ElevenLabs, which develops AI models that convert text into human-sounding speech, is reportedly valued at $22 billion by its backers while pacing toward $600 million in annual recurring revenue (ARR), despite being four years old. Klarna uses the technology for first-line phone support serving 35 million U.S. customers, while Deutsche Telekom, Cisco, Adobe and a growing number of governments are also customers. The company sells to creators for audiobooks, dubbing and music, and increasingly competes with customers such as Decagon, which trained its voice product on ElevenLabs and now runs queries through its own models. Co-founder and CEO Mati Staniszewski said the quality gap between voice models remains significant, though he expects it to narrow over three to five years. He said ElevenLabs aims to achieve a conversational AI system capable of passing the Turing test by combining intelligence with emotional intelligence. Enterprise customers account for more than 55% of revenue, with much of the remaining 45% coming from small and medium-sized businesses, developers, builders and creators. Staniszewski said open-weight models can serve informational customer-service calls, while frontier models remain preferable for sensitive financial-services interactions where errors are less acceptable. Government deployments vary by national requirements and can use open-weight, closed-source or government fine-tuned models. In Poland, ElevenLabs is working on a healthcare deployment in which voice agents remind patients about appointments; 18% currently fail to attend, and the system preserves data residency. Staniszewski supports disclosure when callers are speaking with AI agents, although he expects public expectations to change as personal agents become common. He declined to provide detailed gross-margin information, saying the company is willing to accept lower margins to expand market share and help create customer value over the next five years. ElevenLabs uses thousands of contract workers to annotate speech, including timing, delivery, emotions and accents, with help from voice coaches. Staniszewski said the company is preparing for a possible IPO in coming years, but the timing will depend on the circumstances. He also said ElevenLabs supports precautions around frontier AI deployment, does not train text or intelligence models, does not allow agents to create more agents, and requires every customer to complete KYC (know-your-customer) checks.