Artificial intelligence promises to revolutionise healthcare, but can technology realistically solve the systemic challenges facing healthcare systems?
On 9 April 2026, Regenesys Education will host the South Africa AI Summit 2026 in Sandton, bringing healthcare professionals, technologists, and policymakers into the same room to confront one of the most seductive claims in modern innovation: that artificial intelligence could transform healthcare. That claim is not baseless. The World Health Organisation identifies real applications for AI in diagnosis, clinical care, administration, education, research, and public health. In practice, ambient AI has already been linked to reduced documentation burden for clinicians at Sutter Health, while NHS England and Leeds Teaching Hospitals are piloting AI-assisted prostate cancer diagnostics to shorten waiting times. Rwanda’s Health Intelligence Centre offers another example, using integrated data and analytics to support early warning, resource allocation, and system performance.
Cutting Through The Hype
That evidence matters because it clears away two lazy positions at once. The first is the cynical argument that AI in healthcare is all hype. It is not. The second is the far more dangerous belief that once the software arrives, the hard work is over. It is not. The OECD’s latest work on scaling AI in health is blunt about what responsible adoption actually requires: a skilled workforce, secure and interoperable infrastructure, governance, monitoring, assessment, and procurement systems that can distinguish a useful tool from an expensive gimmick. In other words, AI succeeds when the system around it is already capable of carrying it.
The Public Healthcare Reality
South Africa does not enter this conversation from a position of comfort. Statistics South Africa’s 2024 General Household Survey found that only 15.5% of people had medical aid cover, while 84.4% did not. The same survey found that 73.1% of households said they would first go to public clinics, hospitals, or other public institutions when someone falls ill or is injured. That single fact should reset the entire debate. The real test for AI in South Africa is not whether it can impress in elite private settings. It is whether it can work where the majority of people actually seek care: in the public system, under pressure, with uneven resources, and in provinces where access is already fragile.
The Workforce Gap AI Cannot Fill
That pressure is not abstract. The National Department of Health’s 2024 report on nursing workforce shortages says the gap across nursing categories in 2019 ranged from 26,000 to 62,000 nurses. The same report links the shortage to rising healthcare needs, retirements, unfilled vacancies, budget constraints, and the strain created by South Africa’s quadruple burden of disease. That means the country is trying to modernise healthcare while still fighting a staffing battle at the bedside. AI can support a nurse with documentation, decision support, and patient tracking. It cannot physically replace the nurse who is missing from the ward, the clinic, or the emergency room.
Infrastructure Is The Real Test
Infrastructure creates another hard ceiling. Stats SA reports that 82.1% of households had some kind of internet access in 2024, which sounds encouraging until you look closer. Only 17.4% had fixed internet at home, and in rural areas, just 2.7% of households had internet at home at all. That gap matters because healthcare AI does not run on optimism. It runs on dependable electricity, functioning devices, usable records, and stable connectivity. A 2026 study on Pretoria primary healthcare facilities found that power outages disrupted diagnostics, treatment, and administration, extended waiting times, and contributed to poor health outcomes. In places where staff cannot reliably access registration systems, blood results, or oxygen-related interventions during outages, talk of advanced AI can sound less like reform and more like fantasy.
South Africa Is Not Starting From Zero
Still, dismissing AI outright would be just as shallow as worshipping it. South Africa is not starting from zero. The Department of Health describes the Health Patient Registration System as a cornerstone digital solution that enables a longitudinal health record, and official digital health materials present the emerging South African digital health system as part of a more transparent and accountable path toward universal health coverage. That is important because AI performs best when it can draw on structured, linked, and interoperable data rather than fragmented paper trails and disconnected platforms. The point is not that South Africa lacks a digital base. The point is that the base is still being built, and AI will only be as strong as the foundations beneath it.
Governance, Ethics, And Regulation
The governance environment is also evolving, but it is not finished. SAHPRA published regulatory requirements for AI and machine-learning-enabled medical devices in August 2025. The HPCSA’s ethical guidelines, published in late 2025, make the position even clearer: practitioners remain accountable for clinical decisions, AI should assist rather than replace professional judgement, patient privacy must be protected, and systems must meet standards for interoperability, data quality, and cybersecurity. Meanwhile, South Africa released a draft National AI Policy Framework in October 2024, and the 2026 Estimates of National Expenditure state that government still plans to develop a national AI policy and implementation plan by March 2028. That is progress, but it also tells a more honest story than the hype does. South Africa is not operating inside a fully settled AI governance regime. It is building one in real time.
The Danger Of AI Theatre
This is where the conversation becomes uncomfortable. AI is most useful when it is unglamorous. It helps when it cuts admin time, flags a risky scan faster, detects a disease trend earlier, or improves how scarce resources are allocated. It becomes dangerous when leaders treat it like a headline solution for problems that are really about staffing, procurement, maintenance, electricity, infrastructure, and access. South Africa does not need AI theatre. It does not need glossy pilots that photograph well and collapse under ordinary service-delivery conditions. The real risk is not simply that AI could fail. The real risk is that it could be used to avoid confronting the failures that matter more.
Where AI Can Actually Help
That is why the most sensible path is also the least dramatic. South Africa should use AI, but it should use it where value can be measured and failure can be contained. Radiology triage, documentation support, patient identification, disease surveillance, referral prioritisation, and planning tools for resource allocation are sensible starting points. Each use case should be locally validated, clinically supervised, and judged against hard outcomes such as reduced waiting times, better continuity of care, fewer missed cases, and time returned to frontline staff. The Department of Health’s recent Health Technology Assessment symposium report described HTA as a cornerstone of an equitable, efficient, and sustainable health system. The OECD makes a similar point: AI adoption in health should include evidence assessment, cost-effectiveness, approval processes, monitoring, and procurement standards rather than blind enthusiasm.
Can AI Really Fix South Africa’s Healthcare System?
So, can AI really fix South Africa’s healthcare system? No. That promise gives technology a job that belongs to governance, public management, budgeting, workforce planning, and infrastructure. But can AI make parts of the system faster, smarter, and more responsive when used properly? Absolutely. The honest position is less fashionable than the hype, but far more useful. AI is not a cure for a broken health system. It is a force multiplier for a system willing to fix what is broken first. That is the debate worth having in Sandton.
