On October 7, 2025, at the AI for Africa Conference in Cape Town, something rare happened: major institutions put money behind the question everyone’s asking but few are rigorously answering—does AI actually deliver social good, or are we just building expensive solutions to problems we don’t fully understand?
The AI Evidence Alliance for Social Impact (AEASI) launched with £2.75 million ($3.7 million) to do what Silicon Valley often skips: generate hard evidence on whether AI solutions in Africa and Asia actually help people, or just help pitch decks. It’s part of a broader $7.5 million collaboration with Google.org, bringing together Community Jameel, J-PAL (Abdul Latif Jameel Poverty Action Lab), the UK’s Foreign Commonwealth & Development Office, Canada’s International Development Research Centre, and IDinsight.
Here’s why this matters. AI hype has reached African markets with promises of transforming agriculture, healthcare, education, and financial services. Startups raise millions claiming their AI will diagnose diseases, predict crop yields, or personalize learning. But evidence? Often thin. AEASI directly tackles this gap by funding rigorous research on what works, what doesn’t, and critically—what causes harm.
George Richards from Community Jameel frames it clearly: “We need evidence of which AI solutions work effectively, safely and fairly in order to maximise its impact.” Not whether AI can theoretically work, but whether it does work when deployed in contexts where infrastructure is spotty, digital literacy varies, and stakes are high.
The alliance brings together AI developers, policymakers, civil society, and researchers to evaluate real-world impacts. This means looking at job displacement alongside productivity gains, examining whether AI healthcare diagnostics actually improve outcomes for patients without access to specialists, and tracking whether agricultural AI tools help smallholder farmers or just generate data for foreign companies.
For African founders building AI solutions—whether in healthtech, agritech, or fintech—AEASI could provide validation that attracts serious capital. Conversely, it might expose which AI applications are oversold. Either outcome beats the current state: hype without accountability.
The initiative connects to AI4D, a UK-Canada program launched in 2023 to foster “safe, inclusive and responsible AI ecosystems” across Africa and Asia. With UNESCO, the African Union, and South Africa’s G20 presidency backing the Cape Town conference where AEASI launched, there’s finally institutional weight behind evidence-based AI deployment rather than just innovation theater.
British minister Kanishka Narayan positioned it as backing “African-led innovation that puts people first.” The real test? Whether AEASI’s findings actually influence which AI projects get funded, scaled, or shut down; and whether African researchers lead that evaluation process, not just participate in it.
