ElevenLabs, the AI voice company has turned India into its largest enterprise market outside the US. We sat down with Karthik Rajaram to find out how. Then there’s the story of the rise of artificial intelligence, all happening in San Francisco, shipping in English, and slowly spreading to the rest of the globe. ElevenLabs is narrating a different story.
Also read: Use NotebookLM beyond studying: 5 cool ways to get more for free
In the third month since assuming responsibility as India’s Country Leader, Karthik Rajaram exudes the confidence that comes from recently reviewing the numbers. “India,” he declares without hesitation, “is, interestingly, the the largest enterprise market for ElevenLabs outside of the US, the fastest growing rate. And when I’m talking the largest market, I’m not talking number of users and the other metrics that typically people talk about, but I’m talking enterprise revenue.” The case begins with a straightforward insight into interfaces. All touchscreen devices, all applications, all onboarding procedures assume literacy, comfort with technology, and likely not a native tongue. The voice removes all barriers.”For large parts of humankind’s existence, we did not type, swipe, or click,” Rajaram says. “We just spoke and commerce happened.”
It’s hard to talk enterprise AI in India right now without addressing the nationalism wave Sarvam is riding. We asked Rajaram whether there is resistance when marketing a non-Indian product to the Indian enterprise.
“A resounding no,” he says. “If that was the case, we would not have seen the kind of growth that has made India our largest enterprise market outside the US.”
He makes it clear that enterprises make purchasing decisions not by waving flags. Instead, it is all about results: the success rate, handling time, costs, and if the agent keeps its promise. That is how ElevenLabs is marketing itself, with live projects from Cars24 (400 AI agents doing 30 percent of the incoming calls), Meesho (90 percent success rate and 75 percent reduced costs), Razorpay, and TVS Motors in 25 countries speaking nine languages.
Also read: OpenAI lets you add an emergency contact in ChatGPT: Here’s what that actually means
Coverage is the obvious India challenge, and ElevenLabs supports 11 Indian languages with a target of all 22. However, Rajaram sounds far more enthusiastic about the next phase beyond coverage, emotional inference.
“Let’s say somebody says ‘I can’t pay the EMI this month’ in a flat resigned tone versus someone saying it with anxiety and frustration, that is very different. The workflows are different, the way you want to handle that is different.”
In his opinion, understanding the emotional content and propagating it through the whole pipeline in order to have a real empathic response is what makes it unique and something other voice AI products still fail to achieve. There are technical challenges as well because India uses low-end Androids in noisy settings. The Razorpay experience in tier 2, 3, and 4 cities demonstrates that the product works in less than optimal conditions.
Voice cloning in India isn’t theoretical. Political audio manipulation is documented and ongoing. Rajaram doesn’t soften it.
“The concerns are very, very real, very valid. Anybody who trivializes them is being naïve and irresponsible.”
ElevenLabs’ response is built around three principles, moderation, accountability, and provenance. Voice cloning requires explicit opt-in, backed by voice CAPTCHAs. End-to-end traceability means bad actors can be identified and permanently banned. A proprietary speech classifier detects AI-generated audio at 99 percent precision. Every agent deployed on the platform must identify itself as AI.
“Everybody has the right to know whether they’re talking to AI or a human,” Rajaram says. It’s framed as a design principle, not a compliance checkbox – built into the platform architecture from day one rather than bolted on later.
The Turing test question, what if an agent sounds so human nobody can tell, gets a telling answer. “We’ve shifted from how AI sounds to what can AI do,” Rajaram says. Can the agent act, understand, take action grounded in the company’s data and workflows? Human-sounding, he argues, is already solved. The real game is whether the agent can do the job.
Voice is how India has always done business. The AI just finally caught up.
Also read: Claude Mythos found decade old Firefox bugs that years of fuzzing missed