ElevenLabs' CEO charts path to $22 billion valuation with focus on conversational AI and disclosure

ElevenLabs, a four-year-old startup that powers the voice layer for conversational AI, has become a critical infrastructure for enterprise customer service. The company's technology runs first-line support for major firms like Klarna (handling 35 million U.S. customers), Deutsche Telekom, Cisco, and Adobe, alongside government agencies and content creators. Its backers now value it at $22 billion despite steep competition emerging in the space it helped create.

At a recent Toronto entrepreneurship conference, co-founder and CEO Mati Staniszewski reflected on the company's trajectory and where voice AI is heading. The conversation revealed a company conscious of its role in reshaping how humans interact with machines — and balancing aggressive growth against mounting ethical questions.

From commoditization to emotional intelligence#

Staniszewski walked back an earlier prediction that audio models would become commoditized within a couple of years, acknowledging the quality gap remains substantial. However, he framed the longer-term challenge differently: building systems that pass the Turing test for conversation not just through raw intelligence, but through emotional understanding. "You need to understand the emotions of the other side, to be able to slow down or speak up," he explained. That milestone, he suggested, remains years away.

The company generates $600 million in annual recurring revenue, with 55% coming from traditional enterprise deployments and the remainder split among SMBs, developers, and creators. The mix reflects ElevenLabs' expanding reach across industries.

Competing with customers#

A wrinkle complicates the growth story: customers are increasingly becoming competitors. Decagon, a conversational AI platform, built its voice product on ElevenLabs' foundation and now runs its own queries through proprietary models. Staniszewski sees this not as a threat but as an inevitable blurring of boundaries between model companies, platforms, and applications—a pattern visible across the AI sector as firms like Anthropic expand from pure research into commercial services. "In the past you'd have very clear splits where one starts and ends," he noted. "Today that line is much more blurry."

The company allows customers to choose reasoning models from a menu, and Staniszewski observed that the choice between proprietary "frontier" models and open-weight alternatives isn't binary. Informational customer service calls work fine with open models, but high-stakes scenarios—financial transactions, authentication—demand closed-source systems with stricter guardrails.