A language is never just words. It carries how a people farm, heal, argue, and pass things down the generations. So teaching an AI to understand Bhili is not only about opening an app, though it does that, letting a farmer finally ask a real question in the language they think in. It is also an act of preservation: putting a spoken language on the record for the first time, in a form that can outlast the people who carried it in memory. Listen to Santosh Kevlani, Head of Voice AI at EkStep Foundation, on why bringing a language into AI opens a door and keeps something from being lost at the same time. Watch the full story of how Bhili speakers are building their own voice data: 100pathways.com/diffusion-in… @Shankar4EkStep @jagadishbabu @shalinikap @TanviLall26 @EkStep_Org

The first call came on 8 January. Within three weeks, Nandan Nilekani was calling what Amul and the EkStep team had built a miracle, in the presence of the Prime Minister. The two teams had met every day at 3 pm to get there. Dr. Jayen Mehta traces that speed back to groundwork laid long before. In 1995 Amul decided to think and work like an IT company in the food business, and today 36 lakh farmers pouring milk twice a day add up to about billion recorded transactions a year. Sarlaben puts that knowledge on a feature phone or a WhatsApp chat for a dairy farmer who never gets a day off, and it turns a milk record at the cooperative bank into a credit history, so a collateral-free Kisan Credit Card loan of up to 2 lakh rupees comes within reach. Then the rest of the world noticed. At the AI Summit, the President of France asked how the same system could work in France, and the leaders of Greece, Mauritius and Bhutan said much the same. India already produces more milk than any other country, and Dr. Mehta sees a chance for it to become a dairy to the world. Listen to Dr. @Jayen_Mehta, Managing Director, Amul (GCMMF), on why a system built for 36 lakh small farmers is the one other countries now want to borrow. See how the Amul pathway came together, step by step, on the Diffusion Cube: cube.100pathways.com/explore @NandanNilekani @narendramodi @Shankar4EkStep @jagadishbabu @shalinikap @TanviLall26 @EkStep_Org @Amul_Coop

Most leaders measure success by what they build; this District Collector, by what can run without her. For an administrator, she says, the happiest thing is to become redundant. She thinks of the role as a midwife's: you help, but the mother gives birth, not you. So you put people in the water, promise to come if they go under, and take the leap of faith. On the Bhili project, that meant handing it to the community that speaks it, to carry forward on their own. Listen to @MittaliSethiIAS, IAS, District Collector of Nandurbar, on why the real work of leading is to make yourself unnecessary. The full conversation: lnkd.in/gSQJySBr @Shankar4EkStep @jagadishbabu @shalinikap @TanviLall26 @EkStep_Org

When an outside team digitises a language, it becomes data about a community. When the community does it themselves, it stays theirs. The Bhili speakers were not the subject of this work, they were the ones deciding, for the first time, how their own language should be set down and what each word should mean to a machine. That is why the result is not just more accurate, it is accountable to the people it belongs to. Listen to Santosh Kevlani, Head of Voice and Language AI at EkStep Foundation, on why a community has to own its language, not simply donate it. If your language was skipped too, you can help build its pathway. cube.100pathways.com/explore @Shankar4EkStep @jagadishbabu @shalinikap @TanviLall26 @EkStep_Org

Picture this conversation happening only over WhatsApp text. It would be slower, stiffer, and harder for both of us. That friction is the whole point. Most people would far rather talk than type, especially in their own language, because talking to another person is the most natural thing we do. So Gaurav Gupta's argument is simple: do not ask people to fill forms, let them talk. When you get that right, you hear it in what they say back. They tell you that you have made their life so simple, the way they would talk about making a packet of Maggi. That, he says, is the proof of the pudding. Listen to Gaurav Gupta, Chief Growth Officer, EkStep, on why the simplest way to reach people is to let them talk, in their own language. See where this is already working, and what it takes to carry it to the next place: cube.100pathways.com/explore @Shankar4EkStep @jagadishbabu @shalinikap @TanviLall26 @EkStep_Org

More than 80% of AI pilots stall before they ever reach people. The reason is rarely the technology. No single organisation can carry an AI solution from an idea to millions of people on its own. It takes the institution that already holds the relationship and the data, the architects who shape the problem into a solution, the specialists who bring in capabilities like voice and language, and the wider ecosystem that keeps it safe and running. David Menezes calls this a team sport. When those players come together around a real human purpose, rather than the newest model, a pilot finally becomes something that reaches people at scale. The full conversation: library.100pathways.com @Shankar4EkStep @jagadishbabu @shalinikap @TanviLall26 @EkStep_Org

A man of forty-two tells the Collector he has never once seen a film. None has ever been made in his language. This is what the work is really about: not whether the agriculture scheme has reached you, but whether you can enjoy the next Shah Rukh Khan film in a tongue of your own. Dr. Sethi keeps the subtitles on even in a language she speaks, wanting every word. She wants the same for the ten million who speak Bhili, and for every language still waiting: a richer inner life, in words of their own. Listen to Dr. Mittali Sethi, IAS, District Collector of Nandurbar, on why this was never about schemes, but about dignity and joy. The full conversation: library.100pathways.com @MittaliSethiIAS @Shankar4EkStep @jagadishbabu @shalinikap @TanviLall26 @EkStep_Org

Somewhere close by there is a job going, and someone who needs it exactly that badly, and the two almost never find each other. That is the quiet failure at the heart of India's last-mile labour market, and it is a discovery problem, not a talent one. Resumes and job posters were never built to close it. What can is a system a worker reaches by voice, in their own dialect, on the phone they already own, run less like another private app and more like a public good that the local community and the government stand behind. Get that right and the economics follow: reaching a worker that once cost about 500 rupees can now cost closer to 10. That is how a hidden labour market finally switches on. Listen to Tushar Bansal of BlueDots, on what it takes to connect a worker and a job that are already standing next to each other. See how this hyperlocal jobs pathway was built, and how to reuse it: library.100pathways.com @Shankar4EkStep @jagadishbabu @shalinikap @TanviLall26 @EkStep_Org

Ask an AI to understand a language, and the first thing it needs is that language written down somewhere. Bhili never was. Ten million people speak it, and the digital world had skipped past all of them, which is why an app could reach a village in Nandurbar and still not understand a word its farmers said. The fix did not come from a lab. It came from the Bhili speakers themselves, who became the authors of their own dataset and built it word by word and story by story, from a language that had only ever been spoken. What they made does two things at once: it lets AI finally answer a Bhili speaker in their own tongue, and it lays a path that the hundreds of other languages the internet forgot can follow. Listen to Mr. Santosh Kevlani, Head of Voice and Language AI at EkStep Foundation, on what it takes to bring a skipped language into AI, and why the people who speak it have to lead. The Bhili pathway is written down for the next language to follow: library.100pathways.com @Shankar4EkStep @jagadishbabu @shalinikap @TanviLall26 @EkStep_Org

A District Collector keeps the subtitles on even for a film in a language she speaks, because she wants every word. Now think of a language that never had one. Bhili is the country's most spoken tribal language, yet for its ten million speakers the internet offered scheme delivery, never their own films or songs. When Nandurbar brought it into AI, the first question was not about word-error rate, but about why: to make people's inner life richer in their own tongue, with the community and not for it. Listen to Dr. Mittali Sethi, IAS, District Collector of Nandurbar, on why bringing a language into AI is a question of dignity, not just technology. The full conversation, and the pathway for the next language: library.100pathways.com @MittaliSethiIAS @Shankar4EkStep @jagadishbabu @shalinikap @TanviLall26 @EkStep_Org

The point of all this is not to deploy more AI. It is to give people more agency. Not an algorithm that replaces the expert, but a farmer who becomes a better farmer, a teacher who becomes a better teacher, a health worker who makes better calls. Productivity that starts with one person, then a team, and then the hardest part of all, a whole society. Listen to Mr. Nandan Nilekani, in conversation with Mr. Dario Amodei at the India AI Summit, on what AI is actually for. See what building AI for everyday users looks like in practice: 100pathways.com @NandanNilekani @DarioAmodei @Shankar4EkStep @jagadishbabu @shalinikap @TanviLall26 @EkStep_Org

Every wave of technology has created enormous value, and none of them has shared it equally. The people who could benefit most are usually the last to see it. AI will follow the same path unless adoption is made deliberate. That is the whole case for diffusion: not to celebrate a few clever deployments, but to make the know-how of deploying well travel fast, so the benefit reaches people sooner. Listen to Mr. Nandan Nilekani at Raisina, on why getting AI to everyone is a choice, not an inevitable outcome. More on why diffusion has to be made deliberate: 100pathways.com/insights @NandanNilekani @raisinadialogue @Shankar4EkStep @jagadishbabu @shalinikap @TanviLall26 @EkStep_Org

The first build is always the slow one. MahaVISTAAR took about nine months. Ethiopia, learning from it, built something comparable in about three months. Amul's Sarlaben, drawing on both, launched in about three weeks. Nothing says every deployment will move this fast, but it shows what happens when hard-won deployment knowledge is written down and handed on: the next adopter starts where the last one finished. Listen to Mr. Nandan Nilekani at the India AI Summit, on how shared pathways turn months into weeks. The pathways behind this compression are documented and ready for the next adopter to reuse: library.100pathways.com @NandanNilekani @DarioAmodei @Shankar4EkStep @jagadishbabu @shalinikap @TanviLall26 @EkStep_Org

Silicon Valley is racing to build the most powerful model. India is asking a different question: how do you put AI into the hands of a billion people, in their own languages, on the devices they already own. The edge is not a bigger model. It is public digital infrastructure, the habit of building at population scale, and a stubborn focus on the common person. Listen to Mr. Nandan Nilekani at Raisina, on why India's real advantage in AI is adoption, not just invention. Read more on India's case for building AI at population scale: 100pathways.com/insights @NandanNilekani @Shankar4EkStep @jagadishbabu @shalinikap @TanviLall26 @EkStep_Org @raisinadialogue

AI labs release a more capable model almost every week. That is not the hard part anymore. The hard part is adoption: getting AI into the institutions and services that actually touch people's lives. When capability races ahead and adoption lags, the people who stand to gain the most end up waiting the longest. Closing that gap is the real work, and it is what AI diffusion sets out to do. Listen to Mr. Nandan Nilekani, in conversation with Mr. Dario Amodei at the India AI Summit, on why adoption, not capability, is the bottleneck that matters now. Find out more about why adoption is the bottleneck that matters: 100pathways.com/insights @NandanNilekani @DarioAmodei @Shankar4EkStep @jagadishbabu @shalinikap @TanviLall26 @EkStep_Org

A tribal language spoken by 10 million Indians couldn't be understood by AI. When a farming app in Nandurbar worked only in Marathi, the Bhili-speaking community built the data themselves. Recording words, translating textbooks, documenting oral stories. That's what AI diffusion really means: getting powerful AI out of the labs and to the people who need it most. @Shankar4EkStep @shalinikap @jagadishbabu @TanviLall26 @EkStep_Org

An idea can travel from a conversation with the Prime Minister to a working app in millions of hands in 34 days. In India, it just did. On January 8, the idea was raised with PM Modi. By February 11, it was live. The Sarlaben app now serves the world's largest dairy cooperative: 3.6 million farmers, 40 million cattle, 2 billion milk transactions a year. It answers questions about cattle health in the farmer's own language, and the data stays sovereign. Listen to Mr. Nandan Nilekani on why this is what AI diffusion looks like when it is built for the common person, and why India is placed to show the world how fast it can happen. @NandanNilekani @narendramodi @mygovindia @Shankar4EkStep @jagadishbabu @shalinikap @EkStep_Org @TanviLall26 @Amul_Coop

When a citizen talks to a government bot, who are they really speaking to? And what's at stake? Getting these answers right is not a technical detail, it is a matter of trust. Santosh Kevlani, Head of Voice and Language AI at EkStep Foundation, on why responsible voice AI has to start with the problem, not the technology, and why it takes rigorous human testing before it ever reaches a citizen. @Shankar4EkStep @shalinikap @jagadishbabu @TanviLall26 @EkStep_Org