to speak with the engineer who actually develops that product [35:04] rather than their manager. [35:06] If it's a client conversation, similarly, you want to understand [35:09] what the client is saying from the person on the ground [35:12] rather than a manager summarizing that information and giving that to you. [35:16] And I think that information flow will change with AI, [35:18] where you will have access to everything that's happening [35:22] in a better granularity than you could ever have before, [35:25] because it summarizes that back and forth. [35:27] Second, I think, in general, roughly at ElevenLabs, [35:32] most people will have close to ten direct reports. [35:36] So pretty wide-set, which helps us build that small team approach that we have. [35:42] And that, too, only works if you can amplify a lot of what's happening [35:47] across all the teams, all the people that you work with, [35:53] to summarize information about what's happening in their teams, [35:55] how this is performing, and what are some of the gaps. [35:57] That definitely helps and helps shift it from reactive to proactive, I think, too. [36:01] What I mean by this is that, frequently in the past, I think you would [36:05] rely on specific individuals surfacing the information to you. [36:09] Now, as we think about ElevenLabs and how we run, [36:12] it's like you can summarize all that's happening, [36:15] all the data of what's working, what's not, and get that signal. [36:18] I get that signal, and I can proactively engage where I think it's not working, [36:22] which helps. It helps, of course, [36:24] everybody across the company, because you have that same ability wherever you are. [36:28] I think that maybe the last part, which is very related to those two, [36:31] is why it is even possible today. [36:34] We took a choice to be extremely transparent [36:36] with a lot of the data that we have, [36:40] and a lot of the docs that we write. [36:41] So, everybody has almost access to all the docs that are in the company, [36:46] which ultimately helps you create a system [36:48] where you can actually tap into that knowledge. [36:51] - Okay, so I want you to say more about this because I think... [36:53] This is one of the things I want to talk to you about, [36:54] because you're relatively young. [36:56] This is your first company. [36:58] Your first company you built, you got into AI right away, [37:01] and then you were perfectly positioned for this huge explosion. [37:04] Obviously, people have been talking about AI for 80 years, whatever, [37:06] but you're at the right place, right time with,I think, the right set of skills [37:10] based on what the company has built so far. [37:11] So, how different do you think, in the future, [37:16] the way you run your organization can be than what it is today? [37:19] - I think the small-team approach will definitely be there [37:22] across many of the companies that I believe will be created in the future. [37:27] It also helps with a very different piece, which is like, [37:29] if you just think about adopting AI technology. [37:32] In our case, it doesn't have to be a top-down mandate of, like, [37:35] you need to use this technology to get better, [37:38] and then the teams enable others. [37:40] Given it's relatively small and independent, people bottoms up, [37:45] adopt what they think is best, and are enabled to run with it straight away. [37:49] So, that helped a lot in that sense. [37:52] I'll give it to others, which we believe strongly on one. [37:54] In a lot of the small teams that we have, [37:58] even non-technical teams, we bring engineering talent to those teams. [38:02] So, our talent team, our ops team, our legal team, [38:05] all of that will have engineers who help both [38:08] automate some of the work but also elevate everybody else in how they are using AI. [38:14] I think that will be a bigger pattern in the future across the companies, too, [38:18] where I think increasingly some of the biggest and smartest companies [38:21] will try to infuse all teams with engineering resources [38:25] so they can get smarter. [38:27] - Is there any function that is centralized at ElevenLabs [38:30] that all these small individual teams can then access? [38:35] - Yes. - I spent some time with Luca Ferrari [38:36] of Bending Spoons. - Yeah. [38:37] - And he had this wild idea, wild idea to me, where he's just like, [38:41] well, HR is actually really important. [38:43] He thinks like the fact that tech people are dismissive of it is ridiculous. [38:47] And he's like, "But my HR is 50 people, and they're all engineers." [38:51] It's like one centralized HR that every single other... [38:55] because he owns I don't know how many companies, they all utilize it. [38:58] But it's just like 50, I think, 50 engineers in HR, [39:02] and then all the different structures, teams, and companies he has [39:05] tap into that for that resource. [39:08] - Of course, there are centralized functions. [39:10] Legal is a good example, of course, of how we are running, [39:13] that needs to be a very central enablement of how we help people [39:18] across go-to-market and other teams, [39:19] how they are kind of learning the craft of how to sell ElevenLabs. [39:23] And as you said, there are so many different products. [39:26] You need to be very particular about how you deliver that to the specific customer. [39:31] So, it's not confusing the customer, [39:34] and what we are offering or what we do. [39:36] So, yes. So, there's a good amount of those functions. [39:40] But they all too, similar to Luca's approach, [39:43] all of them will have a good amount of engineering in them [39:46] and figuring out how you scale that operation [39:48] so it's not becoming hundreds of people in all locations, [39:52] but how you can use the few people that you have, [39:54] and really bring that knowledge everywhere. [39:55] One of the biggest ones for us is probably around RevOps, like how you... [40:00] Or like the revenue engineering effectively function. [40:03] We have so many tools that help you get the knowledge, [40:08] whether it's code during the conversations, transcribe, bring it back, [40:13] make it easier for you to fill on a field, [40:16] so you capture that knowledge, and then don't have to spend manual time. [40:19] And then of course, increasingly we dogfood a lot of our own work. [40:22] So, trying to figure out how we can create AI agents [40:24] to help replace part of that work, [40:26] but still operate within the team. [40:28] So, a good example is AI SDR. [40:30] People going on the website today, on ElevenLabs, [40:33] you can fill in the dropdown form, [40:35] and leave information about your company. [40:37] But you can also speak with a voice agent [40:39] and leave that in a quicker and different way. [40:43] So, we've seen a lot of people go through that flow. [40:45] And then, interestingly, two things happen. [40:47] One, people are making the experience easier; [40:51] they enjoy it more, [40:52] but two, they leave a lot more information [40:54] than they ever would for the dropdown form. [40:55] They tell us so much about the problems, about the wider set of use cases, [40:59] so we can connect them better. [41:00] - That's really interesting. [41:01] So, because they're speaking and not typing, you get more information. [41:04] - 100%. [41:06] I mean, it's, of course, a quicker way of doing it but also [41:09] people just feel more at ease instead of going through this manual dropdown form. [41:13] I feel it's just like a better way on leaving the information. [41:19] But now, because we think it's going to be such a big part where [41:23] everybody, every company will have their revenue engineering function, [41:27] but in our case, we want that kind of AI SDR function to be part of the company. [41:32] We want it to help everybody across wherever you are. [41:36] We are a global company, so it's central, [41:39] and then deployed to each local team, [41:42] so each local team can refine it for the local nuance, [41:46] but develop, of course, centrally. [41:48] - I found one of my all-time favorite quotes [41:50] when I was reading the book "Zero to One." [41:51] The quote says: "The single most powerful pattern I have noticed [41:55] is that successful people find value in unexpected places. [41:58] And they do this by thinking about business from first principles [42:02] instead of formulas." [42:03] That is exactly what AppLovin has done with their advertising platform. [42:07] AppLovin connects you with over a billion potential new customers [42:11] inside mobile games. [42:12] AppLovin allows you to capture undivided attention. [42:16] AppLovin ads are full-screen video ads [42:18] that are watched for an average of 35 seconds. [42:21] That is retention that blows other ad platforms out of the water, [42:25] and you can launch on AppLovin in minutes. [42:28] You set the goal, and AppLovin achieves it. [42:30] There's no complex setup, no expertise needed, [42:33] and AppLovin scales quickly. [42:36] They can put your ads in front of over a billion potential customers. [42:40] Other businesses have seen immediate results, [42:43] have scaled to hundreds of thousands of dollars of spend per day, [42:46] and increased their revenue by millions. [42:49] So, you want to get started quickly before all of your competitors are on AppLovin, [42:53] and you can do that by going to applovin.com. [42:56] That's applovin.com. [42:59] Do you still think of ElevenLabs as a single company just with... [43:02] How many different products do you guys have right now? [43:05] - We have three kinds of core product lines, [43:09] and then we have an additional three that are developing bets. [43:14] - But how many different applications? [43:17] - Well, within each of those we'll have, I'd say... [43:19] No, but let's say six are the kind of core and then within that, you can do, [43:22] of course, maybe like 20, 30 different things. [43:26] - Yeah, it seems like you have a ton. [43:27] - Yes. [43:28] - Like, ElevenReader, for example. [43:30] - Yes, but it's like ElevenReader, or we have ElevenProductions, [43:33] which helps you with the human loop aspect to correct the content and localize [43:37] that to another language. - But where's all the revenue coming from? [43:39] - The biggest lines are on the Agents' side and the Creatives' side. [43:43] So, the number of use cases that deploy conversational agents [43:47] is just skyrocketing today for us. [43:50] - And are you seeing a specific industry that's adopting them faster or no? [43:55] - I think the quickest today for us is fintech. [43:58] Super quick. [43:59] - Like Revolut? - Revolut, exactly. Klarna, PagBank, [44:03] like all of those companies, are just moving at another speed, [44:07] or Customers Bank in the US here. [44:10] Then the healthcare, retail e-commerce, and telcos are four of the biggest ones. [44:18] Fintech is moving the quickest. Of course, it's slightly different [44:21] because of the regulatory aspects. [44:23] But then I think the healthcare, telcos are kind of the second, [44:26] and retail e-commerce has just started this year, going through the wave. [44:30] - I want to ask you one more time. [44:31] I just want to go back to this in case we missed anything. [44:33] Is there anything else that you learned working at Palantir [44:36] that we didn't talk about that you think is valuable? [44:38] - Well, this is a stretch, and I'm biased, [44:41] but in Palantir, frequently, there was this concept of [44:46] how you need a little bit of understanding [44:52] of the art to do your job well, [44:54] or people need to effectively try to be artists, [45:00] even if we aren't. [45:03] One of the phrases that was used was "artist colony," [45:05] and if that was ever publicly said. [45:07] I never fully appreciated the strength of that, [45:10] but I do feel a lot more about this now, [45:13] as we kind of intersect that AI and creative space in many ways. [45:17] Whether it's building the audio research models, [45:19] there's a lot of nuance in how that's delivered. [45:21] Like, every voice that is delivered, like, George and ElevenReader. [45:27] Everybody will have their own preference, [45:29] and how you make sure you capture those voices, [45:32] how you deliver those preferences, is a tricky challenge. [45:34] And then two, as you work with some of the brands that are trying to define [45:38] how they communicate with the world, that too requires a lot of the art in that [45:43] and we're trying to bring, of course, a lot of the people to combine that. [45:46] But blending that art and science, Palantir definitely tried to do. [45:51] I think I will let others judge to do it successfully, but we also will try to do, [45:57] and hopefully are doing good steps in that direction, [45:59] whether it's the voice marketplace that we talked about, [46:02] whether it's having a lot of creatives in the company, [46:04] building projects with us, or around us, [46:07] or the kind of for-the-point new engineers or for-the-point creatives [46:10] almost to work with the customers. [46:12] I think that blend will be increasingly important. [46:14] And everybody talks about taste now, how taste will be defining front of AI, [46:19] and behind the buzzword, I think there is a lot of truth to that, where, [46:23] yes, increasingly, everybody will be able to create everything, [46:26] and what defines above good product experience, good deployment experience [46:31] will depend on how the design language sounds, [46:36] how tasteful it is, how you bring that across. [46:38] So, I think Palantir tried to do it. We are trying to do it too. [46:41] - Yeah, I think what you're getting at is there's this guy named Edwin Land, [46:44] he's the founder of Polaroid. [46:45] I discovered him because he was Steve Jobs' hero, [46:48] and Steve Jobs was talking about... - Oh, yes. Yeah, I remember it. Yeah. [46:51] - I've talked to him about it on Founder's episodes [46:52] over and over again. - Yeah. [46:53] - I should get a framed picture of Edwin Land [46:55] and put him in the studio or something. [46:56] But he was the one who invented the idea that Steve Jobs used, [47:01] where he's like, "I want to build a company at the intersection [47:02] of technology and liberal arts." [47:04] And the reason he just came to mind is when you brought up taste. [47:06] He has this great quote where he says, "Taste is as rare as a unicorn." [47:10] And so, it's like people talk about it all the time, [47:12] but it is definitely a limiting factor, [47:14] and I think it will always be a limiting factor. [47:17] But I love this idea. [47:18] I think this is my issue with a lot of why I don't live in San Francisco. [47:22] Like, I drop in there, we do a bunch of recordings, [47:24] obviously have a bunch of friends who are tech founders, [47:26] but it's just like, I feel the new crop of tech founders is like, they lack humanity. [47:33] Where, one thing I liked about Steve Jobs or Edwin Land or the founder of Honda, [47:37] they would say over and over again, it's like, "I'm just inventing technology [47:40] to enhance humanity." - Yeah. [47:42] - Where I think some of these people are just like, [47:43] "I'm inventing technology to replace it." [47:45] So, this leads me to another thing that I wanted to ask you about, [47:47] since you worked at Palantir. [47:48] You've seen Alex Karp be one of the only people, [47:52] and he's done it for a year and a half, and now it's really, like... [47:54] I think his perspective has caught on, where he's just like, [47:57] "If you're running these labs and you keep going on TV or giving interviews [48:00] talking about 'I'm building nuclear weapons-grade technology, [48:04] and it's going to take everybody's job...'" [48:06] I see you at too many conferences, which I want to ask you about. [48:09] - No, I haven't been there. [48:10] - Okay. So, I was just there, I was only there for a few hours, [48:12] and Karp was the one who spoke first, [48:15] and he's just like, "What do you think is going to happen?" [48:17] He's like, "Your technology's going to get nationalized." [48:20] Do you have any opinion on his perspective on this? [48:23] - Not directly. [48:25] So, of course, we've seen what happened with the Anthropic case recently, [48:28] given that we, and most of our team is in Europe, [48:32] spent a lot of time with the European teams and governments [48:34] on how we think about building sovereignty and how important this will be. [48:39] So, no direct answer to your question. [48:41] I do agree so much of the other part of what you said, which is [48:44] that AI really needs to work for the people. [48:46] It needs to amplify human potential rather than replace it. [48:52] And I hope we will do a lot of the work, [48:55] and we are doing a lot of the work in that space. [48:57] But I hope in some of the interviews that you mentioned, [49:00] and some of the conversations from other companies, [49:04] that this will be an increasing theme of how they can bring that to reality, too. [49:09] - The reason I asked you that question is because... [49:13] I think me and you were at a dinner with Scott Wu, [49:15] founder of Cognition. [49:16] I just recorded an episode with him. [49:17] I've spent a bunch of time with Scott over the last few years, [49:19] even before they launched Devin. I met him. [49:22] And what I think he does well, and what you do well, [49:25] is that you guys are always talking about the positive benefits [49:27] that your products are creating for people. [49:31] Like, I think he told the story of a woman who had lost her voice [49:35] before she got married, [49:36] and then she worked with you guys to recreate, [49:39] and so therefore she could essentially redo her vows in her own voice. [49:44] It's like a perfect example of that. [49:46] - Voice is such an incredible identity that, like, probably... [49:50] And the case that you mentioned is some of our proudest work. [49:53] Where we can work with people who lost their voices to ALS, [49:56] due to throat cancer, and work with them on bringing it back. [50:00] That is one example. [50:01] Recently, we worked with a musician who lost his voice, [50:06] and he wanted to still perform, [50:10] so we worked on recreating that voice. [50:13] We set out a concert, [50:14] and he did a concert with his old band together with an AI voice. [50:18] And now he's touring, [50:20] and in the UK he's going across places and touring across. [50:26] Or a congresswoman last year lost her voice [50:28] and still wanted to inspire others to do the work, [50:31] and was in Congress for the first time with an AI voice, [50:35] trying to bring that to life. [50:38] I mean, the common theme is that the voice carries such an additional element [50:43] of emotional impact and recognition. [50:46] The moment you hear someone's voice, you recognize it if you know someone. [50:50] Of course, that has a second part, which is how we safeguard, [50:53] and how you think about safety across that future where voice can be created. [50:57] But yeah, that's the work across there. [51:03] Today, it's over 10,000 people where we worked on bringing their voices back, and [51:07] we hope to continue that. [51:10] And outside of the accessibility space, [51:12] of course, what happens in education, wider culture. [51:16] We partnered today with 800 organizations [51:19] to help bring the technology to the people out there. [51:22] Actually, that's a crazy story, recently. [51:25] Sorry to interrupt you. - No, go. [51:26] - There was... [51:29] First of all, a legend, a guy called Tim Green. [51:31] He used to be a bestselling author, [51:35] one of the top NFL players, [51:37] and then, unfortunately, he got ALS and couldn't do a lot of that work. [51:43] And from all the things you could imagine him doing, [51:45] he decided to go to the complete extreme, and he did a podcast. [51:48] So, he started a podcast. [51:50] - I'm going to have to find this guy and chase him down. [51:51] No, I'm kidding. [51:52] - Yeah, well, you should have an interview with him. [51:54] And he was... - That's not what I meant. [51:56] That's not what I meant, Mati. [51:57] - Okay, okay. That's it. Fair. [51:59] - Man, I will destroy him. No, I'm kidding. [52:02] - But he does extremely well. [52:06] He has great guests. [52:07] Had Howard Lutnick recently on his podcast some of the NFL players. [52:11] And this year, he won an Emmy for his work. [52:15] So it's like, wow. So, he's incredible. [52:18] - Wait, so you made that voice, though? [52:20] - Yes, we do his AI voice, yeah. - That's incredible. [52:23] - So, that was one of our proud small brick contributions to his work. [52:29] But it's just so crazy... [52:30] Like, it's the most extreme thing you can do, and he does it. [52:36] He inspires... I don't know how many people got inspired, [52:39] but we had hundreds to thousands of people reach out, [52:42] thanks to him, like, "how can we get our voice back, too, in the same way?" [52:46] - Well, I think you hit on something which is one of the reasons, [52:49] and I discovered this accidentally [52:51] is, like, what makes podcasting so powerful is the voice. [52:53] Where it's like, you know, obviously I love to read. [52:56] I mean, there are books just scattered around the house everywhere. [52:59] Obviously have, like, I don't know... [53:00] - Oh, you have quite a big pack there. - Yeah, I have like a thousand books [53:02] in my other library, like probably 600 unread. [53:06] And there are some authors where I literally just fall in love with what they do [53:09] and read every single thing they've ever written. [53:12] But if you would ask, like, my emotional attachment to [53:14] my favorite authors or my favorite writers, [53:16] like Cormac McCarthy, for example, - Mm-hmm. [53:17] - in fiction, compared to my favorite podcasters, [53:19] it's just not at all comparable. [53:22] It's like, I feel like I know because of the voice and [53:24] the human-like element of that, [53:26] I just feel like I know them in a way that I could never know, [53:29] like my favorite writer. [53:31] - For sure. I mean, you also... [53:33] You're in the conversation now, and it's like you... [53:36] The voice carries so many other dimensions than text. [53:39] Like text, of course, you imagine, you interpret, [53:40] but it doesn't have the emotion, it doesn't have the intonation, [53:43] it doesn't have the imperfections, it doesn't have the pauses. [53:45] - The imperfection is really important [53:47] because we have some guests that come on here, and they're like, [53:50] "Okay, I said 'like' too much." It's like, that's how you speak. [53:52] It's like, yeah, we can edit it out if you want, but like I slur my words, [53:56] - Mm. - I say, "Uh," [53:57] I make all these kinds of weird things. [53:58] But that is just how I am. I don't ever want to appear on a podcast, [54:02] and if you meet me in person, it's just, "This guy doesn't even sound the same." [54:04] Like, the imperfections. [54:06] People admire imperfection, or excuse me, authenticity [54:11] way more than they do perfection. [54:12] - That's true. [54:14] You know, it's actually finally... It even applies in a non-human way, [54:17] in a voice agent's way, too. [54:20] Initially, we are trying to create a perfect voice agent [54:22] that doesn't do any imperfections. - It doesn't sound human. [54:25] - It didn't sound human. And then of course the obvious thing, [54:28] and that was the clearest, it's like the "Ums," the "Uhs," the pauses [54:33] and suddenly, the performance of working with that voice agent skyrocket. [54:36] Everybody was like, "Whoa, this is so human. [54:38] This is good. I am happy to speak with that." [54:40] So yeah, it's making it imperfect. [54:42] It's almost now the element of that. [54:44] But voice does carry that information. [54:48] It kind of connects you in a completely different way, [54:50] and why I think also it's such a hard research challenge, [54:53] because here, you know, in text, you don't have that many of those dimensions. [54:59] You need a lot of data, of course, to create a good language model, [55:01] but you don't have the dimensions of every voice being different, [55:05] every voice sounding different to every person. [55:09] So, even doing benchmarks for text-to-speech is extremely hard, [55:13] because usually different models will have different voices. [55:17] That already makes them uncomparable. [55:19] - Did you know, when you were younger, that you wanted to be an entrepreneur? [55:23] - I would say I didn't know this was a path [55:25] for a while. - You did or did not? [55:26] - Didn't. I was, you know, I... [55:28] - Because you're European. [55:30] - In Poland it was... [55:31] Yeah, no, for sure, a little bit of that though. [55:33] I know you're joking, but I think it's true. [55:34] - No, I'm being serious. [55:36] - Taking the risk of starting something, [55:42] it wasn't like a common conversation ever happening. [55:46] I mean, I had the luck of having an incredible family [55:48] that kind of gave me the opportunity to study abroad, [55:51] and then that kind of opened your eyes, [55:52] like, okay, now you can work with some of the great companies, [55:55] and then when you work with those great companies, [55:56] then you realize you can actually do things. [55:59] Like, Palantir was great for that for sure, [56:01] where it's like you can actually go and work with a customer, [56:04] try to figure out the problem. You can take... [56:05] That seemed very risky to me, but you can do that. [56:08] So kind of step-by-step that opened the eyes of like, okay, maybe this is a path. [56:12] Maybe it is possible to start your own thing and... [56:15] - From the time you had that realization to the time you started ElevenLabs, [56:17] what was the... [56:18] - So then, as you kind of this... - Was it a year, two years you were... [56:21] Like, what was it? [56:22] - So, over my time at Palantir and my co-founder's time at Google, [56:27] we would start doing hack weekend projects together. [56:30] So, through the years, we tried to explore new technology and build together. [56:34] I think a few years prior, we knew that we would love to work on something together, [56:39] but you want to work on something about which you think it's a true problem, [56:43] and you're obsessed about. [56:44] And that came to us in 2021, so a few years. [56:47] Two years. Two to three years. [56:49] - How many acquisition offers have you had? [56:50] - Like, concrete ones, three or four. [56:53] - When was the last one? [56:55] - Last one was last year. Like, June last year. [56:58] - You going to sell? [56:59] - No. [57:00] - The reason I ask you... [57:01] - AI is changing the world. [57:02] We can build at the frontier of that change. [57:06] We are going all in. This is, I mean... [57:09] - Then don't pay attention to VCs, dude. [57:11] Their incentives are different than yours. [57:13] The reason I ask you is that there are two reasons this came to mind, [57:15] and some of this is like all intuition. You can't even describe this. [57:18] But Evan Spiegel sat in that exact same chair that you're in. [57:21] You know, people gave me shit. [57:23] They're like, "Why you want to interview Evan? [57:26] Look at his market cap." [57:26] I was like, I don't give a shit about his market cap. [57:28] I don't look at companies that way. [57:29] - Yeah, that's it. - I'm obsessed with products. [57:31] And like, that dude has soul in the game, and I hope he wins. [57:33] I have no idea, like, I don't know anything about, you know, [57:38] I don't use Snapchat, Specs, anything, whatever the case. [57:41] It's just that he is differentiated. [57:43] There's just something about him that I like, [57:45] that I just want him to work out well, right? [57:48] And I'm getting the same exact vibe. [57:51] I was like, man, I want Mati to win. I want him to succeed. [57:53] There's just something very likeable - Thank you. [57:55] - about you. [57:55] And then, the second reason is because, obviously, Scott Wu, [57:58] he's come on the show like every few months because I really like Scott a lot. [58:02] - It's because he's a genius. He's so good as well. [58:03] - Well, yeah, but he's articulate and brilliant and optimistic, [58:09] but everybody is running at that dude right now. [58:11] Everybody. [58:12] The amount of people who want to buy his company, [58:15] and this is something we talked about. [58:16] And I hope he holds out, because I would like to see a lot more people [58:20] just be like, "No, I'm in this forever. It's not just to get a big bag of money." [58:25] You know, go listen to Travis from Uber. [58:29] You know, he made billions and billions of dollars from Uber, [58:31] and he's just like, "That did not make me happy. [58:34] I need something to work on." [58:36] And see, that's why I asked you the question because I asked Scott that too, [58:39] and it's just like, well... [58:40] There are obviously a ton of people [58:41] trying to either get him or his entire company and all his talent, [58:46] and he's just so far said, "No, I'm not selling." [58:49] - Love Scott. I think they should build an independent company too. [58:52] I think they have an opportunity to be one of the hyper AIs of the future, [58:56] or however you call the hyper clouds of the future. [59:00] So, I think he can do it. [59:04] They have an incredible team, and I think we can do it too. [59:07] I think it's... [59:08] I really mean it. I think the opportunity that [59:12] currently exists for entrepreneurs with the wider shift, [59:17] the things haven't been written, and it's like it's such a good time [59:21] to build something special, and... [59:23] - I think shows like this are really important, because most of the podcasts [59:26] are made by VCs, right? [59:27] It's just like the content [59:28] that entrepreneurs... - Yeah. [59:30] - The content that entrepreneurs are... [59:31] - It's a little self-serving in many ways, I feel. [59:32] - Well, the content entrepreneurs are consuming are created by VCs. [59:35] It's a weird thing to me. - Yeah. Yeah. [59:37] - And what I would say is it's like the amount of f****** founders [59:39] that I've talked to that have sold their company are like, [59:41] "Shit, I had something. It was going really well. [59:45] They gave me a bunch of money. Now they tell me what to do." [59:47] I was like, "What did you think the money was for?" [59:49] Like, no one's just going to give you a bunch of money, [59:51] and then you still retain the independence to do whatever you want. [59:53] And then it's in all these freaking biographies from Ted Turner to, [59:56] like, there's just a million people that talk about it after the fact. [59:59] And they got huge bags, billions and billions of bags. [60:02] They're like, "I'd give the billions back if I could just have my company back." [60:04] And listen, man, I think you're smart and driven, [60:06] but highly likely, [60:08] ElevenLabs is probably the best idea you will ever have in your life. [60:12] And you're how old? - Oh, for sure. [60:13] Thirty-one. [60:14] - Okay. So, you're going to sell your best idea at 31? [60:17] You got four decades ahead of you, [60:19] maybe five. - Yeah, but it's... [60:20] - Hold on. And you're going to work on your [60:21] second, third, fourth, fifth best idea? [60:23] Dude, the money's not worth it. [60:25] It's not worth it. You're going to get the money anyways. [60:27] - Well, you're convincing me something I'm already convinced of, [60:29] so that's easy. - I know. But they all say this shit. [60:31] The founders all say this, and it's really hard when they're just like, [60:34] "Dude, I'm going to drop 50 billion on you." [60:36] It's just, the numbers are getting crazy. - No, I... [60:38] - So, I understand why people do it. I'm not knocking them. [60:40] I'm just saying I would like... [60:42] I'm very interested in entrepreneurs, because there's no price. [60:45] Like, again, I've repeated this over and over again. [60:47] It's super important to understand. [60:48] Everybody's like, "Oh, if you love what you do, you do it for free." [60:51] No, there's another level. [60:52] If you love what you do, they couldn't pay you to stop. [60:54] How much money would you have to give Steve Jobs and say, [60:57] "I'll give you two trillion, Steve, but you can't work on Apple"? [61:00] He'd say, "Go f*** yourself." [61:02] There's not a dollar amount in the world that could stop him from doing that, [61:05] and the world is better because he didn't stop doing it. [61:07] - I'm with you there. [61:11] I agree. Many people, I think, will say it. [61:13] In our case, we had a lack of acquisition offers, [61:17] which we turned down, and that was not an option. [61:20] And any of those acquisitions would... [61:22] You know, it's not two trillion, but they would have made life easy, of course. [61:31] I think that in itself should never be an interesting proposition. [61:34] And like you said, it's like, [61:36] we do appreciate this is likely the best idea, [61:40] the best timing to have the kind of lack of when it's all happened [61:44] is such an incredible coincidence, [61:47] and I don't know how the world will look like in five to ten years, [61:51] how will the universal high-income peace and conversation come in? [61:54] How will the wider society adopt the technology? [61:56] How will you actually provide a... [61:57] Like, all of those questions are out there and I think they will be a big part. [62:00] So, of course, the best thing, the thing we can do for the world [62:04] and for us is just to continue building. [62:06] - Did you read Scott Wu's piece in "Colossus," the "Colossus" magazine? [62:11] - No. - That's done by my friend Patrick. [62:12] - Yes. [62:13] - I'll text it to you right when we're done. [62:14] You can put in Eleven. [62:16] - Yeah, I'll listen to it on the way. [62:18] - Yeah, put it in ElevenReader and listen to it. [62:20] What I loved was his perspective at the end, he's like, "Listen..." [62:23] You know, he thinks the same thing: [62:25] Right person, right set of skills, right time. [62:26] He thinks just teaching AI how to code, [62:30] teaching computers how to code [62:32] is one of the most interesting problems he could think of. [62:35] And he's like, "Listen, I could accept that if I try and fail," [62:39] but what he felt was intolerable, I forgot the word he used, [62:41] but it's something like intolerable, is like, "I didn't even try." [62:44] - Yeah. Yeah. [62:45] - He's like, "I just want to give this one opportunity, [62:48] the best opportunity in my lifetime," which he understands. [62:51] I think he's around your age, too. - Yeah. [62:52] - He's like, "I'm just going to give it everything I have, [62:54] and see what happens." - Yeah. [62:56] It's like some version of the biggest risk is not taking any risks, [63:02] and some version of... Like, you should just go after it. [63:06] - Does your co-founder think about this the same way? [63:08] - Yeah. He's also all in. [63:10] He doesn't like public presence, as in interviews and podcasts, [63:14] but he truly is a genius. [63:17] I hope one day he comes on the podcast, too. [63:18] - Anytime he wants. [63:19] - He is, of course, a great researcher, but beyond the researcher, [63:21] he also can bring a lot of those ideas in business and other parts of the space [63:25] and kind of understands what's happening and how it happens. [63:28] But on the research side, [63:32] the kind of only AGI pill, [63:35] and how the world will change, is very deeply in his mind. [63:38] And he knows that he can be... [63:41] He and now the wider research team are all part of that change, [63:44] so it's a unique opportunity. [63:47] I think all of us realize that how unique the timing is, [63:51] and what we can do with that timing, and what we can do for the world. [63:53] I think that's like, that we are four and a half years in as a company, [63:59] and that scale, it's crazy. [64:01] - Because your opinion is that the next form factor isn't a device. [64:04] It's actually just the way you're going to interact with AI is just your voice. [64:08] - 100% will be one of the biggest ways. [64:10] Like, in general, intelligence keeps developing. [64:14] The next bottleneck of how you actually get access to that intelligence [64:18] will be how you communicate with that intelligence, [64:19] how you collaborate with that intelligence and we can solve that. [64:22] We can solve that for everyone out there. [64:25] - Twelve months from now, [64:25] how do you think the way you communicate with AI [64:28] is different than it is today, then? [64:30] - Voice is definitely part of it, but it kind of understands you. [64:34] It's able to connect the IQ plus the EQ part of it. [64:38] Emotionally understand and know how you're feeling, [64:41] can adjust based on that, can pause, can think, [64:44] can reinsert itself into the conversation. [64:46] Like, in some crazy way, for decades, [64:49] we learn how technology around works and learn the language of that technology, [64:55] the keyboard, the screen, the coding languages, even. [65:00] You need to learn how the technology works so you can control it. [65:05] And now, what I think we can solve is flip it back to how we want to communicate, [65:10] how the most primal way is voice conversation, [65:13] and you can bring technology on our terms. [65:16] So, I think 12 months from now, what will happen is similar as we are speaking. [65:20] I think the technology will be able to understand us in an incredibly better way, [65:25] both our knowledge, both the emotional part of that conversation, [65:30] and deliver similar conversations we are having now. [65:33] - Yeah, and if that's the case, then the market for this and even the use case, [65:37] the amount of people using AI will drastically explode. [65:39] I think there's actually a historical equivalent [65:41] that just popped to mind when you were speaking, [65:42] where it's like Alexander Graham Bell was talking about the difference between [65:45] the telegraph and the telephone. [65:47] And the telegraph could send messages over long distances, [65:50] but you had to learn how to program it and use it, [65:53] and you had to learn this language to send it. [65:56] And he's like, "The phone, you just pickup and do exactly what you do already." [65:59] - Yeah. 100%. [66:03] I think, even I'm looking around the room, [66:06] I feel like, it will be so many devices as well [66:07] that will just be able to work on your terms. [66:09] Hopefully, the screen, the phone will kind of be able to be back in the back pocket [66:13] if you won't need it in the same way as you do now. [66:15] So, it'll be cool. [66:16] - When you first started the company, you mentioned this from [66:21] "The Hitchhiker's Guide to Galaxy." [66:22] So, was you and your co-founder's ideal version of your product, the Babel Fish? [66:27] - It was one of the slides that we thought that, yes, Babel Fish will exist, [66:31] and we'll hopefully make it happen, [66:32] but less so that we'll create Babel Fish itself, [66:35] but we will enable everyone out there to have Babel Fish in their existing devices, [66:40] existing presence, existing work. [66:42] That was definitely one of the kind of North Stars that we are thinking about. [66:49] - Okay, another question I have for you. [66:51] Why do you always pop up at all these conferences? [66:55] - I don't do that many, I think. [66:56] - I get these emails, so Daniel Ek, [66:58] obviously a close friend of mine. - Yeah. [67:00] - He hounds me on this, and he's right. [67:02] He's like, "Dude, you have one of the most elegant businesses in the world." [67:05] He's like, "All you have to do is sit inside a room and make podcasts, [67:07] and every single thing that you want in life will come to you." [67:10] And his line for this is like, "Stay away from the circus." [67:13] - Yeah. Yeah. [67:14] - And so, the only ones I do, obviously, do events for my partners, like ramp, [67:17] where I saw you at the ramp dinner. [67:18] And I obviously love Michael Dell, so I do his too, and I do nothing else. [67:24] And because of Daniel's advice, just like staying away from the circus. [67:27] Like, the people who show up at these things, they aren't doing any work. [67:29] They're just there to distract. Like, you don't have to do it. [67:31] And so, then, I get all these emails and invitations, like, "Come to this thing," [67:34] and I open it up, and I see your face everywhere. [67:38] What are you doing? [67:39] - No, I usually try to do two to three per quarter, usually conference events. [67:47] But in our case, it's a little bit different because [67:49] for a lot of the people that are usually on the conferences, [67:51] they are partners or prospecting partners or clients. [67:55] So, it is usually a great way of forcing function, of having them [67:59] in one place and trying to figure out how we build together. [68:02] That's the benefit of the breadth of the work we do, that a lot of... [68:05] Especially on the conversational agents work, [68:08] that applies to most of the businesses of how they think about communicating [68:11] with their customers, how they are thinking about changing customer journey. [68:14] So, a lot of the events are a great way for us [68:19] to catch time with a lot of the people that we work with. [68:21] - And then turn them into customers. [68:23] - And turn them into customers. [68:24] - All right. So, you are working. - Or expand the one to two or five, [68:26] whatever it is. You know, the common trope of... [68:29] And I remember thinking back in the day, when we started. [68:32] I got the first invite for one of those conferences, and it's like, [68:35] "Oh, this is super cool, super fun." And then, you go. [68:39] I'm European, so I drank alcohol at the time as well. [68:44] And then I felt tired. I didn't think I did anything, [68:47] and it seemed fun, but it wasn't very productive. [68:51] And then, I had the period was like not... [68:55] - Did you have nothing to sell at the time? [68:56] - At the time, I didn't. [68:57] I didn't know what the purpose of the conference was. [69:00] I think it was the lack of preparedness on my side, [69:02] where I was going into the conference and just flowing through it. [69:05] And I feel it was a kind of terrible thing. [69:08] You go there, you have the sessions and some of them are, of course, good, [69:12] but most of them are not relevant to you or your business, [69:14] or they are the circus part that you mentioned, that, like, [69:16] you feel you are doing something important. [69:18] It's completely unimportant. [69:20] And that's kind of... [69:22] Then I didn't do any events, and then I realized that it's actually... [69:26] You can do them well. You do need to prepare. [69:28] You need to prearrange a lot of the one-on-ones [69:31] you want to do during the conference. [69:32] Everybody is there at that time. [69:34] You don't want to do too many of them because a lot of the... [69:37] There's like some repeating crowd across those events, [69:39] so you want to do the ones that are kind of different. [69:43] And of course, you want to be pretty explicit to the other side too, [69:45] that they know what they are... Like, you know that you are not... [69:47] You're pitching them and they are relaxing. [69:49] That's like a bad recipe for disaster. [69:52] But many people want that too. [69:53] They are there to do exactly the same thing: [69:55] Do business and figure it out. [69:57] So now it works really well for me. Like, I do good prep. [70:00] We prearrange a lot of time. [70:01] I never go and try to be there for [70:04] just the sessions or just the content or when... [70:06] Of course, I speak frequently on those conferences, [70:08] but the main thing is trying to grab time with the people that are there, [70:13] and that has been working really well. [70:15] - Are you still drinking? [70:16] - No. On the weddings, but apart from that, it's... [70:19] - Every time I go to Europe, I have the same thought when I get back. [70:22] I should drink more. [70:24] So much fun. [70:26] It's like they just know how to have fun. [70:28] - There are social circumstances where I would drink. [70:31] They are just so infrequent now that I wouldn't do it. [70:35] - Okay, next time we are at one of these rare conferences together, [70:38] let's make sure we have a drink. [70:39] All right. Deal? - Yep, deal. [70:41] Let's do it. - All right, man. Thanks for doing this. [70:42] Appreciate it, always. [70:44] - I hope you enjoyed this episode. [70:45] Please remember to subscribe wherever you're listening and leave a review, [70:48] and make sure you listen to my other podcast, Founders. [70:50] For almost a decade, [70:51] I've obsessively read over 400 biographies of history's greatest entrepreneurs, [70:56] searching for ideas that you can use in your work. [70:58] Most of the guests you hear on this show first found me through Founders.