Evidence Action recently announced the incubation of AI for Good, a project systematically mapping how AI could accelerate global development outcomes. Unlike many tech-for-good initiatives, this effort aims to identify applications with potential for transformative scale, rather than incremental improvements, across healthcare, education, financial inclusion, and other critical sectors in low- and middle-income countries.

Some particularly high-leverage possibilities include AI-driven diagnostics that could extend healthcare access to millions who currently lack doctors, predictive tools for disease outbreaks that allow scarce resources to be deployed more efficiently, and adaptive learning platforms that could provide personalized education to tens of millions of children where teacher shortages are chronic. Early models suggest these tools could reduce costs dramatically while multiplying the reach of essential services.

At the same time, there are substantial uncertainties. We do not yet know which interventions are technically feasible, culturally appropriate, or cost-effective at scale. Evidence Action’s approach—systematic mapping, cost-benefit analysis, and pilot testing—offers a promising framework for de-risking experimentation. But the key question is how to prioritize AI applications that are scalable, transformative, and aligned with local incentives, rather than pursuing low-impact pilots.

I'm particularly excited about this project because it seems that it will invest valuable resources into work that has been similar to my own, at a larger (and seemingly more effective) scale. I’m curious what the EA community thinks: if AI could reliably increase access to essential services by an order of magnitude, which applications should receive early attention, and what design principles would maximize positive impact while minimizing risks? How should we think about the most effective ways to shape this emerging opportunity?