Nandan Nilekani calls for open AI framework for next-gen models

Infosys cofounder Nandan Nilekani

Infosys cofounder Nandan Nilekani has encouraged business leaders to move beyond the buzz around current AI tools and focus instead on creating resilient systems designed to stand the test of time.

“How do you design an architecture that allows you to bring in new AI capabilities, new models, new agents, new whatever, and make them plug into your whole architecture?” Nilekani asked during a fireside chat with Robbert van Rutten, Shell’s Chief Information and Digital Technology Officer, at Shell’s Changemakers of Tomorrow 2025 event in Bengaluru on Tuesday.

In simple terms, Nilekani is saying not to get distracted by new AI models being released every day. Instead, focus on building strong underlying systems so future technologies can plug in smoothly.

For Nilekani, the risk is not in falling behind on the latest model, but in getting locked into it. “It’s not about specific products because the products will come and go… many companies are spending billions building technology and the leads will keep changing,” he said. “The real challenge is making sure we don’t lock into that, and that we can glide on top of the changing technology.”

The Infosys co-founder drew on his experience building India’s Aadhaar digital identity system and the Unified Payments Interface (UPI), which now processes more than 130 billion transactions annually. In both cases, he highlighted, the key was designing for scalability and flexibility rather than betting on one technology.

As companies like Shell wrestle with applying AI across sprawling supply chains and global operations, Nilekani’s message was clear: today’s cutting-edge models will quickly become yesterday’s news, but a robust architecture can keep pace with whatever comes next.

Also, read: Nandan Nilekani bats for data-driven small language models; bullish on Sarvam AI

Data is the key

Nilekani reiterated that the real driver of AI isn’t the model but the data underneath it. AI systems are only as good as the information they are trained on, and most of the large enterprises remain weighed down by decades of fragmented, siloed, and inconsistent data.

Unless organisations fix their information systems and standardise data flows, the “magic” promised by AI tools will remain out of reach.