Nandan Nilekani flags implementation gap, says AI models outpacing real-world use

Infosys cofounder and chairman Nandan Nilekani

• AI models advance rapidly but enterprise adoption lags behind
• Execution, not opportunity, is the main risk for businesses
• Legacy systems and lack of reskilling hinder AI deployment

Did our AI summary help?

Artificial intelligence models are getting more powerful by the day but businesses are struggling to put them to work, creating a growing gap between AI capabilities and real-world deployment, Infosys cofounder and chairman Nandan Nilekani has said.

The problem is no longer about access to better models but about whether enterprises are ready to deploy them at scale, Nilekani said at the Infosys’ AI Investor Day on February 17.

“The technology is far ahead of its deployment. Model performance is going up, but the progress in implementing is not really there because implementing this is hard stuff,” Nilekani said.

While AI presents a major opportunity, the bigger risk for companies is execution. “It’s not an opportunity risk, it’s an execution risk,” he said.

The comments come at a time when investors are pressing large IT firms for clarity on how generative AI will translate into sustainable revenue growth, margins, and competitive differentiation, even as IT stocks in India and world-over have seen a sharp selloff in recent sessions over AI-disruption fears.

The comments come at time as India hosts week-long AI summit in New Delhi, where tech titans, policy makers and other stakeholders gather to discuss and gauge the emerging tech.

Nilekani said massive investments and competition are pushing AI models forward rapidly but enterprise adoption is lagging because organisations are not built to absorb such fast change.

Also read: Writing code will no longer be the goal as AI reshapes tech jobs, says Infosys' Nandan Nilekani

Bottlenecks

According to Nilekani, the key limitations are organisational and structural.

“Fundamentally, it’s about organisational change, business change, retraining your people, changing your data so that it’s no longer in silos,” he said.

Most companies are still running on legacy systems and operating models designed for a predictable, deterministic software world, while AI systems are inherently non-deterministic.

Decades of accumulated technical debt are slowing down AI adoption, he said. Large enterprises continue to spend the bulk of their IT budgets maintaining old systems, leaving little room for large-scale AI deployment.

“Modernisation of legacy systems cannot be deferred anymore,” he said, adding without cleaning up core systems and data, AI benefits will remain limited.

Also read: Infosys Investor AI Day: There is no opportunity gap due to AI, says Nandan Nilekani

Enterprises that can close the deployment gap through change management, talent reskilling, and system modernisation will be better placed to turn AI’s growing power into business outcomes, he added.