Notice: Test mode is enabled. While in test mode no live donations are processed.
Artificial intelligence is advancing through healthcare faster than many clinical, regulatory and educational systems can adapt. AI can already summarize medical records, draft documentation, detect patterns in diagnostic data, support clinical decisions and communicate directly with patients. The next question is no longer simply whether physicians will use AI. It is what responsibilities physicians should retain as AI becomes more capable.
The American Medical Association (AMA) and Digital Medicine Society (DiMe) provided an important answer by releasing a framework defining the physician’s role in the digital and AI era of medicine. The official [AMA framework overview] identifies five responsibilities that the organisations argue should remain fundamental even as individual clinical tasks change.
The development was subsequently highlighted in [TechTarget’s report on the AMA-DiMe AI framework] which emphasizes patient trust, clinical judgement and safe implementation as central themes.
For healthcare systems in Africa and across the Global South, the framework deserves particular attention. AI may help health systems extend scarce clinical capacity, but successful deployment will depend on maintaining accountability, equity, evidence and human relationships while introducing technology at scale.
The framework comes during an unusually fast period of technological change. AI is moving from experimental pilots into documentation systems, diagnostics, patient navigation, administrative workflows and clinical decision support.
DiMe’s own research illustrates the implementation challenge. In its [3 Key Insights for the 2026 Health AI Horizon] the organisation reported findings from 2,041 healthcare leaders across 90 countries and highlighted gaps involving confidence, workflow integration, governance and workforce readiness. The message is significant: healthcare organisations increasingly need not only more capable AI, but stronger systems for using AI safely and effectively.
The new AMA-DiMe framework takes the discussion one level deeper by asking what physicians themselves should continue to contribute when technology begins performing more of medicine’s routine or cognitive work.
The first responsibility is perhaps the least technological and potentially the most important: preserve trust through human connection.
The framework argues that empathy, communication and shared decision-making must remain at the centre of medicine. As patients increasingly encounter AI-generated information before, during and after clinical appointments, physicians have an important role in helping them interpret information, understand uncertainty and make decisions consistent with their circumstances and preferences.
This distinction matters because delivering information is not the same as delivering care. An AI system may produce a medically relevant answer, but a patient can still need reassurance, contextual interpretation, discussion of trade-offs and an understanding of how treatment choices affect their family, work, finances or cultural preferences.
In the Global South, preserving trust may be particularly important when digital health tools are introduced into communities with different languages, health-literacy levels and relationships with formal healthcare institutions. Technology that performs well technically but fails to earn community trust will struggle to create meaningful health outcomes.
The second responsibility is clinical judgement.
AI can identify patterns, generate differential diagnoses and support clinical decisions, but the AMA-DiMe framework maintains that physicians remain responsible for synthesizing evidence, balancing competing risks and making decisions that reflect an individual patient’s goals and circumstances.
This issue has become more urgent because AI performance is improving quickly. A provocative August 2026 [JAMA perspective on autonomous AI and physician-AI care] argues that AI may eventually outperform physicians or physician-controlled AI combinations in some cognitive medical tasks. The authors also acknowledge that much of the available evidence comes from simulations or discrete tasks and that real-world implementation still faces clinical, engineering, regulatory, liability and workflow barriers.
The AMA-DiMe framework therefore addresses an important distinction. The goal is not to claim that physicians must manually perform every task forever. Instead, it defines judgement, synthesis and accountability as capabilities that medical education and practice should deliberately preserve as technology assumes more information-processing work.
That creates a new training challenge. Physicians will need to understand when an AI recommendation is reliable, when it may be inappropriate for an individual patient, how to recognize model failure and when to override automated recommendations.
The third responsibility is for physicians to help redesign medicine itself.
AI creates opportunities to redistribute work among doctors, nurses, community health workers, patients and technology. The framework argues that physicians should help determine which responsibilities require physician leadership, which can safely be supported by other healthcare professionals and which can be performed or assisted by technology.
This is especially relevant for health systems experiencing severe workforce constraints.
The WHO Regional Office for Africa’s State of the Health Workforce in Africa 2026 estimates that the region had approximately 5.72 million health workers in 2024 and continues to face a projected needs-based shortage of around 5.85 million health workers by 2030.
AI cannot manufacture millions of doctors, nurses or midwives. What it can potentially do is help existing health workers use their time more effectively.
For example, ambient documentation could reduce administrative workload; decision-support systems could help frontline health workers identify high-risk patients; AI-enabled imaging could support screening where specialist interpretation is scarce; and conversational systems could help patients access basic health information between clinical encounters.
The fourth responsibility may be the most important for organisations actively deploying clinical AI: physicians must steward responsible technology use.
AMA and DiMe call for AI and digital health technologies to be integrated safely, effectively, equitably and on the basis of appropriate evidence. Physicians should participate in deciding where technology adds value, where human oversight remains essential and how systems should fit into clinical workflows. They should also contribute to continuous evaluation and identify unintended consequences after deployment.
That focus on continuous oversight is important because AI safety does not end when a system passes an initial validation test.
DiMe describes this problem in its [The Missing Middle of Healthcare AI] The organisation argues that governance often becomes weakest after procurement and deployment, precisely when health systems need to monitor real-world performance, patient impact, equity, workflow fit and unintended consequences.
Healthcare organisations therefore need mechanisms for:
This is also why bias requires particular attention in Africa and other underrepresented regions. AI developed mainly using data from North American or European populations cannot automatically be assumed to perform equally well across African populations, languages, disease patterns, clinical workflows or equipment environments.
The fifth responsibility is preparing medicine for what comes next.
AMA and DiMe argue that physicians have obligations beyond individual patient encounters. The profession must establish the competencies, education and standards required for medicine in an AI-enabled environment while preserving sufficient independent expertise for physicians to exercise judgement and oversight.
Medical education may therefore need to evolve significantly.
Future doctors will require traditional clinical knowledge alongside AI literacy: understanding model limitations, evaluating evidence, detecting automation bias, protecting patient data and communicating AI-assisted recommendations transparently.
Continuing professional education will be equally important. AI systems may change far faster than conventional medical curricula, meaning practising clinicians will need mechanisms for continuous learning.
For lower-resource countries, this should also include locally relevant capacity building. Governments, medical schools and professional councils should avoid becoming entirely dependent on imported AI expertise. Training clinicians, researchers and engineers who understand both local health systems and AI will be critical for sustainable adoption.
The framework does not suggest physicians can build this future alone.
AMA and DiMe also published a shared roadmap covering four stakeholder groups: patients and patient groups, clinical professionals, technology developers and investors, and policymakers and payers. It describes short-, medium- and long-term stages for aligning incentives, implementing new models and ultimately making technology-enabled care sustainable.
Patients should help define the outcomes, priorities and experiences technology-enabled healthcare should deliver. Clinical professionals should lead the redesign of care and define appropriate competencies. Technology developers and investors should build trustworthy products based on clinical evidence and patient needs. Governments and payers must create regulatory, financial and infrastructure environments that encourage safe, evidence-based adoption.
That shared responsibility is especially relevant in the Global South. A rural clinic cannot overcome poor connectivity, weak data infrastructure, inadequate financing or inappropriate imported technology simply by asking individual physicians to “use AI responsibly.”
Responsible AI requires a functioning ecosystem.
The AMA-DiMe AI framework for physicians is not a detailed implementation standard, and the AMA itself describes it as a common foundation for deeper work rather than a complete prescription for deployment.
That means health systems should translate its principles into operational practice.
Governments and regulators can establish risk-based clinical AI standards, require locally relevant validation, strengthen health-data governance and support regulatory capacity before adoption accelerates further.
Hospitals and health systems can create multidisciplinary AI governance groups involving clinicians, patients, data specialists, cybersecurity teams, procurement staff and health-equity experts.
NGOs and donors can finance not only innovative pilots but also evaluation, training, cybersecurity, community consultation and long-term monitoring.
Technology developers should disclose intended use, limitations, validation populations and update procedures while making systems auditable by healthcare organisations.
Universities and research funders should support African-led AI evaluation, multilingual datasets and studies conducted in real clinical environments rather than assuming that evidence generated elsewhere will automatically transfer.
The most important insight in the AMA and DiMe framework may be that medicine does not need to preserve every existing physician task in order to preserve the value of physicians.
Technology will change what doctors do.
Some administrative tasks may largely disappear. Some diagnostic and analytical activities could increasingly be performed by AI. New care models may give nurses, community health workers and patients more powerful digital tools. The balance between human and machine decision-making may continue to evolve as evidence accumulates.
But the enduring goal of healthcare remains unchanged: patients need safe, effective, compassionate and accessible care.
For Sub-Saharan Africa and the wider Global South, that creates a major opportunity. AI could help scarce health workers reach more people, improve early detection, reduce administrative workload and bring clinical knowledge closer to communities that remain far from specialists.
Yet scale without governance could deepen existing inequalities just as easily as it reduces them.
The future should therefore not be framed simply as physicians versus AI. A more useful question is how physicians, nurses, community health workers, patients, governments and technology developers can redesign healthcare around the capabilities that each contributes best.
The AMA DiMe AI framework for physicians provides an increasingly valuable foundation for answering that question: preserve human trust, strengthen clinical judgement, redesign care thoughtfully, govern technology responsibly and prepare the health workforce for continuous change.
For organisations working to expand healthcare access in Africa, those principles offer more than guidance for doctors. They offer a blueprint for ensuring that rapidly advancing AI serves the needs of patients first.
The AMA and DiMe framework arrives at a pivotal moment for healthcare. As artificial intelligence becomes more deeply integrated into clinical workflows, the central challenge is no longer whether AI will influence medicine, but how health systems can ensure that it does so safely, equitably and in ways that strengthen patient care.
By emphasizing trust, clinical judgement, responsible technology stewardship, the evolution of medical practice and professional development, the framework offers a practical foundation for preserving the human values of medicine while embracing technological progress. These principles are especially important for countries in the Global South, where AI has the potential to extend scarce clinical capacity, support frontline health workers and improve access to specialist knowledge.
However, technology alone will not solve structural healthcare challenges. Governments, healthcare providers, researchers, funders and technology companies must work together to ensure that AI systems are properly validated, locally relevant, transparent and continuously monitored. Investment in infrastructure, workforce training and inclusive research will be just as important as investment in the AI tools themselves.
The most promising future is therefore not one in which AI replaces physicians, but one in which physicians and other health professionals use AI to deliver better, faster and more accessible care. If implemented responsibly, the AMA-DiMe framework could help guide healthcare systems toward a future where technological innovation and human expertise reinforce one another—ultimately improving outcomes for patients regardless of where they live.
Ayati Devices raises INR 15 crore to scale AI-powered diabetic foot diagnostics, supporting earlier detection and more a
QuantHealth raises $45M to expand AI-powered clinical trial simulations, supporting faster drug development and more inc
DiMe is developing practical AI governance tools to help healthcare organisations adopt, monitor and manage artificial i
