The Would Problem

The Would Problem

Why getting AI used matters more to us than building it 

Welcome! We’re Wadhwani AI US. We build and fund AI for the public systems that already reach the world’s most underserved communities: health clinics, classrooms, and government offices serving the roughly 3 to 4 billion people who live on less than a few dollars a day. We believe technology should reach the people who need it most, whether or not they can pay for it. And AI is how we can do that at scale. It’s been our mission for the last 8 years: leveraging AI to improve health, education, and livelihoods across India and the Global South. 

A framing making the rounds among funders in this space captures the sector's real dilemma well: could, would, should. Research dollars already fund what AI could do. Almost nobody funds what it should do. We built our foundation on would: what it actually takes to get something past a pilot and working at population scale for those same 3 to 4 billion people. 

Wadhwani AI US is a 501(c)(3) foundation and the founder and principal funder of two independently operating programs. The Wadhwani AI India Program is delivered by our partner, the AI unit of LEHS, and is embedded directly inside India's national health, education, and agriculture systems. The Wadhwani AI Global Program, delivered directly by our foundation in partnership with technology organizations in the Global South, is bringing the same approach to new governments across Africa and parts of Asia.

We launched in 2018 with a flagship program in India, developing and deploying AI models directly inside government systems already reaching large populations. We help governments identify where AI can solve a development problem they already have. We design and build the AI itself, trained on that government’s own data. And we train the government’s own staff to run the solutions we’ve built in the long-term. 

Eight years in, here's some of what that's resulted in: 

• More than 25 AI solutions live across health, education, and agriculture, working with 7 central ministries and 15+ state governments

• A team of 350+ people in India building and running that work day to day

• Funding partners that include the Gates Foundation and Google.org

To bring this to life, we’ll highlight some of our solutions. All three are built and run by our India partner, the AI unit of LEHS, and funded through our foundation.

Tuberculosis screening. A health worker holds up a basic smartphone while a patient coughs into it a few times, and within a minute the tool can flag whether that cough pattern is worth a follow-up TB test. No x-ray, no lab, and no specialist required. It - like all of our tools - runs offline and in rural settings. 

Telehealth. Built into India's national telehealth platform, this tool gives frontline health workers real-time, AI-backed guidance while they're treating a patient no specialist is in the room for. It turns what used to be a basic phone helpline into something closer to having a specialist on call.

Childhood literacy. A student reads aloud for one minute into a phone. The tool listens, immediately scores their reading level, and gives the teacher a plan for what that student needs next. That kind of individualized assessment used to take roughly 10 minutes of 1-on-1 time per student, time that most crowded classrooms don’t have. 

None of this was simple technology to build. The TB screening tool runs on a cough sound database that had to be recorded from scratch (and in real clinics). Turning that into a model accurate enough to trust for TB screening, and small enough to run offline on a basic phone, took years of iteration that most AI teams never have the luxury for. While getting the technology right has been necessary - it isn’t sufficient. What keeps tools like these running at national scale, year after year, has been government partners co-owning the build from day 1 and ensuring the solutions are sustainable enough for their own staff to support them without us. 

That is a lot of what we’ll write about here. Every AI lab in the world already covers what AI can do. But at Wadhwani AI we’re just as interested in what it actually takes to get it used, by whom, and helping those left out along the way.