Primary healthcare of the future — putting people into the process through the power of AI

The decision by many donor countries to change their approach to global health and reduce funding for health, especially for HIV programmes in particular, shook many countries, especially in sub-Saharan Africa, which have been the beneficiaries of US government donations over decades. The consequences of the stop-work orders on 20 January 2025 caused much consternation, and data from a number of countries has shown its impact, in particular those related to the HIV response.

Such shocks call on African countries to take greater responsibility for the health of their people and to work with donors to strengthen domestic health systems. In the near term, this means doubling down on a primary healthcare (PHC) approach that builds resilient services close to communities and expands people’s ability to care for themselves safely – “digital self-care” – so that scarce clinical time is reserved for those who need it most.

These sentiments were underscored during a meeting which produced the Lusaka Agenda and more recently in Accra. In brief, these and many other such meetings and conferences, including the Alma-Ata and Astana declarations, have all promoted the importance of strong health systems, in particular primary healthcare.

Given that primary healthcare has been the subject of many meetings and declarations – as well as the consensus that it is the foundation of all health systems – perhaps we should focus first on what it is and what it is not.

What is primary healthcare and why should we care?

Most people equate primary healthcare with primary care, but these are not equivalent. According to the World Health Organization, primary healthcare has three components:

(a) primary care and essential public health functions as a core of integrated health services;
(b) multisectoral policy and action; and
(c) empowered people and communities.

Primary care consists of five components – first contact and accessible care; continuity of care; coordination and integration of services; a full package of services; and being people-centred. In this paper, we make a case for the role of AI and digital health within primary healthcare and organise the role of AI and digital health around the five primary-care components to make explicit where technology adds value – and where it does not. Fundamental to the PHC approach is the understanding that health outcomes are shaped by the places people live, work and recreate.

What is wrong with the current PHC platform and why change?

Given the increasing burden of disease from conditions like diabetes, hypertension, cardiovascular diseases, cancers and mental illness – and the continuing burden of infectious disease in many parts of Africa – coupled with the growing threats from pandemics, there is a need to rethink what an adaptive PHC system should look like.

Why the PHC system? As previously noted, this is the first line of access to health services and where the health system is closest to communities. Any changes in patterns of health conditions and health outcomes are likely to be first seen at the primary healthcare level. It is therefore critical that we strengthen this system not only by building new clinics but also by strengthening community systems.

AI and digital health: key elements of a PHC of the future

Taking into account the wide range of environmental and health-related challenges and opportunities that communities in sub-Saharan Africa in particular are confronted with – and declining donor funding – governments and communities need to work together to redesign their primary healthcare systems.

What could a PHC of the future look like, and how useful can existing and emerging technologies be? To respond more rapidly to social and health challenges, South Africa and other LMICs should consider the use of their current assets and new assets to leapfrog towards a better future. The PHC of the future won’t live only in clinics. It will extend into the places where people already live their lives – retailers, transport hubs, Sassa queues, markets and malls. These are the high-volume channels through which people already move, buy, learn and seek help. Embedding simple, trusted digital tools in those everyday spaces could take prevention and advice to millions long before they step into a waiting room.

AI can support the five primary-care components, listed above, in the following ways:

• First contact and accessibility: responsibly designed tele-triage, telemedicine and multilingual SMS/WhatsApp booking reduce unnecessary visits and prioritise emergencies;
• Continuity: unique patient identifiers and longitudinal records with consented reminders (appointments, refills, viral-load testing) support adherence and re-engagement;
• Coordination and integration: interoperable lab, pharmacy and referral records prevent duplication and maintain care when people are mobile;
• Comprehensive services: embedded decision support for the total package of PHC services, including prevention, curative services, rehabilitation and palliation; and
• People-centeredness: co-design, explainability, offline-first options where required, and accessibility features ensure AI does not widen inequities.

AI and digital health technology has the potential to help Africa leapfrog over its many challenges. What this entails is using innovation and new technologies to make forward-looking decisions without repeating past mistakes. A simple example is to move rapidly to expand access to and use of mobile phone technology without the need to expand the use of fixed-line phones. In 2023 there were 527 million unique mobile phone subscribers in sub-Saharan Africa, and this is estimated to increase to 751 million by 2030.

Just imagine: the device you are reading this article on becomes your trusted go-to for all your health and wellness queries, your personal coach, your reminder for medication and annual check-ups, and even your guide to the right level of care when needed. Imagine this experience delivered not in a distant accent, but in one’s own local language, recognising cultural nuances, and understanding that “context matters” when care is involved. That is the promise of a truly African AI for health – one that listens, speaks and understands the people it serves.

Yet amid this promise, we must be cautious not to be dazzled by the “shiny new tools”. Technology should serve health priorities, not define them. The best AI is invisible – woven into workflows, supporting health workers, empowering communities and respecting privacy. Before scaling any new digital solution, we should ask: does it solve a real problem? Does it work in low-connectivity areas? Does it reduce workload, or merely shift it? Does it speak the language of those it serves?

Adopting a “problem-first, people-first” lens ensures that AI reinforces, rather than replaces, the human relationships that lie at the heart of primary healthcare.

Empowering users of health services – user-facing applications

Digital health and AI tools can transform how individuals and communities interact with PHC services, making care more accessible, personalised and responsive.

Personalised health information with access to products and services: AI-powered chatbots and virtual assistants can provide health education, reminders for medication adherence, and referrals to products (such as HIV self-testing kits) and services (such as HIV treatment) in local languages.

Improved access to services and continuity of care: telemedicine platforms supported by AI triage systems can connect patients in remote areas to providers, prioritise urgent cases and reduce unnecessary facility visits.

Feedback loops and accountability: digital platforms allow service users to report on primary healthcare service quality – including waiting times and stockouts – in real time, which can increase the accountability and responsiveness of facility managers.

Community-based surveillance: AI and mobile phones can be used to report poor social conditions that may result in disease outbreaks, including substandard food or unsafe water.

We frame digital self-care as a strategy to give individuals, families and communities agency to understand their health issues and provide them with reliable information to take care, and secondarily to take pressure off clinics so healthcare workers can focus on people with acute, complex or unstable conditions. Priority packages include HIV self-testing with auto-linkage, refill and side-effect bots, blood-pressure and glucose self-monitoring with escalation rules, and contraception follow-ups. Ensuring pregnant mothers start antenatal care early, those at risk are identified and all mothers have micronutrient supplements. This is aligned to multi-month dispensing and community pick-up models – critical as waiting times lengthen and facility pick-up burdens grow.

Improving provider performance and facility efficiency

AI and digital health solutions are not only patient-facing but can also improve the efficiency, decision-making and job satisfaction of health workers and managers.

Clinical decision support systems: AI-enabled tools can help providers to reduce medical errors, personalise treatment and improve quality of care through risk prediction, early diagnosis and adherence to clinical guidelines.

Integration of components of the health system for improved efficiency: linking laboratory, pharmacy and referral systems with patient records through AI reduces duplication of tests and ensures continuity of care for mobile populations within and across borders.

Workforce management: predictive analytics can allocate healthcare workers across different sections of a facility, between facilities and between catchment areas to optimise deployment, especially to underserved areas.

Data-driven management decision-making: managers at all levels can use AI to analyse routine health data, identify performance bottlenecks and take corrective action.

Avoid the “shiny-tools” trap: adopt problem-first design; human-in-the-loop decision-making; equity and bias checks; privacy-by-design and data minimisation; offline-first usability in local languages; open standards (such as HL7 FHIR) to avoid lock-in; and require pre- or post-evaluation (wait times, file retrieval, ART pickup time, percentage on multimonth dispensing) before scaling. Avoid deploying complex models when simple workflow redesigns or better management can yield greater gains.

Steps to support AI application in PHC in South Africa

South Africa’s Department of Health has a national digital health strategy that can be leveraged to integrate AI applications in PHC. In addition, the department will need to establish governance arrangements and guidelines for the ethical use of AI in health.

To enable full use of digital health and AI applications, the department should improve connectivity in all facilities, with a special focus on rural clinics and hospitals through partnerships with mobile phone providers. As part of its aim to develop an electronic medical record (EMR), the department should ensure the patient record is interoperable with laboratory and pharmacy systems. This requires that all parts of the system use a unique patient identifier – this remains long outstanding.

Equally important is to ensure that provider and user preferences are considered in how AI and digital health applications are designed. Co-designing applications is a critical requirement for successful implementation – which should include change management to ensure that technology sustainably changes behaviour and practice. In addition, preparing users (both clinicians and patients) to use AI-assisted technologies is important. This may include on-site and remote training. To reduce training costs, micro-learning approaches – small chunks of content shared via WhatsApp or SMS – should be considered.

Priorities should explicitly target “time and touch” metrics: reduce median file-retrieval to less than 15 minutes; ensure ART parcel collection less than 30 minutes per guidelines; increase the share of stable PLHIV receiving three- to six-month refills; and restore external pick-up options where feasible.

Governance should include a simple “go/no-go” checklist (problem-first, usability-first, equity-first, evidence-first, cost realism) and a transparent register of algorithms in use. Connectivity and EMR roll-out must be paired with usability testing, clear SOPs for abnormal-result recall, and sustained micro-learning for facility teams via WhatsApp.

Conclusions

While no technology is a panacea, AI and digital health should be positioned as force multipliers for PHC’s core functions – not as replacements for people or human-centeredness. When aimed at today’s operational bottlenecks (staffing, waits, filing, refills) and backed by simple metrics, they can free up healthcare workers to do what only humans can do: build trust, manage complexity and deliver compassionate care. AI and digital health should provide tools that amplify the best of PHC, not replace it. Focused on everyday spaces and local languages, they can make the health system faster, fairer and more human – helping health workers spend less time on queues and forms, and more time on care. DM

Yogan Pillay is the director for HIV and TB delivery at the Bill & Melinda Gates Foundation. Neli Mokhunoane is the senior programme assistant for HIV and TB delivery at the foundation.