New Delhi, Feb 17 (.) The role of artificial intelligence in improving government efficiency and public service delivery came under discussion at the India AI Impact Summit 2026 on Tuesday, with experts calling for rigorous evaluation and stronger governance frameworks before large-scale deployment.
The session titled ‘AI in Governance: Revolutionising Government Efficiency’ brought together researchers and policymakers to assess how AI tools can move beyond pilot projects to deliver measurable impact across public systems.
Dean Karlan presented findings from a study conducted in Togo that examined the use of machine learning to improve targeting of public services. According to the research, AI-supported targeting led to measurable improvements in food security, mental health and socio-economic indicators.
However, the study also highlighted limitations, noting that reliance on phone metadata alone failed to accurately capture treatment effects and revealed risks such as model drift and short-term vulnerability misclassification.
Participants said the findings underline the need for experimentation, iterative testing and evidence-based procurement, especially when AI systems are used to identify beneficiaries, allocate public resources or inform policy decisions.
The discussion later turned to India’s preparedness for implementing AI in governance. While progress has been reported in building computing capacity, breaking data silos and training public officials, panellists noted that significant challenges remain.
Issues such as data heterogeneity, scalability and ethical clarity were flagged as key hurdles. It was also observed that only a limited number of AI implementers fully understand the ethical frameworks guiding their systems.
Mohammed Y. Safirulla, Director of the IndiaAI Mission under the Ministry of Electronics and Information Technology, pointed to sectors such as banking and tax analytics where structured data has enabled faster AI adoption. In contrast, sectors like education and other citizen-centric services face complexities due to diverse datasets and ambiguous decision-making parameters.
Speakers emphasised that carefully designed pilots, third-party audits and strong validation mechanisms are essential before scaling AI systems across departments. A recurring theme during the session was the need to focus first on solving clearly defined operational problems rather than attempting sweeping transformations without adequate safeguards.
. ATK SAS PRS