ElevenLabs’ Vision for Voice Interfaces | CEO Mati Staniszewski

27:39The interesting question that I keep circling back to is like, if you've crossed the Uncanny Valley for a three minute clip, why can't you cross it for two hours? And if you do cross it for two hours where I cannot tell, then why wouldn't a director just give their entire script that they wrote to an LLM. And then it can be like, put it in the style of Goodfellas with the punchiness of whatever, right? And it just does it. And then maybe there's some tuning or whatever. I'm just saying like, and if that's true, if that is possible, that would be insane. No, but, and I agree. And I think this will be possible, but coming up with Goodfellas or writing the script for Goodfellas or like that book itself, that's going to be extremely hard.

28:31Agreed. That's going to be extremely hard. And then there's another piece, which is like, it will be easier to create software, to create content, to create movies. So now you're kind of even more attached to the storyline, to the history of something than ever before. And that history will frequently go back to, has it happened in real life? Has it happened? Is it the story I used to follow when I was a kid? Is it a universe of characters I know and love and follow and other people do as well so we can kind of speak with it? So I think it's going to be possible to do it, but it'll be harder to create the attachment to the human across all of those use cases.

29:12Totally. What percentage of the revenue-ish is B2C for you versus B2B enterprise? It's closely approaching 50-50. Oh, interesting. Yeah. So it's a lot on the B2C as the creative use cases, narration, voiceovers, dubs, music. And then on the B2B, it's huge conversational AI push with our agents platform. So everybody's building for conversational use cases, making in voice, but increasingly also something not only voice, where like an omni-channel experience where you do voice and text interactions. So that's kind of skyrocketing across. Yeah, it makes sense. I mean, on the B2B side, like one of the great use cases that we have seen is support.

29:59Yeah. It's great because it's generally a cost center. And it's great because LLMs are very good at taking that unstructured nature and putting it into a structured format that it can automate away. And in so many cases, like it makes total sense to me that you're powering that workflow. Yeah, and 100%, it's kind of the place where we see just such a quick adoption. But an interesting thing that started happening is this shift from this being a customer support place where you reach out as a user when you have an issue into increasingly being more of a kind of a front-facing experience, customer experience across the entirety of your product or user journey.

30:44So from, you know, like in a traditional way, if you've got an issue, you would reach out through an email. Now we see companies working on bringing that as part of their website experience itself. So recently we worked with a company in Italy, the biggest real estate company, where they effectively placed an agent that as a buyer and seller of a company, you can reach out to to get information about that real estate. And they had like a historically a huge supply and demand mismatch. So the seller of a property wasn't available to pick up the phone to give the information about the property. And on the other side, you had a lot of the buyers that were just not the right buyers for the property.

31:27So now you have an AI agent that you can call in and reach out and get information about the property. And then that information gets aggregated, given to the seller of the property, and then they can reach out to the right ones. helped improve the qualification rates from like 20 to 60 % of people that are progressing to the next step. And before, you would only reach out if you had a problem going through that experience. So it kind of is interesting where increasingly it's like kind of the part of the experience itself. Insane. So insane. Even in our own work, we love trying to do it. Like how can we elevate our experience?

32:01We of course had, because of the B2C and B2B angle, we had so many people reaching out for help, credits how to use the product so we shifted it where now you have our agents both on the website itself so you can speak with an agent about 11 labs the moment you open the website of 11 labs then when you log into the product you can you have an agent that can help you navigate through the product experience will open items for you speak with you about how it works and and then of course on top of that you can reach out to support which is also powered by our agents So now you have like a full fleet of from understanding to using to reaching out for help.

32:39And to even if you reach out for like sales part, we work with our own agents to help with that process. And it's been amazing. It's been both amazing for the experience itself elevating and of course solving the issue. But second, and you kind of spoke about this briefly earlier, we noticed a behavior change where if you know you can speak with an agent, your behavior changes. And let's say you're talking about your use case, you are much more keen to tell you everything about the use case because you know you are not judged. You can be a little bit more expansive. You have any time in the world to tell you about the use case.

33:21And this has, in turn, meant that we can provide you better experience. even in the classic kind of BDR, SDR categories where you can reach out to 11 labs, all usual flow, you leave your details, you can request a use case. But then if you want, you can accelerate and give more information by talking to AI BDR and then we loop you into the right person a little bit quicker. And that has seen a great adoption as well. How big is the company now? So we have 330 people now. All everywhere? everywhere we have the biggest basis so we we have um still run it remote so everybody can can work remote but we started investing into hubs where you can work in person and our biggest ones are in london and new york san francisco warsaw and now we are starting some in tokyo bengaluru in sao paulo in brazil as well and you just raised it what was it like six six six 6.6 billion or something?

34:236.6, yeah. And you let people take secondaries. Yes. Actually, there was only a tender. Yeah, only secondary. Only secondary. Only secondary. 100 million? 100 million. 100 million of only secondaries. Yeah. Why? Was it a hard decision? Not hard decision. The goal is simple. We are here for the long term. We want to build something that will stand the test of time, but also will require the next five, 10 years of us continuously pushing. And we want people to be aligned on that long term. And by making it true that A, equity is valuable. Two, you can have early liquidity. So you take some risk.

35:05You get some of the reward from that risk, but you are incentivized to work with us on an even bigger outcome. And I think it really helps. And I'm sure many companies are saying that too, But I do think we have so many extremely passionate missionary-like people that are in here for that long term. And for them, the change in the capital will not change their motivations in terms of work, which is, I think, the most common piece that people mention. And we haven't. We've seen people are even more aligned. How do we get us now as a proxy of the impact we want to create? How do we get 11 laps to 11 squared or$121 billion company?

35:51Where I think it's possible. And I think we can. And now people are here for the long term. Where do you draw the line? I think it's tricky, right? I'm sure at some point it will be tricky. Yeah, the current one, in the same way like in sports, if you hire the best high-performance athletes, for them, if they are doing well, they want to be part of the winning team. That's kind of one. That's what we need to do. But two, the money equation doesn't change. You still want to perform until of your game for the years to come. And I think the same applies to if you set the culture in the right way, and then you have the people that are performing next to you in a great way, then it shouldn't change that motivation either.

36:44I think the harder one, as you think about that, it's harder if it's not going well. And of course, for a short period of time, if you've said it well, you can withstand that. But if it's a longer period of time, then I think it will get challenging because then a combination of you taking, although maybe not because you're still having, you know, a big incentive to make it great over long term. Yeah. Well, and you're not taking up rounds. You're not doing up rounds at that point if it's not going well. Exactly. So you kind of are. Yeah. So even in that case, you shouldn't. I'm sure the motivation will be a little bit harder, but I really believe in our team.

37:23I think they will be able to go through the hard and, you know, the hard times are never as, so the bad times are probably never as bad as they are or as they seem in the moment and the good times aren't as good as they seem. So you kind of are a bit closer to the median or the average of the times. And I think it's true here as well. So I, so on the bad times are never as bad as they seem. I agree. However, I know that I've internalized that. I deeply believe that everything that I've ever been anxious about basically in my entire life either hasn't come true or was not a big deal. And knowing everything that I know, still, I run into, let's just say with the company that we're working on now, I run into something and I'm like, oh my God, like I do not, like, it just feels terrible.

38:10Like I can't in that moment face a bad time and be like, eh, this isn't going to feel bad. Like I know. But I think it's good. I think it's good. It's like you, you, you care. That's why I think it's like the same way for me too. It's like, and for many people around, it's like, you still care. So every single kind of direction takes additional emotional impact. I think, you know, you can still stay level-headed and make the decisions. I think that's the important, important part, but, but it seems, you know, that, yeah. Yeah. My, my, my observation is that there are so many decisions to be made, right?

38:48So many decisions to be made, that like, each one of them on their own, you don't even notice that you're making these decisions. But over time, they just kind of like, it's just like, like, it's like a quarter percentage of energy on every single decision, you know, and then over time, you just like sum all of those up. And you're like draining or depleting some reservoir of decision making energy that you have in order to like, kind of like, when the one comes that's like oh you know you're like well i just made like 50 decisions yesterday you know it's like yeah it's it's we i've went for a few of these conversations recently with a few of people that were i sees for a long time and now are leading parts of the work and they just feel that from being able to what felt like meaningful work they're replying to constant things or tiny things across.

39:44And so first of all, for sure, I see like, you know, the kind of the more you get, the more you will get asked for decisions across. But then I think the two things are important to help out. One is you have, of course, the different segments of the company that applies in different ways. I think in our case, we have an amazing set of leads across that will effectively take a lot of those micro decisions and kind of trust them completely. They will do extremely well. That in turn means that the bad things get filtered out always back to you. And then the second thing is that I think there's this kind of balance of like which decisions you should really take versus the individual should take.

40:32and it was like plus minus 10%, they should just run with it. In our case, what we are trying to like instill for everybody that we are small teams, this will only work if people take ownership and try to make a lot of those decisions themselves. We are assuming and we are trying to put them in a position where we fully trust them and then they can flourish. They won't be punished by doing that and running ahead, but they feel empowered to like run towards the fire, take decisions, fix it, not try to ask. they're like you know fuel these up on like is it right and kind of go through that iteration cycle relatively quickly very of course now as a company we run the company on micro teams so we have 330 people on the product side it's like 20 teams of 5 to 10 people working on different segments of the product work and they are executing and then you kind of need to live with the part that you need to take more of decisions.

41:37But then in turn, it works well. Are you holding your breath every big model release? Like what I find fascinating about what you're doing right now is that your growth is fueled by the fact that you are in the eye of the hurricane, right? Like you are in a middle of a market that really matters. And it turns out when you're in the middle of a market that really matters to people and consumers, especially consumers, that also really matters to OpenAI and Anthropic, right? And Google. And, you know, like we saw this in our world, like, for example, like coding really mattered to Anthropic and OpenAI.

42:25Now, you know, Cursor has so far made it out the other end of that kind of, and maybe you're the cursor of this world and maybe cursor will continue to be ahead of them but these are not the uh incumbents of old right right like these are like it's just these are big giants right that are very talented that if you're in one of the key areas it is it just i'm not even saying oh they're gonna do it or not. It's just got to be a crazy time and somewhat nerve wracking and like knowing that you're right in the sweet spot of something that they really care about. Because consumers really care about it.

43:08You probably get the annoying VC question of like, why won't OpenAI do it? That's not really my question. It's more like when you're in this eye of the hurricane, it's insane. Look, we will fight and I think we can win. It is eye of a hurricane. And so far, we've been able to win time and time again. But I think we are in a unique position where we do both. We do foundational research model and the application around it. So in turn, this means two things. One, to your kind of main point, on the research side, we need to stay competitive. So that are better than OpenAI, Gemini models. And I think we can.

43:51We did it on text-to-speech. Then we repeated this on speech-to-text. Then we did it on music, orchestration, bring that all together. And we will continue battling them out and hopefully winning. It's a tricky one. But that's one. So definitely checking the mark and holding the breath when things happen. But at the end of the day, models only one piece of it. The second piece is as important, which is that application side how do you create the best platform for creating an agent beyond having the best voice or the best lm you need how does it integrate with your crm system with sales for service now goal drive how do you build functions in it how do you decide when to refund or take an appointment scheduling how do you then deploy that in production how do you test evaluate version control and then kind of we spoke about this is the platform functionality but what most customers really care about is the outcomes themselves.

44:47How this will solve, how will that impact my CSAT score and customer support if I do deploy that agent, help me understand that it was the cost benefit ratio of doing that work. And none of those other companies usually do that on the agent side. And that's where we focus and spend a lot of the energy. So now we are in this interesting spot where benefits of a lot of the foundational work ultimately end up benefiting the agents. And we hope to do both. We hope the product is going to be best and the incoming audio foundational layer that we do will be best too. Same applies very much on the creative side.

45:23If you are creating a voiceover or a movie, you are going through a set of iterations on selecting the voices, redoing the lines, adding combination of voice, sound effects, musing the track, maybe reaching to some of the assets you created from the past Maybe you created the voices that you care about and you want to bring those voices inside of the production flow. And then you want your other teammates to check and check whether that's good. And suddenly that process became a pretty complex flow to go through. And you need a pretty solid platform to be able to do that. And we kind of combine all of that in one place for you.

45:58Yeah, but now you're fighting all these flanks. We are. And that's good. You have all these flanks. No, it's great. I just, I can't, do you sleep well at night? These are a lot of flanks. You know, it's one way of looking at a lot of different battles you need to do, but it's also there's fewer points of failure. Because if you have an amazing research, amazing product experience across view, then even if your research isn't amazing and product is, you can still benefit from other research in this space. And the opposite is also true. So, you know, it's the way we, like, as I think about the flanks and small teams and winning, it's like, if we win one of those, we are an amazing company.

46:41If we win many of those, we can create a generational company. And the teams feel that too. Like, if we all do great work, then, and I think there's no other way. You cannot create something generational if you do, like, you know, one or two problem solves. You need to solve a double-digit amount of different problems to really create something special. And I think we are headed towards that. Why do you care about doing something generational? Like, why does it actually matter? I don't think, yeah, it's a good question. I don't think in itself, you know, like in a way, the work we create has like a positive impact on the world.

47:18You want to extend that impact as much as possible. kind of your creation should outlive you in many ways. And if it can compound and give the value to the ones around it, like what better way of impact of humanity where you kind of extend the technological advancement for everyone out there. So like in my mind, the generation, I'll just shorthand for that, where all of us can create something that will kind of go beyond what we can do ourselves in the future, which is just an exciting proposition of leaving something for the world. Yeah, fair enough.

47:55How, like on the research side, is it as we all read and see, like, is it as intense? Like, pretty soon we're going to need like an agency to manage, like a CAA to manage the like contracts of these researchers. Is one of the flanks that you're protecting Zuck not getting into your research, your lab? I'm joking, but I'm also being serious. You know, in a way, I'm very happy that for the world this happened where the IQ and the research talent is valued in the way they are. It's like in similar, like for me, it's the most amazing athletes you can imagine. And yes, they should be valued in a high way.

48:51There is a tricky piece and a similar in sports where one is more on the micro level, but the next years are the most critical for that world of research and how that innovation will be adopted. So the value is definitely there for the next years. The interesting thing will be, to what extent will the models themselves be able to produce that over time? I think very hard. You still have the set of innovation breakthroughs you will create. But no, I am one of big flanks. I think beyond the flank of compensation, you need to also create a flank within that of like how do you set them to be able to create the best work of their eyes and then release that work and see that work used.

49:43And I think we've been able to do that very well where a lot of the research work that's done at 11labs gets you to the users almost immediately. It's like such a short iteration cycle between the real world application and what you are creating. And then it's a super small, very high density team where there's like, everyone is amazing. Can I explore the like timeframe point on the researcher thing? Because I think it's super interesting. Is what you're saying like, hey, right now for the next several years, whatever that timeframe is, they are crucial to advance the state of the art on frontier models.

50:23But at some point, the models will get so good that their value will relatively diminish because the models will start to train themselves. I'm not putting words in your mouth, I'm just trying to clarify. Yeah, it's a variation of our previous conversation. So less so on the comp part, but most of the researchers that are in the team and most of the researchers even outside of the team when we speak with them, many will believe that AGI will happen in the years to come. And in that world, the whole compensation question is an interesting one. I don't know how this will work in many of those aspects in that reasoning thought.

51:00but that's where a lot of the kind of the thinking process goes through, where the current work and towards AGI will just change the dramatic outcome. Do you believe that? I'm less... Are you allowed to say that you don't believe that, even if you don't believe that? From like the self-serving perspective or like AGI being against me, if I say that? From not AGI being against you, from the... Yeah, I guess like self-serving, but also like... I don't know, I just haven't really heard anybody that's in the eye of the AI hurricane not be maximally bullish. I'm maximally bullish there will be a lot of breakthroughs, but I'm more on the longer-term spectrum.

51:43I don't think it will happen in the next two or three years. So that's where I would disagree as I think about the conversations with a lot of the research folks, but they know more than I do, so I should trust them. But I do think it's going to... A lot of the innovations on the foundational stack will happen and then bringing that to the real world will take a much much longer time frame just like the um the the business process of bringing that into an organization people adapting to how to work with this will just take such a long time so you know it's it's it's it's like closer five to ten years in that dramatic shift um uh and then it's it's like you know Now, what's the distribution of compensation then look like?

52:26It's an interesting one too. Is the knowledge work that's replaced or is it the creative work that's adjusted or is it something completely else? I'm here is more in a category where I think there'll be more of the middle to middle where the super smart people or super creative people will use those tools to their advantage. So that gives you that superpower. so I think you still will be in that category but you will want to you will need to want to do that work too have you had any crazy insane recruiting stories like have you had to do things that you were like I cannot believe I just had to do this to win or save this candidate the many but I don't know which ones I can mention no names but the AGI piece of like the world will change is yes That has happened a number of times.

53:20So we're just, which is, so the previous part was like, we had people, it's like, hey, how, one we didn't hire, but it was, hey, I want to maximize what I can get in the next three years, because I think in three years, the world will not lose. So it's a part of their negotiation strategy. Because they also view their own work, along with everybody else's, as diminishing. Yeah, no, exactly. If we achieve some artificial general intelligence. Exactly. That's insane. That is, that was interesting. That's really dystopian. Dystopian and very like, so we didn't end up proceeding there, but it was impossible to convince the other person that it, you know, like how it has impact for everybody else in the company, how that kind of sets up a bad precedent, et cetera.

54:07But it did, it was like, it's very hard to convince someone that AGI will not happen if they are spending, you know, most of their work and life in that space. But that's kind of the question that I have. Like, it is very hard to convince the smartest people who are the most entrenched in this technology that AGI will not happen in the next few years. Whatever AGI is, we can figure out that definition. But like something that does not exist today. But like, I guess like, why don't you or I believe them? that like that's that's the question that i keep coming back no i i part of it's like there's some percentage that i do it's it's you know it's uh it's a big question where i think there's probably much more knowledgeable people than i that can take it but um no i think there's like a good degree of percentage percentage points that that this might happen and they are fully right it's insane it's insane what do you take what's your take i'm probably one of the uh least disqualified people to talk about this relative to the folks that that you work with every day but my point is like man i'm still trying to get an llm to not sound like an llm like i'm still in the world where i'm getting sent text that sounds shitty like an llm and everybody else is talking about this world where compensation won't matter in a couple of years and i'm like a lot will have to go right for us to get from this point to that point, you know?

55:43And so I'm like, all right, like there's a few things, at least in the enterprise that have worked really well. Like coding has worked really well. Support has worked really well. Like there's not that many things that have actually worked really well yet. And then on the consumer side, like I've seen some behavior change, certainly my own, where I have a sounding board basically at all times. But I'm not, it has not, it's amazing. It is profound. It makes me more productive. But I don't feel the way that I think about my compensation or the value of my real estate or, you know, like cancer getting solved.

56:34Those have not, I haven't changed my milestones in my head. And maybe I'm like some laggard, but like I live in Silicon Valley and like at KP. So if you asked a lot of other people that are not here, I don't know, maybe they would reflect more of that sentiment of like, I don't know. I don't know, that's my take. I'm more in the longer timeframe because I think it will like so long to adopt, Even if you have like a, you know, like a knowledge, knowledge, like the smartest human in your pocket available that you can, you can speak with at any point and the kind of thousands of them. It's still like before you can realize all the outcomes for the world will take a longer period of time.

57:13Like even in company building, everybody's probably telling you like, oh, you can, you don't need 330 people. You can do this with a hundred people. and then what ends up happening like reality hits you across the face opens open eyes like hiring diagram what is it it's it's like scaling as quickly as it ever ever did ever did right yeah and by the way uh even in that case they're like you know 50 or more of all of software is probably being written by ai within that company and they're hiring a ton of engineers no 100 i it's so so that's yeah case in point these are like i mean i just i don't like if everybody that's smarter than me is telling me this is definitely inevitability then why isn't the behavior of the places like why is no no i that's i think we agree there i think it's going to take much much longer time before it's adopted and across all products across all dimensions and in uh your 330 you're just crazy to think to me as well we we were just um a year ago we So it's like a lot of people joined in the last 12 months.

58:17But the deployment of the technology, like what is go-to-market, what is engineering. So on the go-to-market side, working internationally across all the companies and being able to speak through the process, be a partner into how to use that technology is a lengthy and time-consuming process. You really need people there. And on the product side too, which goes back to that point of like, you can have the best foundational model, but to derive value from any of those models, you need a pretty solid product layer on top. So I'm with you on the longer timeframe. Fascinating. Like in your neck of the woods, where do you live?

58:53So I'm living between London and New York at the moment. Okay, London and New York. Like, do these conversations even exist? Like - They do, but less, you know, it's an interesting one. For NSF, they are definitely more frequent, yeah. I guess that's like, you know, the density of founders and companies and investors. It's just much higher. So I'll use that as my litmus test of whether the AI ecosystem is growing in those two. That's right. That's right. Or Waymo comes over there. Or Waymo comes over there. I heard Waymo's coming to New York or is in New York. That would be a good test for Waymo.

59:32That's about as good as it gets. That's about as good as it gets. There's WAIF in London, which is great, which is self-driving software for cars. and it works really well. Yeah, yeah. Anyway, and maybe on the headcount piece. So like, as you forecast your headcount for next year, are you gonna double? So two answers there. One is we mostly think about headcount as like a corollary to the output we want to achieve. So if we know there are specific projects we want to run or scale, then we'll increase the headcount. They are not like a goal in itself. um but as we look about kind of where we want to be in the next 12 months yes i think we will double double the head count i think i think somewhere between somewhere over 500 people in the next 12 months is is is likely as we think about scaling especially now we work very internationally so we have a lot of our clients here in us but then also across south america asia europe uh so that's going really well to lots of flanks that we need to win across and and those products especially in enterprise setting um are starting to get more complex and we want to be able to deliver to any complexity so that will require more people um and then research too like we see increasing move from from from from the models we create into how do you combine them with other modalities so we can derive that value.

1:01:01So that will also evolve a little bit. Do you sleep well? That is like a genuine question. I love it. Yeah, I think so. Like you're enjoying this. It's not like overwhelming you. So I sleep well. Caveat, the travel makes it a little bit harder because I travel a lot. And it's like I do a quarter without travel, almost a quarter with a lot of travel. And that's tricky because then you swap times on so much that even though I'm trying to like measure and get asleep in the right cadence. I don't know whether it's good for me, but just from emotional perspective, look, it's a once in a generation shift that's happening.

1:01:43AI is changing the world around us better. Regardless of the timing of a lot of those things, it's changing so much in the process work that we do and getting us smarter, quicker across the entirety of our lifetime. And we can be the voice of that change. We can be the voice of technology, which is a rare opportunity and chance that both I and I think the entire team have luck to be working on. For sure, man. You do quarterly travel. I do. I try to be in a one continent in a one quarter roughly and then maximize my travel during that quarter. And then another quarter focused on the deeper work.

1:02:23That makes sense. I appreciate you doing this. Thank you. Congrats on the startup. I'm excited for you too. Thanks, man. I appreciate it. I can't wait to see. I cannot wait to see how this unfolds. And I cannot wait to see your role in it. I cannot wait. I think regardless of how that unfolds, we need to do a test of checking whether you can actually tell the uncanny value for some of the voices. I think the quality is pretty good. Well, the amount of freaking podcasts that I've done, you have plenty of training data. There's plenty of things that you can use. You can work together. There's plenty of things that you can use.

1:02:59Are you hiring? What are you hiring for? Where are you hiring? We are hiring. We are hiring constantly. The current top roles would be helping us on a lot of the product work. So if you are excited about building on the frontier of agents, frontier of creative work, we need people to bring it to the enterprise, work deeply with the customer and figuring out how to deploy that across some of the Fortune 500s. So whether A, you are just in a great full stack or a great deployment engineer, forward deployed engineer, that's great role. Two, on the go-to-market side, similarly, I do think the role of how close you are to GPT 3.5 can be extended to other parts as well.

1:03:45But if you're excited about the technological shift and doing zero to one, then on GoToMarketSide, we love technical, highly analytical people to join the team as well. Yeah, this would be top two. I love my Cs. When you hear the word grit, what do you think of? When the hard times are there, are you still fighting and going after it and try to do the best you can to go through it? Thank you, man. Thank you. That's it for now. If you liked the episode, please leave us a review or go back into the archives where we've done more than 200 episodes with some fantastic folks. This podcast is a Kleiner Perkins production, and I'm Juven.

1:04:28Thanks for listening.