Can AI Solve Social Problems? What Google's $30 Million Bet Tells Us.

Published: 2024  |  Perspective updated: 2025

AI is everywhere as a talking point. At every conference, in every strategy document, in every conversation about what comes next. The question that interests me more than whether AI is transformative — it clearly is — is whether it will be transformative in ways that benefit everyone, or primarily in ways that make convenient things more convenient for people who were already doing well.

Google's answer, at least in part, is to put money behind the former.

The Google.org Accelerator: Generative AI launched in 2024 with a $30 million global open call for nonprofits, civic entities, academic institutions, and social enterprises to develop generative AI-powered solutions to social problems. Selected participants receive six months of mentorship, technical training, pro bono support from Google employees, and cloud credits. The programme focuses on three areas: knowledge, skills and learning; scientific advancement; and resilient communities.

This is not a small gesture. It is a genuine bet that AI can be meaningfully applied to some of the world's most persistent challenges — not just the commercially convenient ones.

What the first cohort built

The first cohort of 21 organisations built solutions that collectively aim to serve more than 30 million people by 2028. The range is striking.

One participant, Tabiya, used Google Cloud tools to develop an open-source AI-powered conversational agent to tackle youth unemployment — and has since reached more than 8,000 jobseekers in half the time and at a quarter of the cost compared to previous approaches. Another built a generative AI tool to make World Bank development research more accessible to practitioners and policymakers who need it but cannot wade through dense technical documents. A third developed AI-powered tools for student writing feedback, targeting schools in under-resourced communities.

These are not marginal improvements to already-working systems. They are attempts to reach people and solve problems that the market has not prioritised because there is no obvious commercial return.

The access gap

A 2024 Google.org survey found that while four in five nonprofits feel that generative AI is applicable to their work, nearly half are not using it — due to a lack of awareness, training, tools, and funding. That gap between potential and practice is exactly the kind of problem an accelerator programme is designed to close.

The social sector is not inherently less capable of using AI than the commercial sector. It is less resourced to experiment, to fail, to iterate, and to acquire the technical expertise that makes experimentation productive. A programme that provides funding alongside technical mentorship addresses both the financial and the knowledge barrier simultaneously — which is why the format matters as much as the money.

The harder question

I want to be honest about where my optimism has limits here.

Programmes like this are important. They are also, relative to the scale of the problem, modest. The global femtech sector alone — one slice of the health and social impact space — is projected to reach $206 billion by 2033, yet still accounts for just 8.5% of total digital health funding. The social impact applications of AI are similarly dwarfed by the commercial ones. Grant funding and accelerators are necessary but not sufficient.

The deeper question is whether AI development — at the level of foundation models, data access, and infrastructure — is being shaped with social impact in mind from the beginning, or whether it is primarily being built for commercial use cases and social impact is being added at the edges.

Google's accelerator suggests a genuine commitment to the latter. Whether that commitment influences how the underlying technology is built — not just how it is deployed — is the question I keep asking.

Why this matters for financial services

The parallel to financial services is direct. The same gap between potential and practice exists in financial inclusion. The tools to reach underserved customers, to deliver financial education at scale, to identify vulnerability earlier and respond more effectively — many of them exist. The will, resource, and commercial incentive to deploy them at the necessary scale is more variable.

The organisations that close that gap — that genuinely use AI to improve outcomes for the customers who are hardest to serve, not just the easiest — will be the ones that deserve the trust and loyalty of those customers. And increasingly, they will be the ones that Consumer Duty, TISFD, and a growing body of regulation will reward for doing so.

Social impact is not a side project. It is where the next generation of genuinely useful products will come from.