ElevenLabs CEO says businesses should disclose when AI answers calls

Why it matters

ElevenLabs is moving from voice generation into customer-facing agents, where growth depends on both operating economics and whether users trust a machine that sounds human. The $600M ARR figure is company-reported, and the $22B valuation traces to July tender-offer talks, not a confirmed new financing.

ElevenLabs CEO Mati Staniszewski says businesses should tell customers when an AI agent answers their calls, even as ElevenLabs builds the voice systems that increasingly handle those conversations. In a TechCrunch interview published September 24th, Staniszewski argued that disclosure makes sense while customers still expect a person on the other end. He expects that norm to change as people grow accustomed to using their own agents.

The position puts a practical condition on ElevenLabs' push into customer service: natural-sounding speech can make automated support easier to use, but it can also make it harder for callers to know who - or what - is responding. Staniszewski said businesses could offer a choice, such as letting a customer skip a 30-minute wait for a human and speak with an agent instead. He said callers often choose the agent and are surprised by the experience.

From bad dubbing to customer calls

Staniszewski and co-founder Piotr Dabkowski started ElevenLabs in 2022 after a problem they knew from growing up in Poland: the rough dubbing of American movies. The founders set out to make content easier to experience across languages, according to ElevenLabs' account of its founding. Staniszewski had worked at Palantir before starting the company; Dabkowski had worked as a machine-learning engineer at Google.

That original language-access problem has widened into a bid to supply the voice interface for business software. ElevenLabs sells tools for speech generation, dubbing and other audio work, alongside conversational agents for companies. Its customers use the technology in customer service, while creators use it for audiobooks, dubbing and music. ElevenLabs' platform now spans creative products, APIs and agents that can operate across voice and other channels.

TechCrunch reported that Klarna uses ElevenLabs for first-line phone support serving 35 million U.S. customers, and named Deutsche Telekom, Cisco and Adobe among other customers. That deployment brings the founder's original goal - making speech understandable across languages - into a more operational setting: whether a business can resolve a customer's problem without making the interaction feel deceptive or frustrating.

Growth comes with a margin trade-off

Staniszewski told TechCrunch that ElevenLabs is pacing at $600 million in annual recurring revenue, with more than 55% of the business coming from what he called classic enterprise. Those figures come from ElevenLabs and are not audited revenue. In May, ElevenLabs said it had surpassed $500 million in ARR during the first four months of 2026, after ending 2025 at $350 million. ElevenLabs has not explained in these disclosures how its ARR calculation relates to recognized revenue.

Staniszewski declined to give detailed gross-margin figures in the interview. He said ElevenLabs is willing to let margins fall if passing savings to customers helps prove the product's value and expand market share. That is a consequential choice for ElevenLabs, which sells both its own models and the infrastructure customers use to build voice products: growth in usage does not, by itself, show how much revenue remains after inference and model costs.

The interview also surfaced the tension between being a supplier and becoming a competitor. TechCrunch noted that Decagon trained its voice product on ElevenLabs before developing models of its own. Staniszewski described the boundaries between model, platform and application companies as increasingly blurred. ElevenLabs is extending its own reach into customer workflows at the same time its customers can build more of the stack themselves.

RuntimeWire reported earlier this month that ElevenLabs was also moving down-market with Reception, a $29-a-month AI receptionist. The launch gives ElevenLabs another route to sell conversational agents beyond large enterprise contracts, where deploying the technology depends on integrations, testing and oversight. In September, ElevenLabs also hired Ashley Kramer as its first revenue chief as it builds out enterprise sales and partnerships.

The $22 billion figure is not a new round

The valuation in TechCrunch's headline has a separate timeline from the interview. Bloomberg reported on July 2nd that ElevenLabs was in talks for a tender offer at a $22 billion valuation. That report was 84 days old on September 24th; it does not establish that a deal closed at that price. ElevenLabs' last publicly announced primary financing before the tender-offer report was its February 4th Series D: $500 million led by Sequoia Capital at an $11 billion valuation, according to the company's announcement.

A tender offer provides liquidity for existing shareholders; a primary financing puts new capital into ElevenLabs. ElevenLabs said in May that it had closed a $100 million tender offer alongside its Series D, but that announcement did not confirm the reported $22 billion terms. The latest interview therefore combines a current operating claim with an older, unconfirmed secondary-market valuation report.

Staniszewski's bet is that voice agents will become useful enough for customers to choose them, and recognizable enough that companies can disclose their use without losing that choice. ElevenLabs' reported scale gives it a large base from which to test the premise. The unresolved measure is whether customers keep using the agents because the conversations work - and whether ElevenLabs can support that expansion without margins becoming the cost of winning the market.