IIT Kanpur’s AI solution could help governments identify households needing welfare support

IIT Kanpur has developed a cost-effective solution to help governments identify households in need of welfare support by using data already available in existing government systems.

The institute’s AI-powered ‘Family Score system’ assesses the socio-economic conditions of households without requiring additional surveys, offering a faster and more cost-effective approach to welfare identification at scale.

The first phase findings have been presented to the Andhra Pradesh government.

In collaboration with the government of Andhra Pradesh, Phase 1 tested whether AI could provide a reliable alternative to conventional household surveys for assessing socio-economic conditions.

The findings indicate that an ensemble of AI models developed by the IIT can outperform conventional surveys in determining the socio-economic conditions of households, while requiring minimal cost and effort from the government.

This approach addresses a key challenge in welfare delivery, identifying the right families at scale without the time, manpower and resources required for large-scale household surveys.

“The findings were presented to the Chief Minister of Andhra Pradesh, N. Chandrababu Naidu, and senior government officials by representatives of IIT Kanpur. The Chief Minister and the Andhra Pradesh government team appreciated the capabilities demonstrated by the system and its potential to help identify and reach the right families with targeted welfare support,” an official statement by the institute read.

The AI system uses existing government data to assess the socio-economic well-being of families and classify them into defined categories. It uses an ensemble of approximately 15 government-owned datasets which provide indicators of household socio-economic conditions.

IIT Kanpur Director Manindra Agrawal said, “Family Score is an effort to use AI and existing government data to better understand the socio-economic conditions of families and enable more targeted welfare delivery. “The initial results are encouraging and demonstrate the potential of this approach to reduce the dependence on costly and time-consuming surveys. Our collaboration with the Government of Andhra Pradesh reflects our aim to put rigorous and responsible AI in service of citizens, while ensuring that the process remains transparent and open.”

Nitin Saxena, Dean, Wadhwani School of AI & Intelligent Systems, IIT Kanpur, said, "The strength of this system is not just its accuracy but its interpretability. Every score comes with a reason, so officials remain fully in control of the decision. That is what responsible AI for governance should look like."

Manish Srivastava, CEO, Airawat Research Foundation, said, "The team from the AI CoE for Sustainable Cities presented AirawatOS — an operating system for a city. It unifies data from satellites, sensors, records and field reports into one trusted, live picture, and uses AI to turn it into evidence-backed, traceable recommendations — while people stay in charge of every decision. Rather than replacing any department's authority, it connects them: each agency keeps control of its own data, while a shared network lets them coordinate securely across floods, permits, and projects that span many institutions."

The current coverage in the government-owned datasets indicates that family scores can be generated for roughly 1.35 crore of the 1.72 crore (75%) total households in Andhra Pradesh.