India: From digital to AI public infrastructure

India may be preparing for its next big digital leap. And this time, it could be about putting intelligence on top of the digital infrastructure that already exists. On September 21, Rajiv Kumar, President and Managing Director of Microsoft India, said India is on a roadmap to move from Digital Public Infrastructure (DPI) to AI public infrastructure over the next five years. He specifically referred to Aadhaar, UPI and Digi Yatra as examples of the country’s successful DPI. 

The statement raises a simple question: what exactly is AI public infrastructure?
To understand this, think of India’s DPI as a network of digital highways.
Aadhaar provides a digital identity. UPI provides a common rail for digital payments. Digi Locker provides a way to access digital documents. These are not individual services competing with private companies. They are common infrastructure on which thousands of public and private services can be built.

Adding an intelligent layer
AI public infrastructure could add an intelligent layer to these digital highways. Instead of simply enabling a transaction, an AI-enabled system could understand a person’s language, analyse information, personalise a service or help the person navigate a complicated process.
In simple terms, DPI connects people digitally; AI infrastructure could help technology understand and assist them.
This idea is not entirely theoretical.

At the India AI Impact Summit in February, Nandan Nilekani, one of the architects of India’s digital public infrastructure, argued that the important question is how quickly AI can be productively diffused among people. Drawing a parallel with UPI, he said open networks allow many innovators to build applications using AI. He also argued that AI agents could remove complexity for users by allowing them to interact in their own language. 

That is an important distinction. India does not necessarily need to build every AI application itself. Just as UPI created an open architecture on which banks and technology companies could build payment services, an AI-enabled public infrastructure could allow many organisations to build applications for healthcare, agriculture, education and public services.

The Singapore model
Singapore provides an interesting glimpse of how such an approach can work. Its Government Technology Agency says it is building shared AI tools and data infrastructure for Government agencies. These include AI-powered citizen-service chatbots, productivity tools for public officers and analytics platforms. Instead of every agency developing its own AI system from scratch, shared infrastructure can reduce duplication and speed up deployment. 

Singapore’s updated National AI Strategy, released in May this year, also calls for AI to be embedded more deeply across Government to improve public services.  But there is an important lesson here for India: AI infrastructure is not simply about computers and data centres.

The World Bank’s latest research, based on Government responses from 60 economies, shows that AI adoption in Governments is already widespread but often shallow. Forty-four per cent of internal Government AI use consists of individual employees using AI for tasks such as searching, summarising or obtaining assistance.
The World Bank says effective AI adoption also requires reliable data, systems that can share information, trained personnel and appropriate governance. 
India has already begun creating some of these building blocks.

Bharat EduAI stack
In February, the Ministry of Education launched the Bodhan AI Centre of Excellence and announced the development of the Bharat EduAI Stack, an open and interoperable architecture intended to build, integrate and scale AI-powered learning solutions across school education, higher education, skilling and research. 
Education could therefore become one of the early testing grounds for India’s AI public infrastructure. Similar approaches could eventually emerge in agriculture, healthcare, financial services and Government administration.

The real promise is not that every Indian will suddenly have access to a sophisticated AI model. It is that AI could become embedded in services people already use.
A farmer could interact with a Government service in his or her own language. A student could receive personalised learning assistance. A small entrepreneur could navigate regulations without understanding the complexity of Government systems. A citizen could get help accessing a public service without knowing which department handles it.
India’s first digital revolution created common rails for identity, payments and information. The next could put intelligence on those rails.
And if Rajiv Kumar’s prediction is correct, the next five years could determine whether India’s AI story is merely about building powerful technology, or about making that technology useful to millions.