NHA reached 1.4 million citizens to drive scheme awareness using Sarvam’s voice agents
How Sarvam AI partnered with NHA to drive Ayushman Vay Vandana Yojana awareness across 5 languages, reaching 1.4 million citizens in 6 days.
Background
The National Health Authority (NHA) runs one of the world’s largest public health insurance programs, Pradhan Mantri Jan Arogya Yojana (PM-JAY), serving hundreds of millions of people across the country.
Reaching people at that scale is not straightforward. Awareness remains uneven, especially for newer schemes. Traditional outreach methods break down under these conditions. Call centers are expensive to scale. Digital channels miss those without consistent access to smartphones. Supporting many languages at once stretches teams and systems.
In several states, awareness remains a constraint, particularly for newer schemes like Ayushman Vay Vandana Yojana. When Sarvam partnered with NHA on a voice-based outreach model, it was an opportunity to test whether these constraints could be addressed in practice.
Challenge
For this pilot focused on Ayushman Vay Vandana Yojana awareness, NHA faced several challenges:
- Limited scale and reach: Manual operations cannot feasibly reach millions of beneficiaries with personalized information in their preferred language within compressed timeframes. Traditional awareness campaigns are costly, inconsistent, and often ineffective at penetrating rural and semi-urban areas.
- High cost-to-serve: Each human-led interaction carries significant operational costs, making population-scale one-to-one outreach economically prohibitive.
- Digital divide constraints: A significant portion of the target population lacks smartphones, reliable internet, or digital literacy - excluding them from app-based outreach, SMS campaigns, and web portals.
- Inconsistent quality: Variation in agent training and language proficiency leads to inconsistent beneficiary experiences across geographies.
- Capacity constraints during peak demand: Call volume spikes during campaigns overwhelm agent capacity, leading to long wait times and abandoned calls.
- Accessibility barriers for senior citizens. Senior citizens, a key demographic for Ayushman Vay Vandana Yojana, cannot engage with unidirectional channels like SMS or IVR recordings that don't adapt to their questions.
- Limited real-time analytics. Manual operations provide limited visibility into citizen sentiment, pain points, and engagement patterns needed to optimize campaigns.
Solution
Sarvam’s voice agents addressed these constraints directly. Voice removes the need for smartphones or digital fluency, and when delivered in a person’s own language, it feels familiar and trustworthy. From initial conversations to rollout, the program went live in under two weeks.
Sarvam’s Samvaad platform works on basic mobile phones, with no internet or smartphone required. It supports natural conversation across 11 Indian languages, adapting to how people actually speak across accents, dialects, and speech patterns. Instead of asking citizens to navigate rigid IVR menus, the system meets them where they are.
Deployment from concept to live campaign took under two weeks through close coordination between teams. The use case was shaped jointly based on NHA's understanding of beneficiary needs, messaging was refined iteratively, and outcomes were reviewed continuously. Sarvam's Samvaad platform architecture enables rapid configuration without months of custom development.
The solution runs on India's sovereign AI stack, addressing data residency and security requirements critical for healthcare programs.
Impact
Reach & Engagement
Over 1.4 million unique citizens were reached in six days through 2.57 million outreach attempts, achieving a 53% connectivity rate. Engagement quality proved equally important: 84% of connected citizens stayed engaged for more than 15 seconds, with an average engagement time of 45 seconds per call.
Among connected users, 410,000 citizens expressed interest in learning more (28% interest rate), and 230,000 indicated willingness to enroll (16% of connected users).
Enrollment Impact
The campaign corresponded with a 42% increase in average daily enrollments in target states compared to the previous 30 days—rising from 8,700 to 12,400 daily enrollments. Total enrollments during the campaign week increased from 57,100 to 74,400. Week-over-week growth reached 30%.
Geographic variation provided additional insight: Uttar Pradesh saw a 53% increase in daily enrollments, while Punjab experienced a 112% increase. While enrollment outcomes reflect correlation rather than direct causation, the timing and magnitude of increases - combined with 230,000 citizens expressing enrollment intent - suggests the campaign played a meaningful role.
Operational Validation
The platform's real-time analytics enabled continuous monitoring throughout the campaign. Multilingual support operated reliably across 5 languages (Hindi, Kannada, Marathi, Odia, and Punjabi) without separate QA processes. The sovereign infrastructure met compliance requirements while performing at the scale needed for national campaigns.
Conclusion
The pilot shows what changes when outreach is built for how people actually communicate. Voice agents, running on sovereign infrastructure and designed for India’s language diversity, can work through real constraints at population scale. They remove the dependency on smartphones, avoid fragmentation across languages, and can be deployed quickly enough to match how government programs move. They also introduce a level of visibility that traditional outreach does not provide.