Episode Summary

In this episode of Interpreting India, Nidhi Singh is joined by Shalini Kapoor, chief strategist for Data and AI at the EkStep Foundation, and Tanvi Lall, director for strategy at People+ai. They unpack why so many AI initiatives get stuck after impressive demos, and what it takes to move from pilots to real, sustained adoption. Drawing on research spanning 1,000+ use cases across 25 countries, the guests introduce the Use Case Adoption Framework (UCAF) and explain how India can translate AI ambition into population-scale impact—especially across public services, agriculture, health, and other high-priority sectors.

Why do AI pilots stall in “pilot purgatory,” even when the technology works? What does a concrete AI use case look like beyond a chatbot demo? And what institutional changes—trust, accountability, workflow redesign, safeguards, and data readiness—are required for adoption at scale?

Episode Notes

AI progress is often measured by the number of pilots launched, but this episode argues the real unit of progress should be how many AI use cases are reliably in production and embedded into everyday systems. Shalini Kapoor distinguishes AI innovation (models, chips, and breakthroughs) from AI adoption, emphasizing that adoption is frequently harder because it demands institutional integration, behavior change, and clear accountability—especially when AI advice affects livelihoods, health outcomes, or legal decisions.

Tanvi Lall explains “pilot purgatory” as the frustrating middle state where use cases never move beyond controlled deployments. She shares how recurring barriers—compute constraints within real institutions (not just cloud credits), fragmented workflows, late-stage safety design, lack of sustained funding, and weak organizational readiness—prevent diffusion. The conversation highlights the UCAF approach to defining a use case as a commitment to improve a specific outcome for a specific persona in a specific context, and why trust and accountability are as central as the technology layer.

The episode also explores “horizontal enablers” that make scale possible—data readiness, multilingual language support, voice interfaces for last-mile access, workforce integration, and guardrails. A detailed example (Mahavistar in Maharashtra) illustrates what scaling can look like when government partnership, data pipelines, voice infrastructure, safeguards, and long-term funding align. Finally, the guests look ahead to what AI adoption in India could look like over the next five years, arguing that the most impactful AI will feel “ordinary”—quietly embedded into routine decisions—supported by shared adoption infrastructure rather than one-off pilots.