BENGALURU: The race to build more powerful artificial intelligence (AI) systems may be accelerating at breakneck speed, but businesses are struggling to keep up, Nandan Nilekani, Non-executive chairman of Infosys, said.
Speaking at the Infosys Investors AI Day in Bengaluru, he said the technology’s rapid progress has created a “deployment gap”. Model performance is rising quickly, but companies are not implementing AI at the same pace.
“Technology is far ahead of its deployment,” he said. “The tech will keep getting better and better… but enterprise deployment is not going to go up.”
In simple terms, the technology is improving faster than businesses can realistically use it. The challenge, he suggested, lies not in the models themselves, but in organisations.
A major obstacle is the weight of legacy systems. According to Nilekani, many large enterprises are spending between 60-80% of their IT budgets simply maintaining old systems. “There is no business value out of that,” he said. Companies want to reverse this ratio, investing more in new systems than in maintenance, but cannot do so without major clean-ups.
He described how, over decades, firms have layered new technologies on top of old ones rather than replacing them.
Further, security risks compound the issue. Many systems were designed “in an era before you could have online attacks”. At the same time, state and non-state actors are becoming “better at using AI”, increasing the threat to organisations that have not modernised.
AI is often linked to higher productivity, but Nilekani warned that this can be misleading. Much of the progress people talk about comes from “Greenfield” projects, which means building something new from scratch. In reality, most companies run on older “brownfield systems” that carry technical debt, data locked in silos and poor documentation.
There is also a risk that today’s excitement about AI will create new problems. “Five years from now, there’ll be more AI legacy systems than any other legacy system,” he said. That means companies may soon have to clean up badly managed AI systems just as they are now struggling with old software.
On jobs, Nilekani said, the real issue is not opportunity but execution. “It is not an opportunity risk, it’s an execution risk,” he said. What matters is whether companies can retrain their people, redesign how they work and manage large-scale change properly.
Additionally, he said investment in AI models is also rising fast. But the gap between what the technology can do and what businesses can actually use may define this era, he said. The hard work lies in cleaning up old systems, retraining staff and changing how organisations operate.