IntroductionIn recent years, artificial intelligence (AI) has transformed medical practice and research across both individual and population levels of care. The integration of AI into healthcare has generated substantial research interest, with numerous studies documenting its applications across clinical settings and evaluating its utility through diverse classification frameworks. These classification frameworks have organized AI implementations by target recipients1—clinical users, direct-to-consumer services, and business operations—or by examining the overall progress, challenges, and emerging opportunities of AI2.In primary care settings, individual studies have validated the clinical utility of specific AI applications. For example, AI-based diagnostic tools for skin conditions have demonstrated measurable improvements in diagnostic agreement compared with reference standards for primary care physicians3. Similarly, AI-based assessment of retinal images for diabetic retinopathy screening has shown favourable sensitivity and specificity when compared with the gold standard of ophthalmologist evaluation4. While these studies have validated individual AI applications in primary care settings, they focus predominantly on technical performance without clarifying how these tools strengthen specific primary health care (PHC) functions. This limitation hinders sustainable implementation and prevents AI solutions from becoming deeply embedded within healthcare system operations.The value of AI in healthcare ultimately depends on its capacity to improve the quality of care and address fundamental challenges facing healthcare delivery systems5. Primary care forms a critical link between healthcare delivery and national health policy, playing a vital role not only in improving population health within individual countries but also in advancing global health equity by addressing disparities in access to essential services6. Despite this central role, the primary care workforce continues to face critical shortages, escalating levels of burnout, and limited access to specialist consultation7. As the foundational tier of health systems, primary care must be assessed from population health and health system perspectives, and AI deployment and assessment should similarly adopt this multidimensional viewpoint.Herein, we propose a framework to systematically detail how AI can enhance primary care performance across multiple dimensions, bridging the gap between individual technological applications and comprehensive primary care improvements within diverse health system contexts. Specifically, we introduce the concept of a “1.5-Tier Healthcare System” as a model for collaboration between human expertise and AI systems. We outline the conceptual pillars of the framework and review recent studies that have implemented and evaluated AI-empowered programs and tools in real-world primary care settings. Further, we explore how this framework can be adapted across diverse healthcare contexts—specifically high-income urban health systems, low-resource rural health systems, and fragile or conflict-affected settings—and explain why South Korea represents an optimal test site for the 1.5-Tier Healthcare System.The 1.5-Tier Healthcare SystemWe use the term “1.5 Tier Healthcare System” to describe a model in which AI serves as a technical bridge between primary care and higher-level hospital services (Fig. 1). Rather than referring to a single tool, this concept offers a broader framework for understanding how AI solutions can reinforce the role of primary care and address the persistent challenges faced at this level of the healthcare system. We organize the system into four conceptual components and discuss how AI can strengthen the core functions of primary care: first contact, comprehensiveness, coordination, and continuity (the 4Cs)8. We also review existing evidence to demonstrate the ways in which AI-enabled tools can enhance the 4Cs within primary care (Fig. 2).Fig. 1: The 1.5-Tier Healthcare System.Full size imageThe framework delineates tiers based on geographic or organizational layers (‘WHERE’) and clinical intensity and scope (‘WHAT’). The 1.5-Tier serves as a virtual integrator, bridging community primary healthcare (PHC) with hospital expertise through digital coordination.Fig. 2: Reinforced primary care functions within the 1.5-Tier Healthcare Model.Full size imageThe current 4Cs of primary care—first contact, comprehensiveness, coordination, and continuity (left)—are each strengthened by AI within the 1.5-Tier Healthcare System (right).First contact: augmented primary access and optimized triageFirst contact represents the initial point at which individuals enter the healthcare system for any new health need, and acts as a critical gatekeeper that determines the efficiency of downstream care8. Strengthening the first contact stage is essential for ensuring timely access to care, preventing unnecessary specialist utilization, and ensuring the remaining core functions of primary care operate effectively9. However, primary care settings often lack the specialized equipment and training required for precise diagnostic evaluation10. These structural limitations can contribute to inconsistent case recognition and suboptimal referral decisions, particularly for conditions in which early detection is essential to prevent progression and where timely intervention substantially alters outcomes.AI-enabled triage systems help mitigate these challenges by supporting more consistent identification and routing at the point of first contact. A prospective study in Australia demonstrated that automated retinal photography combined with AI-based glaucoma screening achieved an AUROC of 0.80, with sensitivity of 65.0% and specificity of 94.6% for referable glaucoma, identifying referable disease in 18 of 161 (11.2%) previously undiagnosed participants10. By expanding screening capacity and flagging individuals at higher risk, this approach facilitates earlier intervention and reduces unnecessary specialist burden, ultimately improving the likelihood of favourable clinical outcomes.A similar benefit has been observed in early-stage mental health care. Under-resourced primary care environments are frequently beset by long wait times and inefficient triage, causing many individuals with mild or emerging mental health symptoms to leave before receiving appropriate support11. AI-based assessment tools can provide timely preliminary evaluations, support risk stratification, and guide patients toward suitable early interventions. By identifying these individuals at the first point of contact, AI reduces delays that typically hinder access to care and enhances the overall efficiency of downstream service utilization. Together, these examples illustrate how AI facilitates automated data intake, real-time risk stratification, and more precise routing at the first point of contact, thereby strengthening the gatekeeping capacity of PHC.Coordination: multi-level and cross-setting care coordinationCoordination refers to the role of primary care as the central hub that manages information exchange and aligns decisions across multiple care settings. Effective coordination prevents duplicate testing, conflicting treatment plans, and missed interventions—issues that disproportionately affect patients with complex chronic conditions12. However, current referral pathways are often mired by inappropriate referrals, incomplete information, and outdated communication systems, contributing to delayed specialty care access and fragmented care13.AI-enabled tools can address these gaps in coordination by improving the accuracy and consistency of inter-professional communication. Large language models can generate standardized referral letters and discharge summaries directly from electronic medical record (EMR) data, achieving up to 99.25% accuracy in zero-shot summarization14 and demonstrating strong agreement with human clinicians15. The automation of medical documentation reduces clinician burden, minimizes communication errors16, and ensures more complete, timely information transfer17.AI also strengthens cross-setting and population-level coordination, an increasingly essential function as health outcomes are shaped not only by medical services but also by community resources and social determinants. Population health management (PHM) focuses on integrating diverse data sources and aligning care delivery across settings within defined communities18. In PHM, AI improves coordination by analysing large, heterogeneous datasets—EHRs, monitoring devices, genomic information, and socioeconomic indicators—to identify vulnerable subgroups and anticipate where disparities in access or outcomes are most likely to emerge19. AI-enabled platforms further enhance coordination between primary care and social or community services. A social risk management initiative in the state of Nebraska (United States of America) used a digital care navigation platform to streamline outreach, assess the unmet social needs of patients, connect them with community resources, and support follow-up through community health workers; this approach serves as a scalable model for integrating social care into population-focused primary care20. By enabling proactive identification and timely intervention across clinical and non-clinical settings, AI reinforces the role of primary care as the central coordinating hub that aligns medical care, social support, and population-level health strategies.Comprehensiveness: data-informed comprehensive carePrimary care is uniquely positioned to deliver lifelong comprehensive care, addressing prevention, acute illness, chronic disease management, and rehabilitation21. While the depth of expertise in individual medical domains has expanded, the ability of primary care physicians to manage a breadth of conditions has become strained22. Enhancing comprehensiveness—the capacity to address the full spectrum of health needs within primary care—is essential for reducing fragmented healthcare.AI-enabled diagnostic tools address comprehensiveness challenges by supporting more consistent recognition and structured interpretation of findings at the point of care, functioning as digital aids that augment clinician judgment. In a randomized controlled trial, AI decision support for spirometry improved primary care clinicians’ preferred-diagnosis prediction by 9.0%p(95% CI, 4.5–13.3; P = 0.001) overall, and by 15.9%p(95% CI, 9.0–22.7; P