In a few days, heads of state, industry, technology and non-profit leaders, and researchers from all over the world will gather in New Delhi for the international Artificial Intelligence (AI) Summit. This event marks a turning point in how the world thinks about the technology. It’s the first global AI summit to be held in the Global South, and the first to focus on AI’s impact on society.
There are reasons to be excited about AI. This rapidly improving, multi-purpose technology is changing how we live, learn, work, and much more. While the arms race among Big Tech firms to develop LLMs dominates the headlines, governments, social enterprises, and non-profits worldwide are pouring resources into AI applications because of its potential to improve the delivery of essential public programmes. We see several pathways through which AI can improve social outcomes when deployed thoughtfully. AI can help ensure programmes target the right people, provide customised and actionable information, maximise the productivity of frontline workers, increase organisational efficiency, reduce bias in decision-making, and boost resource mobilisation for governments.
Thanks to a deep bench of technology talent and the government’s foresight in building world-class digital public infrastructure as well as large investments in computing power and dedicated research hubs, India is uniquely positioned as a global AI powerhouse. As AI becomes integral to social programmes worldwide, India’s experience in using cutting-edge technology for underserved communities holds important lessons for the world.
One of the most important lessons: AI tools must be tested with the same rigour as any development intervention to understand their true impact. India is on track to establish itself as the “AI application capital” of the world. It also has the opportunity to become the “AI evaluation capital” of the world to understand what works, what doesn’t, and why.
Consider better targeting. The government of Bihar is partnering with J-PAL affiliated researchers, Google.org, and a local NGO, Yugantar, to evaluate strategies for sending AI-powered flood forecasts to households. Early results from the ongoing evaluation indicate that training local community volunteers to help disseminate the flood alerts increased access to forecasts, trust in the forecasts, and the likelihood of taking precautionary measures. In Tamil Nadu, J-PAL affiliated researchers are working with the government to test the impact of a low-cost, AI-based tool to identify people at high risk of heart disease.
Second, AI applications must be designed to work in the messiness of real-world conditions. In Karnataka, fingerprint scanners did not improve doctor attendance in state-run primary healthcare centres because the government was unable to penalise absent doctors without risking further shortages in rural areas. Based on this, the state chose not to scale the intervention.
Third, pairing AI with human skills can unlock the technology’s full potential. An evaluation of a computer-based software that delivered personalised instruction in state-run schools in Rajasthan found substantial gains in both teacher productivity and learning outcomes. These gains depended in part on a facilitator who ensured students used the software. With less supervision, student engagement fell, and overall learning gains declined. But the software itself remained effective when used.
The common theme here is that technology matters as much as institutional, behavioural and field realities. Over the past two decades, promising technologies often fell short when context, delivery, and human incentives were overlooked — from laptops in classrooms to smokeless cookstoves to efforts at reducing indoor air pollution. Investing in evidence to guide AI investment and scale-up decisions will help governments, funders and NGOs direct resources where they have the greatest potential to maximise social good, while limiting unintended harms.
The India AI Impact Summit 2026 will showcase the ingenuity and skill of India’s tech workforce, civil society organisations and universities, and the government’s vision — first in building a strong digital public infrastructure, and now in AI. By placing impact at the heart of the summit, India is sending a clear message: AI must be judged by its ability to bring measurable improvements in people’s lives. Truly achieving this vision will also require rigorous and concurrent field evaluations, and sharing the lessons. If done well, India will earn its place as a vishwaguru in AI impact.
Dhaliwal is the global executive director, J-PAL at the Massachusetts Institute of Technology. Mukerji is the executive director, J-PAL South Asia at IFMR
