In 2026, the most important debate about AI in African healthcare is about ownership, decision-making, and whether the technology actually improves patient outcomes, experts said at the Applied Machine Learning Days Africa conference in Johannesburg on Tuesday.
AI is rapidly entering Africa’s healthcare sector, projected to reach $259 billion by 2030, one of the world’s biggest growth markets. Yet this rapid expansion is happening in systems where the average is roughly 2.6 doctors per 10,000 people across the continent, and only a handful of countries meet the World Health Organisation’s (WHO) recommended doctor-to-population ratio.
An estimated 24 million people were living with diabetes in Africa in 2021, a number expected to more than double to 55 million by 2045. These numbers mean AI will not be a “nice-to-have” experiment, but will likely shape who gets care, how they get it, and at what cost.
Professor Annie Hartley, who leads the Laboratory for Intelligent Global Health and Humanitarian Response Technologies (LiGHT), noted that Africa’s AI-in-healthcare debate must focus on ownership, not dependence on global vendors
“We have to have control,” she told . “We have to have ownership of these tools. We cannot rely on other countries to make these tools for us.”
Africa’s healthcare systems face severe resource constraints, from financing to basic infrastructure. On average, countries in the WHO African region spent about $117 per person on health in 2020, compared to a global average of more than $1,200 per capita, and only five countries in the region reached even $271 per person.
On the infrastructure side, almost all African countries fall below the global average of 2.7 hospital beds per 1,000 people, highlighting how limited physical capacity remains even before new AI tools are added on top.
Hartley said Africa’s so‑called constraint, stringent resources, is a design advantage. That limited computing and funding can incentivise more optimised tools, smarter use of local data, and continuous learning from real-world use in clinics and communities.
“When imported AI systems misread African realities, whether due to language, epidemiology, or workflow differences, local teams are forced to be more vigilant, iterating and adapting instead of treating models as fixed products,” she said.
Hartley believes this positions African researchers and practitioners to lead globally in “continuous learning from low amounts of data” by valuing and carefully curating the data they do have.
Start with citizens’ problems, not AI features
“The greatest conversation is not about which platforms to buy, but about what citizens of Africa want to achieve in terms of health status,” Tom Lawry, an AI health advisor and managing director at Second Century Tech, an advisory firm, said. He urged starting with real challenges, such as a few doctors, overworked nurses, and rising chronic diseases, before mapping an AI roadmap.
Lawry’s approach is deliberately low-tech at first. He said Africa already knows most of its pain points in healthcare.
“What is needed now is for doctors, nurses, and teams to map out processes, define problems clearly, and only then identify where AI can add value, whether by boosting clinician efficiency, cutting administrative hurdles, or improving patient outcome,” he said.
He pointed to a diabetes case study in Singapore, where AI was used to identify people with pre-diabetes and pair that with behaviour-change interventions. The result was a measurable slowing of progression to full diabetes, saving lives while avoiding the higher annual costs once patients become fully diabetic. This kind of population-level use case resonates strongly with African realities, where diabetes prevalence is rising quickly, and budgets are constrained.
Funding, pilots and proof that AI works for Africans
Both perspectives converge on the question of how to move from ideas to working systems in environments where budgets are thin and donor funding is dwindling.
Lawry noted that, practically, money tends to follow clearly defined problems that matter to governments and citizens, which is why pilots grounded in real outcomes are so important. A concrete pilot, say, reducing the progression from pre-diabetes to diabetes in a particular district, gives policymakers evidence of health and economic returns, much like Singapore’s programme, where avoiding full diabetes reduced government spending per patient by thousands of dollars annually.
Hartley said that even with donor-funded pilots, African institutions must control their data, models, and governance. That means investing in local labs, capacity, training on local data, as well as designing AI that works with limited infrastructure.
If AI in African healthcare grows to match the broader global market’s outlook, where AI in health is projected to exceed $500 billion globally by 2033, it will sit at the centre of clinical decision-making, population health, and financing.
The real question is whether those systems will be owned, governed and optimised by Africans, for African problems, or whether the continent will once again be a passive consumer of technologies built elsewhere.
