The idea of AI in healthcare often conjures images of high-tech robotics solving the world’s health issues in shiny labs and smart hospitals. Its potential to help the country move closer to universal health coverage is enormous, but any serious conversation about AI in healthcare should be grounded in the realities of the country’s health system.
South Africa’s healthcare sector is complex, with a significant disease burden of HIV, TB, heart disease, diabetes and mental illness. Add to this some of the highest trauma rates in the world, infrastructure and staff shortages and crime and corruption, and the reality starts to emerge.
This complexity is further exacerbated by historical inequality, resource shortages, and social drivers of illness such as poverty, malnutrition, and housing issues. Any AI tools introduced into this system need to be built with an understanding of all of these factors if they are going to be successful in addressing them.
As a practical example of this, South Africa’s quadruple disease burden means that significant resources are required to manage chronic diseases and ensure that medicines are both available and that patients are taking them. AI-powered tools can predict a healthcare facility’s medication needs, track stock levels, and flag procurement gaps before stockouts occur. On the patient side, AI models can predict which patients are likely to default, and chatbots can do regular check-ins and reminders.
A key consideration in delivering any AI tool into the South African healthcare sector is the workforce. On the one hand, the country faces severe and worsening health worker shortages and tools that help drive productivity and efficiency or allow people to manage their own health better are necessary. But on the other hand, this is not a sector where people are incidental to the mission of delivering universal health coverage, they are the driving force behind that mission.
There are some important considerations around how the transition to AI technologies is managed, how people will be retrained and whether the tools are being used to extend the capacity of people who are already overextended, rather than hiring additional people.
Covid-19 demonstrated that innovation moves faster than regulation. In much the same way that we had to rework regulations to allow for telemedicine and remote consultation, we will need to properly regulate the use of AI to ensure that safety and efficacy are not compromised.
Artificial intelligence featured prominently at the BHF’s 25th Annual Conference, where one theme emerged consistently across discussions: innovation must be guided by purpose, not for innovation’s sake. Delegates agreed that AI has enormous potential to improve healthcare delivery, but only if it is underpinned by trust, accountability, appropriate governance and solutions tailored to South Africa’s healthcare realities.
We need to be able to trust both the data and insights AI generates, because the cost of getting it wrong in healthcare are simply too great.
Dr Martin Bekker, a researcher in AI ethics at the University of Witwatersrand, summed it up when he said that AI should be used to eliminate “dirty, dull and dangerous” tasks, provided it improves safety and dignity.
As we grapple with regulating this new technology, the guiding principle should be that AI should be a tool that augments, rather than replaces, the knowledge, expertise, and care of trained healthcare workers. Clinical judgement, empathy and accountability cannot be outsourced. Importantly, the physician remains ultimately responsible for patient care, making a human-in-the-loop approach essential.
If all of these are factored in, AI could genuinely expand access to care. South Africa has around 9 doctors per 10,000 people – roughly a third of the WHO’s recommended ratio. Telemedicine platforms are already creating access to healthcare for people cut off from care across the continent. Community health workers, in particular, could be significantly supported by AI tools that help them identify risk and manage follow-up care.
Another area where AI tools are already showing significant benefit is in the area of fraud, waste and abuse, which happens in both the public and private sectors and costs the country billions of rands annually. Fraud, claims manipulation and duplicate billing don’t just impact the bottom line, they ultimately impact how care is delivered. AI-based pattern recognition is able to process data more accurately and significantly faster than humans ever will, and we are already seeing meaningful results.
Patient safety has to be at the centre of every AI tool we deploy. AI systems are only as good as the data they were trained on and models built on datasets that don’t reflect South Africa’s disease burden, genetic diversity, or clinical presentation patterns present real risk. That safety extends to data privacy too. Health records are among the most sensitive data that exists. As more data flows between more systems there is increased risk for attack.
AI will not fix South Africa’s health system on its own. The funding model, the infrastructure gaps, and the shortage of trained professionals require political will and sustained investment. But if AI is introduced with rigour, local knowledge, and genuine accountability, it could make the system meaningfully better for the people who need it most. Ultimately, technology succeeds only when it strengthens patient care and improves outcomes.