Nandan Nilekani Reveals How One Question from PM Modi Built an AI App for 3.6 Million Dairy Farmers in 34 Days
In the most talked-about speech of the AI Impact Summit’s CEOs Forum, Infosys co-founder and Aadhaar architect Nandan Nilekani laid bare two stark truths: that India has mastered the art of taking technology to a billion people at extraordinary speed — and that AI’s entire promise now hinges on whether its benefits reach ordinary citizens before a white-collar backlash derails it. He illustrated both truths with one arresting story.
A Question That Launched an App
The scene Nilekani described was deceptively ordinary: a meeting with Prime Minister Narendra Modi on January 8, 2026, about applying AI to Indian farmers. Mid-conversation, Modi asked a question that cut through the technical debate like a scalpel. ‘Why can’t we apply it to cows and cattle?’ the Prime Minister reportedly said. ‘Because if the cow is sick, it cannot tell you that it is sick. How can you solve this problem?’ That single question, Nilekani recounted to a packed hall of global CEOs and ministers, set in motion one of the fastest AI deployments in Indian history.
The same evening, the Prime Minister’s Office convened a meeting with S. Krishnan and Abhishek Singh of the Ministry of Electronics and Information Technology (MeitY), representatives of Amul — the world’s largest dairy cooperative — and members of Nilekani’s team. The agenda was not to study the idea or form a committee. It was to build. Thirty-four days later, on February 11, 2026, the ‘Sarlaben’ AI application went live across Gujarat.
“An idea that the PM had on January 8 became a reality on February 11. This, to me, is an example of the speed of execution of AI diffusion in India.”
— Nandan Nilekani, CEOs Forum, AI Impact Summit 2026
Sarlaben: What the App Does and Who It Serves
Named after a familiar, trusted Gujarati female figure, Sarlaben is an AI-powered digital assistant built on two billion data records drawn from Amul’s Automatic Milk Collection System, a Pashudhan livestock database, ISRO satellite imagery of fodder production, and an annual farmer-wise cattle census conducted across Gujarat. The app speaks in Gujarati, processes voice queries, and responds in natural language — designed for the average dairy farmer, many of whom are women, who may have limited digital literacy but acute knowledge of their animals.
A farmer can ask about cattle health symptoms and receive a diagnosis. She can check vaccination schedules, get advice on breeding and milk-production practices, or find out which government subsidy she is eligible for — all through a voice interface in her own language and dialect. Gujarat Chief Minister Bhupendra Patel, who launched the platform on February 11, described it as a tool that gives dairy farmers ‘real-time insights on animal health and productivity.’ The scale it serves is striking: 3.6 million milk producers across more than 18,500 villages.
SARLABEN AT A GLANCE Cooperative: Amul — world’s largest dairy cooperative Farmers served: 3.6 million across 18,500+ villages in Gujarat Cattle monitored: 40 million Annual milk transactions: 2 billion Data records powering the AI: 200 crore (2 billion) Time from PM’s idea to live deployment: 34 days (Jan 8 → Feb 11, 2026) Data sovereignty: All data stored within India; Amul retains full ownership Data sources: Amul AMCS, Pashudhan database, ISRO satellite imagery, annual cattle census
The UPI Parallel: India’s Proven Playbook for Scale
Nilekani drew a pointed parallel that resonated visibly in the room. On December 30, 2016, Prime Minister Modi launched the BHIM payment app on the Unified Payments Interface. That launch seeded what is today the world’s most-used digital payments platform — 21 billion transactions a month, 500 million users, and an architecture that has been studied and replicated by governments from Singapore to Brazil. ‘I get the same sense of excitement now that I got then,’ Nilekani told the forum. ‘AI is at that same inflection point. If we get the diffusion right, the transformation will be of the same order.’
The parallel is instructive beyond sentiment. UPI’s success was not simply about the technology; it required institutional coordination, trust-building across banks and regulators, interoperability mandates, and a government willing to absorb initial friction. Nilekani’s broader argument at the Summit’s fireside chat — held alongside Anthropic CEO Dario Amodei — was that AI diffusion is precisely the same challenge. ‘Technology is just one piece of the puzzle,’ he said. ‘It is about institutions, trust-building, negotiations, guardrails, working with different stakeholders towards a common vision. Diffusion is difficult.’
The Warning Beneath the Optimism
Nilekani’s optimism about India’s AI trajectory carried a sharp edge of warning that the applause in the hall could not muffle. AI, he told the CEOs Forum, is currently running two simultaneous races — ‘a race to the top and a race to the bottom’ — and the latter is moving faster. Deepfakes spreading electoral disinformation, rising energy bills as data centres consume grid power, and the systematic displacement of routine cognitive work are all accelerating. If those remain AI’s defining visible impacts, he cautioned, a societal backlash becomes inevitable.
The historical analogy he deployed was stark. The resentment of blue-collar factory workers towards globalisation — automation and offshoring that hollowed out manufacturing towns from Detroit to Dhanbad — eventually produced a political backlash that reshaped elections and trade policy worldwide. Nilekani’s warning was that AI risks triggering an equivalent reckoning, this time among white-collar workers who find their skills automated away. ‘All of us who have a stake in AI being useful to humanity have to accelerate and redouble our efforts to make the diffusion happen,’ he said. ‘Otherwise, the consequences are going to be very difficult.’
Data Sovereignty as Design Principle
Nilekani was careful to underscore a dimension of Sarlaben that sets it apart from many AI deployments in the developing world: the architecture is built around data sovereignty from the ground up. Amul’s data does not leave India. It does not pass through foreign cloud infrastructure. It is not used to train models whose intellectual property accrues to overseas corporations. This is not an incidental feature; it is a design choice that aligns with the MANAV framework’s ‘Whose data, their rights’ principle, announced at the same summit by Prime Minister Modi. In a week where India also joined the Pax Silica semiconductor coalition to secure its supply chain against external dependencies, Sarlaben represents the same impulse applied at the application layer: build for sovereignty, not just capability.
What This Means for Global AI Diffusion
Nandan Nilekani’s speech was, at its heart, a case study in what AI governance should look like when the subject is not hypothetical risk but immediate human need. In 34 days, a conversation about a sick cow became a functioning application serving 3.6 million farmers, built on two billion real data records, operating in the language people actually speak. The speed is remarkable; the method is replicable.
For the AI Impact Summit’s 88 signatory nations, Sarlaben offers a template that no conference declaration can match: a live, working demonstration that AI’s value proposition for the Global South is not philosophical but operational. As Nilekani closed his address, he expressed confidence that by the time many of the attendees return to New Delhi in the months ahead, there will be ‘many, many more applications’ of comparable scale. Whether that confidence proves justified will be, as he acknowledged, determined not by the quality of the ideas but by the quality of the diffusion.
-Kuppuswamy S
