Walk into a public clinic in rural Siaya County, Kenya, and Africa’s healthcare shortage becomes obvious within minutes: roughly one doctor for every 5,000 people, against a global benchmark closer to one per 1,000. AI diagnostic tools have started closing that gap in real clinical settings, not just research papers, and some of the results are genuinely striking. The harder story is what happens after the pilot ends.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Major Highlights
- At University College Hospital, Ibadan, a speech-to-text system built for African-accented English by Intron Health’s Tobi Olatunji cut radiology reporting turnaround from 48 hours to about 20 minutes.
- In western Kenya, a smartphone-based microscopy tool has cut malaria fever diagnosis from a days-long wait down to about 90 seconds.
- CAD4TB, a computer-aided TB detection tool, is deployed across screening programmes continent-wide; the WHO reaffirmed its backing for the approach in 2025.
- In Egypt, teleradiology platform Rology now returns routine scan reports within 12 hours, and emergency cases within about an hour.
- A 2026 review of AI diagnostic adoption in the DRC found 12–15% gains in radiology accuracy in controlled studies, but concluded scaling is blocked by weak infrastructure, insufficient training, and no clear regulatory framework.
- Africa contributes only about 2.8% of global AI-health research output, concentrated mostly in Egypt and South Africa.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
KINI BIG DEAL
The Ibadan and Kenya numbers are the kind that make for a great headline, and they’re real. But the honest read of this story is that the technology has stopped being the bottleneck. A model trained and validated well tends to perform competitively, that part is basically settled. What isn’t settled is everything downstream of the model: whether the hospital has reliable power, whether there’s a trained nurse who can read a confidence score instead of treating the AI output as gospel or nonsense, and whether there’s a national procurement pathway so every single hospital doesn’t have to negotiate its own pilot from scratch.
Nigeria’s own case is instructive here in an uncomfortable way. The country has some of the continent’s most active builders in health AI, Intron Health being a clear example, but its public hospital infrastructure gap is most visible outside a handful of teaching hospitals in Lagos, Abuja, and Ibadan. A tool that transforms UCH Ibadan’s workflow does not automatically work at a secondary healthcare center three states away without power, internet, or a trained operator. That’s not a knock on the technology. It’s a reminder that “it works in the pilot” and “it works at scale” are two completely different claims, and conflating them is how good tools die in year two of deployment.
The research dependency point deserves attention too. If only 2.8% of global AI-health research comes from Africa, and most of that leans on partnerships with institutions outside the continent for funding and data, then a lot of the scaling decisions for tools meant for African patients are still effectively being made elsewhere. That’s the quieter, less flattering version of this story, and it’s the one that will determine whether these pilots become permanent infrastructure or permanent pilots.
