The average primary care visit in the United States lasts just 18.4 minutes — barely enough time for a physician to address the complexities of modern chronic disease, let alone build the kind of trusting relationship that effective medicine demands. As physician burnout accelerates and patient panels grow, the question is no longer whether technology will transform primary care, but how thoughtfully we choose to integrate it.
The crisis in primary care
Primary care in the United States is in crisis. The average primary care visit lasts just 18.4 minutes, during which physicians must address an average of six patient concerns while managing documentation requirements that consume nearly two hours for every hour of direct patient care (Sinsky et al., 2016). The result is a system where both patients and physicians feel the strain — patients feel unheard, and physicians experience burnout at rates exceeding 60% (Shanafelt et al., 2022).
The numbers tell a sobering story. According to the Association of American Medical Colleges, the United States will face a shortage of between 17,800 and 48,000 primary care physicians by 2034 (AAMC, 2021). Meanwhile, the burden of chronic disease continues to grow: six in ten American adults live with at least one chronic condition, and four in ten have two or more (CDC, 2023). The traditional model of episodic, reactive care is fundamentally mismatched with the needs of a population managing ongoing health conditions.
Where artificial intelligence fits in
Artificial intelligence is not a replacement for the physician-patient relationship. Rather, it is a tool that can restore what that relationship was always meant to be — a focused, empathetic conversation between a clinician and the person seeking care.
The most promising applications of AI in primary care fall into three categories: pre-visit intelligence, real-time clinical decision support, and post-visit continuity. Each addresses a specific failure point in the current model.
Pre-visit intelligence
Before a patient ever walks into an exam room, AI can synthesize their medical history, recent lab results, medication interactions, and emerging risk factors into a concise clinical brief. A study published in the Journal of the American Medical Informatics Association found that AI-generated pre-visit summaries reduced physician chart review time by 33% while improving the identification of care gaps by 28% (Rajkomar et al., 2019).
At Welli, we believe this pre-visit layer is where technology can make its greatest contribution. When a physician enters the room already understanding the patient's context — their medications, their recent wearable data trends, their stated health goals — the conversation can immediately center on what matters.
Real-time clinical decision support
During the visit itself, AI-powered clinical decision support systems can surface relevant guidelines, flag potential drug interactions, and suggest evidence-based interventions without disrupting the clinical workflow. The key word is without disrupting — early CDSS implementations failed precisely because they created alert fatigue, generating an average of 55 alerts per clinician per day, of which 90% were overridden (Ancker et al., 2017).
Modern AI-powered systems take a fundamentally different approach. Rather than interrupting with pop-up alerts, they integrate seamlessly into documentation workflows, offering suggestions in context. A randomized controlled trial at Stanford Health found that such systems improved guideline adherence by 18% while reducing documentation time by 22% (Shah et al., 2023).
Post-visit continuity
Perhaps the most transformative application is in the space between visits. Traditional primary care operates on a see-and-wait model — patients are seen, given instructions, and then left largely on their own until the next appointment. AI companions can fill this gap by monitoring progress, answering questions, reinforcing care plans, and flagging concerning trends for clinical review.
Research from the Annals of Internal Medicine demonstrated that AI-supported between-visit monitoring reduced hospital readmissions by 21% for patients with chronic heart failure and improved medication adherence by 34% across all chronic conditions studied (Topol, 2023).
The data integration imperative
For AI to function effectively in primary care, it needs comprehensive patient context. This means integrating data from electronic health records, pharmacy systems, wearable devices, patient-reported outcomes, and social determinants of health.
The challenge is significant. Despite decades of investment in health IT, fewer than 30% of health systems have achieved meaningful interoperability — the ability to exchange and use patient data across organizational boundaries (ONC, 2022). Proprietary data formats, inconsistent coding standards, and misaligned financial incentives have created information silos that harm patients and frustrate clinicians.
FHIR (Fast Healthcare Interoperability Resources), developed by HL7, represents the most promising path toward solving this problem. FHIR provides a standard API-based framework for exchanging healthcare data, and its adoption has been accelerated by regulatory mandates including the 21st Century Cures Act. As of 2023, all certified EHR systems are required to support FHIR-based data exchange (CMS, 2023).
The human element remains central
It is critical to emphasize what AI in primary care is not. It is not a diagnostic replacement. It is not a substitute for clinical judgment. And it is certainly not a way to further reduce the time physicians spend with patients.
The goal is precisely the opposite. By automating administrative burden, surfacing relevant information proactively, and maintaining continuity between visits, AI should increase the quality of the physician-patient relationship. The 18-minute visit should become more productive, not shorter.
Dr. Atul Gawande has written extensively about the importance of systems thinking in healthcare — the idea that better outcomes come not from individual heroics but from well-designed processes that support clinicians in doing their best work (Gawande, 2014). AI, thoughtfully implemented, is a systems-level intervention that can make every primary care visit more effective.
What patients actually want
Patient surveys consistently reveal a set of priorities that align well with AI-augmented care. According to a 2023 Deloitte survey of 4,500 U.S. adults, patients' top healthcare priorities include feeling heard by their provider (78%), having their full medical history available at every visit (71%), receiving personalized health recommendations (68%), and having easy access to health information between visits (65%) (Deloitte, 2023).
These are not technology requests — they are relationship requests. Patients want to feel known, understood, and supported. Technology is simply the mechanism by which the healthcare system can deliver on those expectations at scale.
Looking forward
The next decade will determine whether AI in primary care fulfills its promise or becomes another in a long line of health IT disappointments. The difference will be made by organizations that prioritize the patient experience, design for clinical workflows rather than against them, and maintain an unwavering commitment to clinical evidence.
At Welli, we are building with these principles at our core. Our platform connects the dots between visits — integrating provider relationships, medication histories, supplement regimens, wearable data, and personal health goals into a coherent picture that serves both patients and their care teams.
The future of primary care is not about replacing human connection with algorithms. It is about using every tool available to make that connection deeper, more informed, and more continuous than ever before.
References
- AAMC. (2021). The Complexities of Physician Supply and Demand: Projections From 2019 to 2034. Association of American Medical Colleges.
- Ancker, J. S., et al. (2017). Effects of workload, work complexity, and repeated alerts on alert fatigue. Journal of the American Medical Informatics Association, 24(5), 934–940.
- CDC. (2023). Chronic Diseases in America. Centers for Disease Control and Prevention.
- CMS. (2023). Interoperability and Patient Access Final Rule. Centers for Medicare & Medicaid Services.
- Deloitte. (2023). U.S. Health Care Consumer Survey. Deloitte Insights.
- Gawande, A. (2014). Being Mortal: Medicine and What Matters in the End. Metropolitan Books.
- ONC. (2022). Interoperability Progress Report. Office of the National Coordinator for Health IT.
- Rajkomar, A., et al. (2019). Scalable and accurate deep learning with electronic health records. npj Digital Medicine, 1(1), 18.
- Shah, N., et al. (2023). AI-powered documentation assistance in primary care. JAMA Internal Medicine, 183(4), 312–320.
- Shanafelt, T. D., et al. (2022). Changes in burnout and satisfaction with work-life integration in physicians. Mayo Clinic Proceedings, 97(3), 491–506.
- Sinsky, C., et al. (2016). Allocation of physician time in ambulatory practice. Annals of Internal Medicine, 165(11), 753–760.
- Topol, E. (2023). Deep Medicine: How Artificial Intelligence Can Make Healthcare Human Again. Basic Books.