AI for the next billion

If you only follow AI headlines out of Silicon Valley, it’s easy to believe the future is being written in a handful of American labs. But if you want to see what ‘AI for all” will actually look like, you have to watch what’s happening in small-town India, in government schools, primary health centres and village internet kiosks where connectivity is patchy, and incomes are modest, but ambition is very high. 

India already has something most countries don’t: real digital scale at the bottom of the pyramid. The latest Internet in India report estimates around 886 million active internet users, with rural India accounting for roughly 488 million, about 53% of the total. For most of them, the internet isn’t a laptop on a desk; it’s a shared, affordable Android smartphone. If AI can work for this India, low-income, mobile-first, multilingual, it can work anywhere. The real question is whether we design for that reality upfront, or keep treating inclusion like a retrofit. 

India’s edge: rails first, AI next 

India’s biggest AI advantage is not a single model or lab. It’s the decision to build public digital rails first, then let innovation ride on top. 

Aadhaar, UPI and DigiLocker have already shown what happens when identity, payments, and documents are turned into open, low-cost infrastructure instead of proprietary products. The effect is visible in everyday behaviour. A recent digital payments report shows UPI processed over 93 billion transactions in just the second half of 2024, with the average ticket size down to around ₹1,400, evidence that the system is now used for micro-payments as much as big-ticket spends. The insight is simple: when the rails are free, reliable and interoperable, low-income users don’t just adopt digital payments; they normalise them. 

The IndiaAI Mission is designed to extend that logic into AI. Public investment into computers, datasets and startups is not just about chasing the latest benchmark. It’s about ensuring that the models shaping Indian lives are trained on Indian realities. Initiatives like Bhashini, BharatGPT, and BharatGen push that further by treating language as infrastructure.  

They are not building “language packs”; they are building systems that can think and respond across scripts, dialects and code-mixed speech. In a country where the default user may type in Hinglish and speak in a local dialect, that is not a technical flourish it’s table stakes. 

The deeper insight is that inclusion has to be baked into the stack: open rails, affordable access, and AI that natively understands the ways Indians actually live and speak. That combination is precisely what other emerging markets are looking for. 

Under the radar: AI built under constraints

The most instructive AI stories in India don’t look futuristic at first glance. They look like small, pragmatic fixes. That is, until you look at the outcomes. 

Take agriculture. In Telangana’s Khammam district, the “‘Saagu Baagu” ’ project under the World Economic Forum’s AI4AI initiative has worked with around 7,000 chilli farmers. Through chatbots on their phones, farmers receive crop-specific advice, AI-based quality testing, and connections to digital marketplaces. Early results show profits rising by about 18% and, in some cases, farmers effectively doubling their income. The tech is not flashy; the interface is simple and vernacular. But the design logic is powerful: built for low literacy, low bandwidth, and real risk, and AI becomes a profit lever, not a pitch deck. 

In public services, Bhashini and partner institutions are building speech and translation pipelines that can handle everything from widely spoken Indian languages to historically ignored tribal tongues. That may sound abstract, but the impact is concrete. A telemedicine platform that can listen to a patient in her mother tongue and respond in kind changes how she experiences the state. A grievance portal that can be used by voice, not just text, changes who feels capable of filing a complaint. When AI is embedded in those touchpoints, technology is no longer the story; access is. 

Even in civic spaces, you see the same pattern. An AI-powered multilingual kiosk at a bus stand in Kanyakumari that helps tourists with information, SOS support and missing-persons complaints is not what most people imagine when they think of cutting-edge AI. But it reveals the core Indian insight: meaningful AI is built under constraints, in messy, real-world environments, not just pristine labs. 

Writing the playbook for inclusive AI 

Other emerging markets are watching India because their constraints rhyme with ours: low-cost Android phones, intermittent data, multiple languages, and populations that are both vulnerable and entrepreneurial. 

India has already proved, through UPI and India Stack, that you can build digital rails that are open, cheap, and massively scalable, and then export that architecture. The same can happen with AI, but only if we treat inclusion as a core strategy, not a CSR sidebar. 

For India’s AI ecosystem, that translates into a clear set of expectations. Flagship models and applications must be natively multilingual and context-aware, not English systems crudely translated. They must be optimised for mid-range phones and low bandwidth, because that’s where the next billion users live. And success has to be measured in real adoption by women, small farmers, informal workers and first-generation internet users, not just in the number of pilots launched. 

India will get some bets wrong. But if we can build AI that actually works in crowded government classrooms, heat-stressed fields, Tier -3III hospitals and small kirana shops, we won’t just be catching up with an AI race defined elsewhere. We will be defining what winning looks like for the following billion users. 

For a country that has already turned digital payments and identity into global public goods, this is the natural next step. In the end, “AI for all” will not be judged by how many foundation models we train, but by how many lives at the margins we help steady, one local language query, one small decision, one affordable smartphone at a time. 

 (The author is Founder, NxtQuantum Shift Technologies; and CEO, Ai+ Smartphone)