Hello and welcome to this special conversation. AI is learning to speak, and ElevenLabs is one of the companies that is betting big that voice is going to become one of the most important interfaces between humans and AI. The company started by building synthetic speech which can sound remarkably human, and today its AI agents can listen, speak, understand and act, handling more than 15 million conversations every week across customer service, insurance, healthcare and other businesses. And the scale of ElevenLabs is quite striking. It was founded in 2022 and the company is now already valued at $22 billion, it has more than 800 employees, and says enterprise customers now account for more than half of its revenues. Its technology is today being used by some of the world's largest tech, insurance and telecom companies. India, of course, is a very important part of that story too. ElevenLabs says India is its largest market by sign-ups and the second largest by enterprise revenue.
Mati Staniszewski is the co-founder and CEO of ElevenLabs and he joins me today in Bengaluru on the sidelines of the ElevenLabs summit here. Mati, thank you so very much for joining us here. Let me just start with the India market itself. The last time you spoke to us you said $100 million ARR, and you're doubling down in Bengaluru with your hiring as well. How far along are you in that number, and what particularly excites you about the market here today?
Thank you. Thanks for having me here and thanks for the conversation. It's a special day for us to start and host the ElevenLabs summit here in Bengaluru. Especially because for us as a company, so many of the use cases started here in India. As a company, we're solving a lot of the way that people and technology interact, and it requires high-quality voice and voice conversation. And India is a voice-first country. And as we met the previous time, so many of these use cases were growing, and today we are seeing many more of them. And as a company, our total ARR — we are now over $300 million in ARR.
India is our second biggest market after the US. We have an incredible team — a team that spans go-to-market, engineering, research, forward-deployed engineering — so truly a cross-functional element to be there with customers to deploy the technology, to build and customize the models to make sure that they speak all the Indian languages, across all the dialects. And since we last spoke, we released Eleven v3, our new generation model that now supports 14 of those languages. So a really big step change since we last met.
Well, you know, you said India is a voice-first country. I mean, we prefer speaking to writing. There are more than 2,000 languages for 1.4 billion people. Could it really be the laboratory of the world for voice AI in the future? Could the first interface between humans and AI primarily become voice? Could India be a test case?
I think 100% yes. And we are already seeing that. Some of these use cases are starting here, and the rest of the world can get inspired by what's starting here. And that spans from companies that are innovating not only on the commercial side. PhysicsWallah, in education, is looking at how to make sure that every person could have their personal teacher in the future — a teacher that understands your context, understands your knowledge, that you can ask a question late at night or early in the morning, and converse and learn through a voice conversation.
Companies like Cars24 or Razorpay are already leading with innovation, where they not only do the customer care, customer support — where you have immediate help with your claim or with your problem to get it resolved — they can follow up. They are now doing part of the sales workflow, where you have an AI agent working as part of that. And it's an AI agent that works in a symbiosis with a human agent, where they both collaborate on closing the deal. And that's already happening. So that's not even a future claim — that's already happening, and we are seeing the rest of the world being inspired by that.
And maybe a last one, just a good example on a very different space, is Havells — how you can effectively automate your home, how you can interact with your home, whether it's to turn down or up the AC, and make that a lot simpler. They are already building that in their app, which now has 2.7 million downloads.
Well, there's a whole lot of private clients that you already have here in India, but we understand you're also signing an MoU with the government of Karnataka. If you could tell us a bit about that, and if the government portion of it could become a larger part of your business.
So today I'm very excited about signing the MoU with the government of Karnataka. It's a big step for us because we believe, for the technology to actually provide value and to amplify human potential, it needs to be in the hands of people. It needs to diffuse across the economy. So we are working on a number of different segments. The first one is all the upskilling, all the work that's required to make sure all the people can have access to the tooling, to the technology that is needed. And that's the first and a big step in that MoU, and that helps just bring a lot of the tooling across.
You know, what does all of this mean in terms of your financials? And I'm talking about the kind of investments you're making for a market like India. What's your hiring plan here? What's the growth path you see in the next few years in the market?
Yeah, and maybe, you know, for us frequently we look at it — if we can help provide impact, help provide value, the revenue is going to come anyways. So we take a long-term view. We want to build a company that will be here for the decades to come. And even as we think about the MoU, a big part of that is bringing services to all the people that lost their voice due to ALS or due to cancer and can now get it back thanks to AI voice, and speak again. People that are visually impaired and need to access the web in any way — those people too will be able to have a voice. So we are aiming to provide a million voices within that commitment, and all of that is non-revenue-generating. And here we know this work is actually helping bring that potential.
Now, to the second part of your question — I think we're just touching the surface. So we are less than 1% of the journey there in terms of the commercial impact we can have. And we are seeing that with the companies and how they deploy the technology. We are seeing that on the revenue numbers. The growth has just started. Recently we had, for the first time, the enterprise revenue crossing the self-serve revenue at 55% of the total. And that's partly a reflection of where the technology is — over the last year it became reliable to be able to deploy it in that setting. So as you think about the investment in the region, we'll look to build the team and a presence, a local presence, to make sure we build everything from the models that are optimized for the space to people that are optimized and work with the customers in a given domain. So that's all going to happen.
You wouldn't put a number to the investment or the hiring plans?
Well, it's the... can you be more specific on the investment — which part do you mean?
In the investment you're looking to make into India, yeah.
I mean, it's, you know, it's a...
I mean, I saw your slide there in your presentation — you were talking about making investments in India. 15 million conversations weekly that the company's already handling. But what does that mean monetarily? I mean, for a $22 billion company.
So of course, that's the valuation number. In terms of our budget, we raised close to a billion, and we are spending a big portion of that on building the models. And as you've seen, the recent model release had one of the biggest areas of coverage in Indian languages. So as you think about it, it's a double-digit percent of the total amount — so in hundreds of millions of dollars as we think about the spend that we want to make sure is here in the region.
But I think the more important part is, as you've seen in the past, it can never be an afterthought. And as you think about building the models, even when we started, v1, the first model we released, Hindi was a big part of that model. The architecture was optimized from the very start to support those languages. And in the past, approaches before we started, companies would do, let's say, English and then bring it to other languages. We didn't like this idea. We think you really need to build locally, and that means not only the team, it also means the architecture. So you can spend as much money as you want, but if it's a second thing you do, then it's not going to support you in the same way. So that was a big part for us.
Well, you know, speaking about the Indian market, and I'm sure you've been asked this several times before — there are competitors like Sarvam that are doing precisely this. What is your moat then, or rather, what is your USP when you're selling to Indian enterprises? Why should they come to ElevenLabs versus a company that's building in India for India?
Yeah, and Sarvam is a wonderful company. I think they're doing great work. And in general, we like competition. We like that there is a good selection of different models, and usually what happens is different models will be optimized and good for different use cases. Sarvam, of course, spans a lot of modalities — we focus on voice and audio in terms of the model work. Happily, our recent model is topping the benchmarks, especially across the Indian languages, and is the best model there. And that's thanks to the partnership with a lot of partners here too.
But as you ask about the moat, and as we think about the wider AI space, one thing is building the models and the second is bringing them to production. And that's the big part of what we sell. So we build models, but we also build a product to help you orchestrate those models, connect that with your configuration, with your knowledge, with your deployment, to actually bring that, test that, monitor that in the workflow. And as you think about the long term, we are building an ecosystem around that work. And concretely, what the ecosystem needs is people contributing simulation tests, contributing integration and voices. So today we have in our voice marketplace 50,000 voices that got created, authenticated, and every person earns compensation as a result. And thousands of them are here in India, where people here in India are earning compensation as a result every time that voice is used. And that's a different model than any company in the world — whether it's the global companies or local.
Well, hold your thought. You know, I asked you about the moat because you've had this very interesting example of Dec Gone — I mean, it's a client that's now a competitor. So what happens if the OpenAIs, the Geminis, the Anthropics of the world build this voice agent as part of their bundle? What remains your moat? Are you going to be the infrastructure layer, or how do you compete then?
And you know, there's of course a number of companies, but not many companies that are actual labs would build their own models. And that's kudos to Sarvam, who are actually building, are actually competing on that stage and are building something special — but a lot of companies don't have that combination. So look, I think the first piece is already happening: OpenAI, Google, all those companies are already building voice models and voice agents. And we are happy to compete with them on research and out-compete them on research. And that's — we have my co-founder, who's an incredible researcher and has been able to assemble a team of 50 people, a good amount of them here in India too, who are building the models that are better than any of the companies that you mentioned.
But ultimately, the main thing as you think about any deployment is the focus. And we are squarely focused on the interaction layer. We want to be the interaction layer for the agent tech economy, and that takes modern models, that takes integrations, and that takes that ecosystem and deployment layer that we combine as the only company with that focus.
All right, take that point, Mati. But let me talk about one of the repercussions of building this voice AI for a market like India. We're big on the BPO market, for instance, which is purely voice-driven voice calls. Something like what you're building could easily replace hundreds of millions of jobs. How do you see that playing out? Do you think AI could help actually create newer jobs for that segment? Or would there be a bit of a displacement period, which we're already starting to see?
