India shapes ElevenLabs’ voice AI roadmap; company eyes 100-member team

BENGALURU: India is increasingly influencing how ElevenLabs researches and develops its voice AI models, with the country’s linguistic diversity throwing up challenges that are shaping the company’s products globally, its field CTO Alex Holt told TOI.Holt’s view is echoed by ElevenLabs co-founder and CEO Mati Staniszewski: “India has been at the frontier of voice experiences and part of our story at ElevenLabs from the start.”India is the company’s biggest market outside the US. ElevenLabs currently supports 14 Indian languages and is working towards adding 22 more, said Karthik Rajaram, who leads its India operations. More than 25,000 Indian developers and businesses use its API each month, up about 2.5 times over the past year.The company plans to build a team of close to 100 people in India by the end of the year, Rajaram said. Its India team has grown fivefold since he took charge at the beginning of the year, while the company currently has eight forward-deployed engineers here. It plans to hire for roles spanning sales, customer success, product and engineering. A Bengaluru office is also being finalised, although the company will remain remote-first.The growing India presence is also feeding directly into product development. In one case, a forward-deployed engineer working with an Indian customer developed a configuration that helped frontier models perform more reliably when users switched between languages within the same sentence. ElevenLabs subsequently incorporated the capability into its platform as a toggle.“This is an example of how working with customers in India improves not just their outcomes but also our products and research globally,” Holt said.The expansion comes as voice AI adoption in India moves from experimentation to production. “The biggest change has been the shift from exploration and pilot projects to deploying voice AI in production,” Holt said.Language remains a key engineering challenge, particularly pronunciation and transcription accuracy across India’s languages, dialects and accents. Holt said the company’s latest V4 model is a significant improvement over V3, but more work is needed.ElevenLabs has built its agents platform around three layers — conversation, intelligence and workflow. The speech layer handles speech-to-text, turn-taking, Indian voices and accents, Hinglish and 14 Indian languages. The intelligence layer combines the system prompt, knowledge base and LLM, while workflows connect the agent to business systems such as CRM, ticketing and telephony.For enterprises, Holt said effective voice agents need intelligence, an understanding of business processes and emotional intelligence. The platform can also route different tasks to models based on complexity and cost, allowing routine interactions to use cheaper models while sensitive tasks use more capable ones.Cost remains critical in India. Rajaram said not every use case needs the most advanced model, making the balance between capability and price important as inference becomes cheaper. “Our approach is to ensure that the value delivered justifies the price and that the unit economics work for customers,” he added.Separately, ElevenLabs has partnered with the Karnataka Innovation and Technology Society (KITS), with IT minister Priyank Kharge, to explore the adoption of Kannada voice AI in areas including skilling, citizen services and grievance redressal. The partnership will also cover responsible AI and voice restoration.You use AI every day. Now get your AI Quotient. Take the AIQ test.