The Gates Foundation and OpenAI recently announced  a pilot initiative to advance artificial intelligence capabilities for health in Africa. Together they are committing $50 million dollars in funding and technical support with the goal of supporting 1,000 primary healthcare clinics in Africa by 2028, starting with Rwanda. In a time of dwindling development aid and its inevitable impact on healthcare in developing countries, this is exciting news! Except for one inconvenient truth: you can’t run AI on hope and goodwill only.

Don’t get me wrong, I am not a skeptic of technology. Having spent four years of working with digital health tools in Uganda, I recognise the genuine promise AI holds for healthcare. AI can reduce inefficiencies, improve patient flow and experience, expand access to healthcare in remote areas, and enhance diagnostic capabilities, among many other possibilities. However, I cannot shake the feeling that we are getting the sequence spectacularly wrong. In our rush to deploy cutting-edge AI, we risk building the penthouse while the foundation continues to crumble.

“The sand”

Approximately 1 billion people access healthcare from facilities without reliable electricity access or with no electricity at all. This energy gap is acute in rural areas with over 50,000 healthcare facilities in rural Africa lacking electricity supply, though many urban facilities also face inconsistent unreliable power supply. With inconsistent electricity and limited renewable energy alternatives, how do we deploy sophisticated AI systems in facilities that cannot refrigerate vaccines properly, cannot keep lights on for nighttime procedures, and cannot guarantee staff will have power to even load the AI interface?

Let’s be wildly optimistic and assume we solve the electricity crisis tomorrow, for example if efforts by the World Bank and others pay off. We would still confront the reality of inadequate healthcare staffing, deteriorating facilities, poor internet penetration, and a workforce with minimal digital literacy. A 2022 survey of 47 African countries found that the region has a ratio of 1.55 health workers  (including nurses, midwives and physicians) per 1000 people, far lower than the WHO recommended 4.45 health workers per 1000 people to achieve universal health coverage. While AI can improve efficiency and enable health workers to see more patients, it cannot replace them. Maybe OpenAI and the Gates Foundation see the future differently, but healthcare at its core depends on human connection, contextual judgment, and the application of tacit knowledge, qualities that remain fundamentally beyond AI’s reach.

Internet access is fundamental to leveraging AI’s healthcare potential, yet the digital divide remains stark: internet usage averages just 23% in low-income countries compared to 94% in high-income countries. In Africa specifically, nearly 400 million people in Eastern Africa and 268 million in Western Africa remained offline as of October 2025. Limited connectivity breeds limited digital literacy. In 2011, in rural Ghana and Tanzania for example, only 40% of health workers had ever used computers, just 29% had received any computer training, and roughly 80% were computer illiterate or beginners. A more recent 2017 study still found that a substantial amount of health workers in sub-Saharan Africa were not sufficiently familiar with computer tools, lacking important computer skills. Zoom out further: 12 of the world’s 20 countries with the weakest digital skills are in Africa, where only 11% of university graduates have formal digital training. The gap between AI’s requirements and Africa’s reality is staggering!

The need for sequencing

Am I advocating for abandoning AI and digital health tools altogether? Absolutely not. I’m arguing for prudent sequencing, for not skipping the foundational steps that ensure sustainability and enable AI to deliver on its promise. Before we invest millions in AI deployment, we must ask critical questions: Will these tools function when the power goes out? When the internet is down? When the only available health worker has never used a computer? If the answer is no, then we’re not solving Africa’s healthcare crisis. We’re creating dependence on systems that will fail precisely  when they’re needed most.

AI itself isn’t the problem. The problem is magical thinking. It’s the belief that we can somehow bypass the hard, unglamorous work of building functional health systems and jump straight to the sexy, Silicon Valley solution. That technology can paper over the cracks in infrastructure, staffing, and resources.

Real healthcare transformation requires doing the boring work first. Training and employing more health workers. Building and renovating health facilities. Ensuring consistent electricity and internet. Establishing dependable supply chains. Then we can intelligently integrate technology that actually works in African contexts. This means digital tools must be built for the realities they serve, not imported wholesale from high-resource settings where power outages are rare, internet is ubiquitous, and every health worker has grown up with computers. Technology designed for San Francisco won’t work in Kampala without adaptation, or without the infrastructure San Francisco takes for granted.

Yes, AI belongs in Africa’s healthcare future, but as the cherry on top. You cannot build a digital health revolution without power, without skills, without infrastructure.

And right now, we’re trying to do exactly that.