Aadhaar gave India a way to verify identity through DPI. UPI gave it a way to move money. David Menezes, Director at EkStep Foundation and People+AI, told StratNewsGlobal that his organisation is now trying to build the equivalent layer for artificial intelligence(AI) — public data platforms that any institution, startup, or government department can build on top of, instead of proprietary AI tools controlled by a handful of companies.
Menezes traced this thinking back to EkStep’s earlier work on India ‘s Digital Public Infrastructure (DPI), including the education platform DIKSHA, which the foundation built and handed over to the Government of India in 2021. When generative AI became widely available in 2023, EkStep applied the same logic to new sectors. In education, that produced ASL, an assisted math and literacy programme running in government schools in Telangana and Karnataka, where a computer listens to a child read aloud, flags pronunciation issues, and adjusts the difficulty of subsequent passages. Menezes said the system was built to run offline and on local servers so it could function in schools with limited connectivity.
In agriculture, the same principle produced MahaVISTAAR, built with the Government of Maharashtra. It initially launched as a smartphone application, and Menezes said usage only accelerated once it was redesigned as a voice-based feature-phone helpline, taking daily queries from a few hundred to roughly 10,000. Bharat-VISTAAR, launched at the national level, is now being built to complement that work by helping farmers across every state access central government schemes through a single helpline, while state tools continue to handle region-specific advisory on weather, pesticides, and market prices.
The same rails extended to dairy. Sarlaben, an AI advisory tool built for Amul following a request from Prime Minister Modi, went live in three weeks, compared to the nine months MahaVISTAAR had taken. Menezes attributed the shorter timeline to Amul’s cooperative data being already well-organised, and to safety guardrails and architecture carried over directly from the Maharashtra deployment.
Each of these systems is built with government-authorized data rather than open internet sources, so farmers know which institution to approach if a response is inaccurate.
Menezes estimated that the AI model itself accounts for roughly 30 per cent of a deployment’s success, with the remaining 70 per cent tied to data quality, institutional backing, and trust.
He added that the largest bottleneck in India’s AI rollout is the number of pilots operating without contact with one another. His response is 100 Pathways, a People+AI initiative now active across twelve countries, including Ethiopia, which documents deployment know-how — including safety guardrails and data architecture — for other institutions to reuse.
Watch the whole interview here.
