AI Could Redraw Specialty Boundaries in Care Delivery

Artificial intelligence could reorganise care delivery by expanding clinical scope across broader disease domains, according to a commentary published in Health Affairs Scholar. Rather than focusing only on efficiency gains, the model centres on wider access to specialist-level knowledge and a different way of structuring the clinical workforce. It challenges a system built around narrow specialties and proposes a new category of clinician able to manage patients with several chronic or complex conditions within broader domains of care. The approach aims to reduce handoffs, improve coordination and use specialist capacity more selectively. It also raises questions about how clinicians train, how care is organised and how payment systems respond when boundaries between specialties become less rigid. Progress depends on changes in education, credentialing, oversight and the practical use of AI in daily care.

From Narrow Specialties to Broader Domains
Healthcare systems are still organised around distinct specialties, each with its own training pathway, certification structure and scope of practice. These arrangements developed because medical knowledge grew too large for any one clinician to master across multiple fields and remain current. Over time, those boundaries became embedded in training programmes, specialty societies, billing rules and standards of care. For patients with complex multimorbidity, the result is often a fragmented experience shaped by multiple referrals and limited coordination across disciplines.

Must Read: What Makes Clinical AI Trustworthy

Artificial intelligence challenges the idea that narrow specialty boundaries remain necessary in their current form. Current systems perform strongly in knowledge-retrieval tasks, pass board exams across several specialties and generate recommendations aligned with specialist guidelines. Their weaker area remains context-dependent, longitudinal reasoning, including the ability to pull together fragmented histories, identify subtle findings and manage uncertainty over time. Even so, the pace of improvement is rapid. As access to specialist-level knowledge becomes more widely distributed, the case for organising care strictly around organ-based specialties becomes less fixed.

The proposed response is to organise care around broader disease-based domains rather than around separate specialties tied to one organ system. These domains include cardiometabolic disease, infectious and inflammatory disease, mental and neurological health, oncology, interventional care, primary care and shared services. In that structure, one AI-supported clinician manages the full range of a patient’s conditions within a domain instead of relying on several separate referrals. A patient with diabetes, hypertension and early chronic kidney disease could therefore be managed within one cardiometabolic framework, while a patient with Crohn’s disease and inflammatory arthritis could remain within one unified practice.

Clinical Roles and Care Continuity
The model works most naturally in cognitive specialties, where clinical value rests mainly on diagnosis, prescribing and long-term management. In these areas, AI extends the ability of clinicians to work across broader disease groupings and manage moderate-to-high complexity patients over time. Procedural fields fit differently. In specialties such as dermatology, gastroenterology and ophthalmology, where cognitive work and procedures sit together, AI is more likely to alter the balance of practice than to replace the specialty itself. Generalist-specialists may take on more evaluation and management work, while proceduralists concentrate more heavily on interventions. Surgical and highly interventional specialties remain less affected by this restructuring.

The central organisational change is the introduction of a middle layer between primary care and subspecialist practice. Generalist-specialists would not replace all subspecialty expertise. Instead, they would manage broader domains longitudinally while subspecialists remain focused on procedurally intensive care, rare diagnoses and the highest-acuity cases. That arrangement reduces referral volume without removing escalation pathways for the most complex patients.

For patients, the intended effect is fewer handoffs, faster diagnosis, less administrative friction and more coherent management across related conditions. If more care moves into broader disease domains, specialist capacity could also be used more selectively for patients who need it most. That creates a different balance between generalist and specialist work, with continuity and coordination becoming more central than the maintenance of rigid specialty boundaries.

Implementation, Economics and System Change
The model also affects how health systems and payment structures operate. When fewer clinicians are involved in the same episode of care, value-based payment and bundled payment models become easier to apply. A major difficulty in specialty care comes from the number of clinicians involved in a single patient pathway. Consolidating care within broader disease domains reduces some of that complexity. At the same time, the model carries economic risk under fee-for-service payment if it expands clinical scope without constraints. Existing specialty boundaries create friction that limits how much care one clinician delivers directly. If AI reduces that friction, previously deferred, fragmented or incomplete care could become additional billable activity, increasing overall service volume rather than reducing it.

The payment approaches identified for piloting include per-member-per-month capitation tied to defined disease domains, global ambulatory codes that bundle cognitive evaluation and management across related conditions and shared-savings arrangements linked to lower total cost of care. These approaches align financial incentives more closely with broader clinical scope. The aim is to ensure that organisational change in care delivery is matched by payment models able to support it.

Health systems and Academic Medical Centers would also need to adapt. If community and generalist clinicians can manage more complex conditions without referral, the volume of routine specialty care at Academic Medical Centers could decline. One response is to focus more tightly on ultra-complex cases that exceed AI capability and require research-grade intervention or complex procedures. Another is to reorganise into more seamless hubs that bring together diagnostics, procedures, laboratory services and pharmacy support. Across the workforce, some specialists may move further into procedural or ultra-complex care while others adopt broader domain-based roles. That transition creates a need for cross-specialty domain certificates, AI-era maintenance of certification pathways, competency-based privileging and clear escalation protocols. Patient safety remains central, particularly where AI may generate hallucinations, miss contextual signals or create overconfidence at the edge of traditional training.

AI could support a substantial redesign of care delivery by extending specialist-level knowledge across the workforce and reducing the need for rigid specialty boundaries. The generalist-specialist model centres on broader disease domains, continuity of care and more selective use of subspecialist expertise. Its practical development depends on changes in training, privileging, payment and oversight, alongside safeguards for patient safety and escalation. The overall direction is a shift away from organising care around narrow specialty lines and towards a structure shaped more directly by disease biology and patient need.

Source: Health Affairs Scholar

Image Credit: iStock