On the second day of the India AI Impact Summit 2026 a session on artificial intelligence in governance convened global researchers and senior policymakers to examine how AI can strengthen public service delivery at scale. The discussion emphasised the need to move beyond pilots and promise towards measurable impact and focused on rigorous evaluation responsible deployment and systems level readiness across government. Speakers argued that defined operational use cases should precede system wide transformation. The session opened with a research presentation that examined machine learning driven targeting of public service delivery in Togo and the study showed measurable improvements in food security mental health and socio economic indicators where AI supported targeting had been applied. At the same time the research revealed limitations as phone metadata alone failed to capture treatment effects and exposed model drift and the difficulty of predicting short term vulnerability. Presenters recommended iterative testing and scientific validation as prerequisites for scale. The discussion then assessed India's preparedness for government level AI implementation noting progress in building compute capacity breaking down data silos and upskilling public sector employees while flagging persistent challenges around scalability data heterogeneity and ethical clarity. It was noted that only a small proportion of AI implementers fully understand their ethical frameworks which creates a governance gap that must be addressed alongside technical capability. Officials pointed to areas such as banking and financial systems where AI adoption has advanced due to the availability of high quality structured data enabling analytics for tax and expenditure tracking. Speakers urged careful design of experiments independent audits and strong validation protocols to build trust and enable scaling particularly when systems are used to identify beneficiaries or allocate resources. The central takeaway was that AI can enhance government efficiency only when grounded in robust data scientific validation and responsible governance frameworks.