AI and the Future of Institutional Capacity in Africa

Governments and investors in Africa are rapidly expanding AI applications. An important question is what happens when intelligence becomes abundant across the continent while institutional capacity remains limited or is still taking shape.

The launch of the African Union's Continental AI Strategy has focused attention on access to technology and innovation. AI may ultimately reshape development priorities by reducing the cost of producing information and expert recommendations. The central challenge lies in organisational capacity. In other words, as information becomes less costly, institutional capacity becomes increasingly important.

The ability to detect problems is not the same as the ability to solve them. An AI tool that picks up high-risk pregnancies in Kano or Turkana does not automatically mean saving more mothers. If an AI tool flags complications but women cannot reach a hospital with enough staff and the resources to treat them, better detection may increase demand without improving outcomes. The potential gains depend on what happens after the warning: timely transport, reliable referral and effective care.

The Changing Economics of Expertise

For a long time, development policy treated knowledge as a scarce good. You needed more engineers, more doctors and more agronomists. Today, AI can increasingly mimic certain forms of expert judgement at very low cost. AI is making certain forms of specialised knowledge and analytical assistance more readily accessible, even as professional judgement and contextual expertise remain essential.

But to describe AI-generated expertise as "cheap" immediately raises political-economic questions. Cheap on what terms? Access to AI applications does not necessarily equate to control over the technology, infrastructure or economic relationships that surround them.

Why Institutions Matter

Healthcare in Africa provides perhaps the clearest illustration of this broader challenge. AI can improve diagnosis, triage and clinical decision-making. Yet value is only created when improved decisions lead to treatment, recovery and improved wellbeing. These requirements are often treated as a neutral list of capabilities.

They are the outcome of struggles over resources and priorities. Decisions about where hospitals are built, who receives healthcare workers and whether services are publicly or privately funded all shape who benefits from technological innovation.

Penda Health's AI clinical copilot provides an instructive example. A 2025 reprint covering 39,849 patient visits across 15 clinics reported a 16% relative reduction in diagnostic errors and a 13% reduction in treatment errors among clinicians using AI Consult. However, a subsequent randomised controlled trial involving 9,691 patients across 16 facilities, published in 2026, found no statistically significant reduction in treatment failure within 14 days. The findings highlight an important distinction between improving aspects of clinical decision-making and achieving measurable improvements in patient outcomes. Embedding AI within existing clinical workflows may be necessary for translating recommendations into better care, but it does not guarantee improved outcomes.

Jacaranda Health's PROMPTS programme is an AI-enabled maternal health platform integrated into public health systems that provides stage-specific guidance, responds to mothers' questions and facilitates referrals when serious risks are detected. According to Jacaranda Health's third-quarter 2025 report, the organisation had reached more than 3.5 million mothers across its operations, while 82% of high-risk mothers referred by its helpdesk reportedly received care. The programme has shown that successful digital innovation depends on integration into existing care routines and referral systems rather than technology alone.

In 2026, Aga Khan University Hospital Nairobi introduced an AI-enabled radiotherapy system that combines AI-supported planning with highly precise radiation delivery. While these technologies can boost treatment speed and accuracy, patients benefit because hospitals possess the specialist staff, workflows and organisational capacity needed to translate recommendations into care.

This institutional aspect also brings with it questions of governance and control. The policy issue is therefore not just one of whether the AI systems work or not, but rather who governs them, who owns the data upon which they are based, who pays for them and who is ultimately accountable in the event of a failure.

Dependency, Governance and Control

For African health systems, the promise of AI is really about something more fundamental than whether it will improve the diagnosis of disease, assist the clinician, or extend access to care. It's about who controls the infrastructure upon which all of those benefits will depend. Private providers can provide useful capability but there is a real danger in being reliant on proprietary systems. The issue is about establishing arrangements that actually build up local capacity and preserve public authority. One has to look at effectiveness in the same way as you assess patient data protection, clinical accountability and long-term affordability. Governments and health systems need to be able to set their own priorities, to be able to scrutinise performance and to be able to change providers without any disruption to care.

Yet the same logic extends well beyond healthcare.

M-KOPA's success in digital finance is not simply a story about algorithms. It is also a story about distribution, customer relationships, repayment systems and operational discipline.

Apollo Agriculture likewise combines machine learning with networks of local agrodealers and agents to provide farmers with inputs, advice and support. The question is not just whether AI is more efficient, but how its benefits, costs, ownership and risks are distributed.

The Politics of Acting on Knowledge

The greatest gains from AI are unlikely to come from simply adding software to existing systems. They are more likely to depend on how effectively governments and organisations translate new technological capabilities into improved services. This requires investment in management capacity, procurement systems, infrastructure, workforce development and service delivery.

Yet institutional capacity alone does not determine who benefits. Governments allocate resources within political constraints, markets distribute technological benefits unevenly, and organisations often serve the priorities of powerful stakeholders. AI may identify social needs accurately, but political systems may still favour profitable markets, influential regions or internationally backed projects.

The challenge for African governments is therefore not simply to acquire more sophisticated technologies, but to build institutions capable of deploying them effectively and ensuring that their benefits are widely shared. This means strengthening local expertise, maintaining public oversight of technological systems and ensuring that investment in AI supports broader development priorities. If the twentieth century rewarded societies that accumulated expertise, the AI age may increasingly reward those able to govern, coordinate and act upon it.

Bruno S. Sergi is an instructor at Harvard University’s Division of Continuing Education and a professor at the University of Messina. His research focuses on economic development and emerging markets. He is the editor-in-chief of the Cambridge Elements in the Economics of Emerging Markets series at Cambridge University Press. His commentary and analysis have appeared in the Los Angeles Times, Newsweek, South China Morning Post, The Japan Times, The Straits Times, The National Interest, Harvard Business Review Italia, and The Diplomat.