Voice AI May Become the Most Inclusive Digital Interface for India, Says Nilekani
As India looks beyond text-based interfaces, Infosys co-founder Nandan Nilekani argues that voice-driven AI could become the most effective way to deliver digital services at population scale.
News
- Amazon Confirms 16,000 Layoffs After Email Leak Sows Confusion
- Voice AI May Become the Most Inclusive Digital Interface for India, Says Nilekani
- Dell to Power NxtGen AI Factory with Nvidia Blackwell GPUs in India
- Google and ICC Join Hands Ahead of World Cups
- TCS Bets on Brazil With Its Largest Delivery Campus in the Country
- Adani Ties Up With Embraer to Build Regional Aircraft Ecosystem in India
Voice-based artificial intelligence could become the most powerful access layer for India’s digital economy, according to Nandan Nilekani, co-founder of Infosys and chair of the EkStep Foundation.
“If a person can talk to the computer and get back instructions, information, or have an agent work for them, that’s the final frontier of access,” Nilekani said, pointing to India’s linguistic complexity, which includes 22 official languages and hundreds of dialects.
He was speaking at a fireside chat at ‘Voice AI: Making the Best Work for India,’ an event organized by EkStep in partnership with NVIDIA.
EkStep, which Nilekani co-founded with Rohini Nilekani and Shankar Maruwada, builds open digital infrastructure aimed at improving literacy, learning and public service delivery at population scale.
Nilekani said India is well positioned to emerge as a global reference point for large-scale voice AI, with implications for language technology, governance and everyday digital interactions.
Delivering reliable voice AI across Indian languages over the next few years, he said, could create a playbook for other multilingual societies.
He cited practical use cases already taking shape, including voice-based access to railway PNR information and tools that help farmers understand crop pricing and market conditions.
Such applications, he argued, are more likely to attract sustained funding, particularly when they are tailored to specific sectors, languages and dialects.
“These are the kinds of applications that will get funded,” Nilekani said, adding that NVIDIA is providing the compute infrastructure required to train and deploy voice models at scale.
The discussion also featured Vishal Dhupar, Managing Director for South Asia at NVIDIA.
In a lighter moment, Nilekani joked that Dhupar’s role allowed him to simply say, “I want a million GPUs,” and see it happen.
At the same time, Nilekani warned of a growing divide in how AI is being deployed.
He described a “race to the bottom” driven by low-quality content, exploitative applications and tools that prey on loneliness and mental health vulnerabilities. India, he said, should aim to lead a “race to the top” instead.
India’s AI efforts over the past five years have focused largely on text-based translation, he noted. Voice AI introduces a far more complex challenge because of accent variation, dialects and widespread code-switching between languages.
On responsible deployment, Nilekani stressed the need for strong guardrails. “It is about ensuring AI behaves as expected, doesn’t hallucinate, and stays within boundaries,” he said.
AI, he added, should amplify human capability rather than replace it. Designing systems from that premise, he said, is what will ensure the technology delivers broad social value.
