\n<\/p>\n
ElevenLabs builds the voice layer of AI, the models that turn text into speech that sounds human. Most people encounter it when they\u2019re talking with customer service \u2014 often without realizing it. Klarna runs first-line phone support for 35 million U.S. customers on it, for example. So do Deutsche Telekom, Cisco, Adobe, and a growing list of governments. ElevenLabs also sells to creators, who use its platform for audiobooks, dubbing, and music.<\/p>\n
It\u2019s not alone in what it does. In fact, it\u2019s increasingly bumping into customers, including Decagon, a conversational AI platform that trained its voice product on ElevenLabs and now competes with it. But its investors don\u2019t seem too concerned. The company, which says it\u2019s pacing at $600 million in annual recurring revenue, is now reportedly valued at $22 billion<\/a> by its backers, despite being just four years old.<\/p>\n To understand more, I interviewed ElevenLabs co-founder and CEO, Mati Staniszewski, at Nrth<\/a> in Toronto, a local entrepreneurship conference formerly known as Elevate. We covered a range of topics in a short time, including whether businesses should tell customers when they\u2019re talking to an AI (he thinks they should), and whether he could discuss the company\u2019s gross margins. Unsurprisingly, Staniszewski said he couldn\u2019t discuss them in any detail, but he was clear that he doesn\u2019t mind them getting squeezed even further if it means expanding the company\u2019s market share.<\/p>\n Our conversation follows, condensed and lightly edited. You can check out the full conversation here<\/a>.<\/p>\n You joined us at TechCrunch Disrupt last year, where you said audio models would be commoditized within a couple of years<\/a>. How would you rate that prediction now?<\/strong><\/p>\n There is still a lot of work to be done, and the quality delta you can achieve just on the model level is still significant. If we think longer term, probably three, five years from now, those differences will be smaller. What we\u2019d love to do, and be the first ones to do, is pass the Turing test for conversational AI. You need to combine intelligence, but you also need emotional intelligence. You need to understand the emotions of the other side, to be able to slow down or speak up. That hasn\u2019t yet been done.<\/p>\n What percentage of the business is enterprise now?<\/strong><\/p>\n We are now $600 million in ARR. Fifty-five percent plus is classic enterprise, and a big percentage of the [remaining] 45% are small and medium businesses, developers, builders, creators.<\/p>\n Like everyone else in AI, you\u2019re increasingly competing with your customers. Decagon trained its voice product on you and now runs queries through its own models.<\/strong><\/p>\n The lines become more blurry. As we think about model companies, platform companies, application companies, in the past you\u2019d have very clear splits where one starts and ends. Today that line is much more blurry. In Anthropic\u2019s case, what was a model company is definitely a platform and increasingly a wide set of applications. I think this will continue.<\/p>\n Your customers can choose the \u201creasoning layer\u201d from a menu of options at ElevenLabs. What are you seeing in terms of the use of frontier lab models versus open weight?<\/strong><\/p>\n It\u2019s less of a binary choice. In customer experience, if you\u2019re calling in and it\u2019s just informational, you\u2019re not executing any actions \u2014 you can use a lot of the open source models because your knowledge base defines what a good experience is. But if it\u2019s financial services, you want to be authenticated, you want information about a transaction, maybe a refund. There\u2019s no room for error. Here, frontier models will still lead.<\/p>\n Some of those open-weight models are Chinese. The U.S. government is a customer. European governments are customers. What are those conversations like?<\/strong><\/p>\n Different. In each deployment, the models and the voices we deploy will depend on the case. If we work with the Polish government or the Brazilian government, they have their own set of requirements. It can be an open-weight model, a closed-source model, their own fine-tuned model. [In Poland] it\u2019s a healthcare case. You have patients booking appointments across the public health system, and 18% never show up. The deployment is agents that call and remind you. They had a set of models optimized on their knowledge, and we integrate while keeping data residency.<\/p>\n Should businesses disclose when someone is talking to an agent rather than a human?<\/strong><\/p>\n I think there should be disclosure at this time. Currently, people aren\u2019t used to it, and the common pattern is you don\u2019t want to feel cheated on that call. But in five years, when everybody has their own agent working on their behalf, you\u2019ll be calling in and expecting an agent. Then I think we\u2019ll shift as a society. There are good ways of doing it \u2014 if there\u2019s a 30-minute wait for a human, offer the customer a choice. In almost all cases, they choose the agent and then they\u2019re surprised by how good the experience is.<\/p>\n What are your gross margins, given what you\u2019re paying for models and inference?<\/strong><\/p>\n I\u2019m going to give a vague answer. Given we have that research element, we\u2019re able to fine-tune and constrain models in extremely smart ways. But if we can pass on any savings to the customer, we do that. The biggest thing is still proving the value and being there with the customer. So if we can invest and prove that value, we don\u2019t mind the margins going lower to actually benefit together as the value gets created in the next five years.<\/p>\n You have millions of hours of customer service calls. Do you train on them? How much of your training data is synthetic?<\/strong><\/p>\n In certain companies, we created the models together. They wanted a specific model for their use case. Otherwise, the big part of the training hasn\u2019t been so much the volume of data, it was annotating the data. We have thousands of people internally on a contracting basis helping us annotate not only what was said, but when people were speaking, how they said things, what emotions were used. We had to bring voice coaches in to be able to detect accents accurately.<\/p>\n It\u2019s been reported you\u2019re looking at 2028 for an IPO. Can you confirm that?<\/strong><\/p>\n We\u2019d love to create a company that stands the test of time. We are preparing the foundation to be able to do it in the next years. But whether we do it will depend on the time and place.<\/p>\n \u201cYears\u201d is very vague.<\/strong><\/p>\n [Laughs.]<\/p>\n Backstage: Where do you land on whether the frontier labs should slow down?<\/strong><\/p>\n Everybody is aligned to work together on finding a way to pace. Whether they should be public about it, and how much of the media conversation or regulation it should involve, that\u2019s another topic. But yes, we should all take the right precautions as we deploy the technology. We don\u2019t train the text models and the intelligence side of models, which is the core key of the debate.<\/p>\n Could ElevenLabs be exposed the way Hugging Face was?<\/strong><\/p>\n We\u2019re a step further, because we don\u2019t deploy self-replicating or recurrent parts of the intelligence of agents. Our technology doesn\u2019t allow you to let agents create more agents. Every customer goes through KYC. Cybersecurity risk is definitely a risk for the wider world, but we have a good set of precautions in place.<\/p>\n<\/div>\n When you purchase through links in our articles, we may earn a small commission<\/a>. This doesn\u2019t affect our editorial independence.<\/em><\/p>\n