Indian artificial intelligence startup Sarvam AI has partnered with EkStep Foundation and AI4Bharat to deploy multilingual voice-based AI agents across India. The collaboration aims to build large-scale conversational AI systems capable of interacting with citizens in multiple Indian languages, helping bridge the country’s digital divide and expand access to digital services.
The initiative focuses on creating voice AI agents that can communicate naturally with users in their native languages. These agents are designed to interact through voice calls and conversational interfaces, allowing people who are not comfortable with text-based apps or English-only platforms to access digital services more easily. The system can help citizens ask questions, provide feedback, verify information, and interact with public services using simple voice conversations.
The collaboration was implemented through a program known as “Listen at Scale,” which tested how voice AI can function at population scale. Over a 31-day deployment period, the program worked with 20 organizations, including government agencies and non-profits, to integrate multilingual AI agents into real-world workflows. These agents were used for applications such as citizen outreach, beneficiary verification, grievance handling, and information dissemination.
The project demonstrated significant reach during its early phase. According to the program’s report, the voice AI systems consumed more than 74 lakh minutes of conversational interaction and connected with around 50 lakh unique users across different regions of India. The deployment showed that voice-based AI could engage large populations efficiently, especially in rural or low-literacy areas where traditional digital interfaces often struggle to reach users.
In the partnership, Sarvam serves as the primary technology provider, delivering the AI models and infrastructure needed to build and deploy voice agents. Its platform supports conversational AI capabilities that can understand Indian accents, languages, and contextual nuances, enabling more natural interactions for diverse users. The company has been developing AI models specifically designed for Indian languages and local use cases.
EkStep acts as the ecosystem partner and program host, providing infrastructure, outreach support, and funding assistance for participating teams building voice AI solutions. Meanwhile, AI4Bharat serves as the research and knowledge partner responsible for studying the deployment results and developing frameworks that can expand the use of multilingual voice AI technologies across the country.
The initiative highlights the growing importance of localized artificial intelligence systems in India. With hundreds of languages and dialects spoken across the country, many global AI systems struggle to accurately understand regional speech patterns and code-switching between languages. Locally developed models that are trained on Indian languages and speech data are expected to play a crucial role in expanding digital inclusion.
Voice-based AI is increasingly being explored as a key interface for reaching populations that may not be comfortable with smartphones, typing, or English-based digital services. By enabling users to interact through natural conversation in their own languages, voice AI systems could make it easier for millions of people to access government programs, healthcare services, financial tools, and educational resources.
Industry experts say initiatives like this could help India develop a sovereign AI ecosystem tailored to the country’s linguistic diversity and social needs. By combining research expertise, nonprofit digital infrastructure, and startup innovation, the partnership aims to build scalable AI solutions that can operate effectively across India’s complex language landscape.
