Africa's Healthcare Crisis Deepens; AI and Automation Offer Path to Oversight

Fewer than one radiologist serves every 100,000 people in several African countries, a ratio that leaves hospitals buried in imaging backlogs and patients waiting on diagnoses where time can determine survival. The World Health Organization puts the continent’s share of the global disease burden at roughly one-quarter, yet Africa holds only about 3% of the global health workforce. That gap is the problem Nexus Intelligence, a Pretoria-based health technology company, is trying to close.

The company’s most striking test came nearly four kilometers underground. At South Africa’s Mponeng Gold Mine, where broadband connectivity is unreliable and working conditions are demanding, a Nexus AI system analyzed a miner’s chest X-ray for tuberculosis and occupational lung diseases, including silicosis, in 45 seconds. No cloud connection. No on-site radiologist.

Nexus co-founders Dr Gerhard Ferreira and clinical engineer Andries Vorster built the platform by licensing a foundational chest X-ray model from Google Research and developing Nexus AI CXR, a regulated clinical decision-support system that screens X-rays for tuberculosis and significant lung abnormalities. The software does not replace radiologists. It prioritizes their work, rapidly clearing likely normal studies and flagging cases that need specialist attention. “That environment stress-tested the complete operating model and showed that a regulated AI medical device can function at the edge, close to the worker, in a setting with constrained connectivity, complex logistics and a high need for occupational lung surveillance,” Ferreira told FORBES AFRICA.

The platform has since been deployed at approximately 40 healthcare sites across six countries, including South Africa, Vietnam and Côte d’Ivoire, processing more than 25,000 chest X-rays. Partners include ministries of health, Global Fund-supported implementers, mines, occupational-health programs, hospitals and radiology networks.

Regulatory approval was not quick. Nexus moved from its Google Research licensing agreement in 2023 to obtaining CE MDR Class IIb certification under the European Union’s Medical Device Regulation in late 2025, a process that took several years. Offline deployment does not mean uncontrolled deployment. Each installation runs version-controlled software, maintains local audit trails, records inference results and timestamps, and supports secure updates during scheduled maintenance windows. Clinical quality assurance, post-market surveillance and human oversight remain part of routine operation.

The question of accountability inside the clinical workflow is deliberate. Joseph Pategou, a biopharmaceutical professional, put it plainly: “The shortage of doctors and radiologists across Africa is precisely why AI has the potential to transform healthcare, but adoption will only happen if clinicians trust the technology and understand the boundaries of responsibility.” In his framing, AI organizes medical knowledge, identifies patterns and prioritizes cases, but the physician retains responsibility for diagnosis, treatment decisions and patient outcomes.

Lorna Omondi, Strategic Partnerships Lead at Google Research Africa, reinforced that boundary. “The clinical decision remains with the clinician. Patients never interact directly with the model. We wanted a partner committed to understanding that safety in this field was of the utmost importance,” she said. Under the partnership’s terms, patient data stays under the control of Nexus and its healthcare partners and is not returned to Google. The arrangement also includes a commitment to provide more than 10,000 in-kind screenings annually for underserved communities.

Much of the global conversation around AI assumes dependable internet. Healthcare, often, cannot. Nexus built its platform to operate either through cloud deployment or entirely offline using dedicated on-premises workstations. In offline mode, the AI travels with the X-ray equipment, analyzes images locally and delivers results during the screening encounter itself. Images, results and audit logs are stored on the device and synchronize later when connectivity becomes available. For occupational health programs, that design reduces unnecessary travel and minimizes production disruptions. What appears to be a technical feature in a hospital becomes an operational necessity underground.

Femi Oke, a Nigerian AI product manager specializing in healthcare interoperability, argued that the continent’s deeper challenge is less about algorithms and more about trusted data systems. “Trust must be designed into the workflow, not added after the model is trained. In healthcare, open standards such as HL7 FHIR can provide a common exchange layer, while consent, provenance and data-quality controls ensure that local data is treated as governed infrastructure, not free raw material,” he told FORBES AFRICA.

By contrast, a parallel conversation is emerging in agriculture. Nearly 60% of sub-Saharan Africa’s workforce depends on farming, according to the World Bank, yet most cultivate small plots exposed to climate shocks, unpredictable rainfall, pests and poor market access. Dr Samuel Babatunde, Director of Operations at Nigerian agricultural production company SBZ Development, cited evidence that AI can increase crop production by 15% to 30%, lower input costs by 10% to 25% and raise net farm income by 20% to 40% when combined with weather prediction, soil analytics and digital market intelligence. “AI’s biggest contribution is reducing uncertainty by predicting pest outbreaks, spotting nutrient deficiencies early, forecasting climate risks, and linking farmers to better markets to bolster resilience and food security,” he told FORBES AFRICA.

Nexus, meanwhile, is expanding its own scope. Beta versions of the platform already include models for silicosis and pneumoconiosis, and the longer-term strategy is to build an integrated lung-health ecosystem combining AI screening with clinical management software and centralized radiology reporting. “More than 90 healthcare workers use it every day, and every one of those interactions ends with a clinician making an informed decision for patients,” Omondi said. “That is what helpful AI looks like in practice: local builders solving local and global problems.”

Whether regulators in other jurisdictions will move quickly enough to match that ambition remains the open question.