India should Claude-ify its non-profits

We are in an era of nonprofit unicorns—organisations that impact a million people or more. They have mastered policy, co-built government programmes, and scaled through communities to reach the last mile, yet many still shy away from technology.

When talking to nonprofit founders, very often a point of discussion comes up: “A simple app can solve this!” Until recently, that meant engineers, product managers, maintenance budgets, and long timelines, a cost most non-profits couldn’t justify.

All this has changed over the last year. Agentic AI tools such as Claude Code and others, can code apps, websites and software for everyone.

Indian nonprofits are especially well placed to use this: Aadhaar and UPI, handling 27 billion authentications and 220 billion transactions, already offer a digital backbone the social sector can plug into. But AI will not automatically uplift India’s social sector; the real risk is a wave of new technology that never gets used. Three disciplines separate what works from what dies in the graveyard: the right use case, the right product design, and the right go-to-market strategy.

Find the right use cases

Technology mostly amplifies existing behaviour; it rarely creates new behaviour on its own. The demand for easy digital payments existed; UPI removed the friction, and adoption followed. But street hawkers will not automatically adopt a financial planning app, nor will rural classrooms take to a tech-enabled curriculum just because it exists.

Use-case discovery must come first. A scholarship applicant wants funding; they just don’t know which scholarship fits, what documents are needed, or when the deadline falls. Remove that friction and adoption follows almost on its own.

Technology can also augment human capacity, making existing work easier or better. Anganwadi workers often lack graded content to share with children. Rocket Learning fills this gap with bite-sized, localised content and classroom tools for social-emotional learning; frontline workers adopt it because it makes them more effective at work they already do, a force multiplier, not a burden.

A third path is government mandate: the Unified District Information System for Education (UDISE+) requires every recognised school to record school, teacher, and student data in real time each year, bringing roughly 1.5 million schools, 10 million teachers, and 250 million students onto one system. Once a government workflow goes digital, adoption becomes system-led, not consumer-led; non-profits need to build with these workflows, not around them.

Get the product design right

The second discipline is product design, where most social-sector technology fails. A Bharat product must be built for the person who will use it: an Anganwadi worker in a noisy centre, a student on a low-end phone, a court staffer buried in overloaded workflows, or a parent uncomfortable reading long instructions.

That means three things. First, the interface must be multilingual and often voice-first, without adding cognitive load: fewer screens, fewer fields, clear buttons, large fonts, visible next steps. AI makes translation, speech, image recognition, and conversational interfaces cheaper, but bad UX can still kill a good intervention.

Second, it must fit existing habits and workflows and, where needed, work under assisted usage, since the ultimate beneficiary is often not the direct operator. ARMMAN’s Kilkari delivers two-way preventive care communication to pregnant women and mothers, from the second trimester until the child turns one, through recorded voice calls and WhatsApp. It asks nothing of mothers beyond picking up a call, no app to download, no new habit to learn.

Third, it must track metrics that show whether it is actually delivering value. Most beneficiaries don’t pay for the product, so there’s no price signal telling you if it works; a finance app for micro-entrepreneurs must look past downloads to cohort retention and transactions actually recorded.

Distribution is the bottleneck

The third discipline is distribution: building an app is now easier; distributing it, and getting people to use it repeatedly, remains hard and expensive.

One route is government. Adalat AI automates transcription, digitises records, streamlines court workflows, and delivers real-time updates. In two years, it has scaled to more than 3,500 courtrooms across nine states and, with ACT’s support, aims to reach 7,500 by 2027. The beneficiary is the litigant, but adoption runs through judges, court staff, and the judicial system: the poor don’t always need to adopt technology directly; sometimes the institution serving them must.

The second route is B2B2C distribution. Scholarlify, a nonprofit, makes running scholarships simple, letting institutions, organisations, and governments manage scholarships from application to disbursement through automation and data insights. For CSR teams and foundations, it becomes the operating rail to distribute scholarships better; for students, it collapses the maze of eligibility, documentation, and follow-up. Adoption here is institution-led; the beneficiaries are youth trying to fund their education.

This is the real shift. AI has reduced the cost of building. The scarce capabilities now are choosing the right use case, designing for Bharat and solving distribution. India’s social sector needs a new cadre of product managers, AI builders, implementation leaders and GTM teams who understand impact. Funders should not only fund programmes. They should fund product discovery, pilots, maintenance and distribution.If AI diffusion stays confined to corporates, coders, and urban consumers, India will miss a historic opportunity; it must enter the social sector, where the need is greatest. Silicon Valley is already chasing the one-person unicorn company. That may be out of reach for nonprofits, but AI has changed their economics of scale too. The next nonprofit unicorn will win on empathy, product discipline, and distribution, not engineering headcount. India should Claude-ify its nonprofits, not to chase technology for its own sake, but to multiply impact.

(The authors are co-founders, Change Engine, an accelerator backing founders to build nonprofit unicorns. Views are personal.)