Sarvam will deploy multilingual voice AI agents throughout India in collaboration with EkStep and AI4Bharat

Artificial intelligence startup Sarvam AI has partnered with EkStep Foundation and AI4Bharat to deploy multilingual voice AI agents across the country under an initiative called “Listen at Scale.”

The project aims to transform public outreach by shifting from traditional one-way communication channels such as IVR calls and SMS to structured two-way voice conversations in local languages. By enabling conversational agents that can interact in regional dialects, the initiative seeks to make digital services more accessible to communities that often struggle with text-based interfaces.

According to a report by the Bengaluru-based startup, the programme engaged around 50 lakh unique users over a 31-day period across sectors including healthcare, agriculture and governance. The voice AI agents were used to automate several processes such as beneficiary verification, collecting citizen feedback and resolving grievances.

As part of the collaboration, EkStep Foundation provided the core platform infrastructure as well as community outreach support. Meanwhile, AI4Bharat analysed the outcomes of the programme to understand its impact and effectiveness.

Under the initiative, 20 selected teams received grants worth a total of ₹1 crore and were allocated five lakh voice AI minutes each to build and deploy their applications using the platform.

The programme has already been used in multiple public sector initiatives. For example, the National Health Authority used the system to reach more than 14 lakh senior citizens, which resulted in a 42 percent increase in daily enrolments for the Ayushman Vay Vandana Yojana.

The system was also used to profile over 4.14 lakh persons with disabilities for the Department of Empowerment of Persons with Disabilities, generating nearly 51,000 actionable profiles for assistive devices and financial assistance.

Work carried out with the Uttar Pradesh Skill Development Mission revealed that about 92.6 percent of profiled informal workers had no prior formal employment records, helping identify a largely “invisible” segment of the workforce.

The initiative also produced state-level insights. In Maharashtra, the AI system flagged inefficiencies or possible middlemen interference in about 2.75 percent of scheme delivery interactions. In Odisha, it confirmed that 77 percent of farmers had received seeds but identified a gap in input procurement adoption at around 63 percent, providing useful insights for policy improvements.

Sarvam AI recently open-sourced two reasoning models — Sarvam 30B and Sarvam 105B — as part of its effort to support broader AI development in India. The startup is backed by investors including Lightspeed Venture Partners, Peak XV Partners and Khosla Ventures.

The company is also supported by the government’s IndiaAI Mission and is preparing to raise additional capital to expand its infrastructure and scale its AI capabilities.