How AI can help health systems turn capacity into access
As health systems confront physician shortages, long wait times and fragmented access workflows, AI can help uncover hidden capacity through demand forecasting, smarter scheduling, automated waitlist matching and more intelligent referral management.
For years, healthcare AI has been asked to help clinicians make better decisions, from reading radiology images to summarizing charts. That work matters, but it does not solve the operational barriers that determine whether patients can actually obtain the care being recommended. Patients can still wait weeks for appointments or lose momentum when referrals and scheduling workflows break down.
The next major opportunity for AI is therefore not limited to clinical decision support. It is also operational decision support: helping health systems understand demand, use available capacity intelligently, and move patients through access workflows with less friction. In that role, AI can help coordinate the operational decisions that stand between a patient and an available appointment.
The access problem is not theoretical
The U.S. faces a projected physician shortage of up to 86,000 doctors by 2036, according to the Association of American Medical Colleges. At the same time, the capacity that already exists is not always easy for patients or health systems to use efficiently. AMN Healthcare reported in 2025 that the average wait for a new-patient physician appointment across six specialties in 15 major U.S. metropolitan areas reached 31 days, up 19% from 2022.
Those numbers reinforce the premise of the first article in this series: healthcare access is not purely a supply problem. Health systems still need to recruit and retain clinicians, but they also need better visibility into where demand originates, where it stalls, and where appointment capacity is being lost. A December 2025 MGMA poll found that no-shows were the leading patient-access focus reported for 2026, followed closely by online scheduling, phone access, and wait times, underscoring how much of the access challenge lives inside day-to-day operations.
That is where operational AI can become useful. Clinical AI often helps answer, “What should we do?” while access-focused AI can help answer, “How do we make it happen?” A correct clinical recommendation still depends on an operational system that can connect to the right patient, provider, appointment type, and time.
Health systems do not just need smarter diagnoses; they also need smarter orchestration. AI can analyze demand, cancellations, referral urgency, provider availability, visit-type rules, patient preferences, and historical no-show patterns to support scheduling and routing decisions. Used well, that intelligence can reduce empty slots, delays, manual workarounds, and the burden on front-desk teams.
Where AI can make access smarter
Demand forecasting is one practical place to start. AI can analyze appointment volume by specialty, location, season, and patient segment to help organizations anticipate pressure before it becomes a backlog. Rolling forecasts can inform template changes, staffing decisions, and capacity allocation before a service line is booked weeks into the future.
Scheduling optimization is another opportunity. Matching visit types to the right provider, modality, location, and slot requires rules that are often spread across EHR configuration, departmental practice, and staff knowledge. Standardizing those rules and making them machine-readable gives AI something reliable to work with and reduces dependence on sticky notes, tribal knowledge, or the one employee who knows why a particular appointment belongs in a particular block.
Cancellations and no-shows create a different kind of capacity loss. AI-enabled waitlist matching can identify patients who are appropriate for newly available appointments, while risk signals can help tailor reminders, outreach, or rescheduling options. The goal is not simply to send more messages; it is to intervene earlier and more selectively so that an open appointment does not go unused.
Referral management creates a similar opportunity because the work does not end when an order is placed. AI can help route referrals based on urgency, access rules, specialty requirements, and available capacity, while closed-loop workflows track missing information, contact attempts, scheduling status, and next steps. That visibility can reduce stalled referrals and show access teams where patient demand is accumulating.
None of these use cases works well if the underlying operational data is incomplete or inconsistent. Provider directories, scheduling rules, referral statuses, appointment types, and capacity definitions must be standardized enough for an AI system to interpret them reliably. Operational AI does not eliminate the need for data governance; it makes those fundamentals more important.
Where hidden capacity remains
In many organizations, the problem is not only a lack of clinicians. Some usable capacity remains trapped inside disconnected systems, rigid templates, unfilled cancellations, manual referral queues, and inconsistent scheduling rules. Patients can therefore experience long waits even while individual slots, providers, or locations remain underused at particular times.
AI can create the most value when it helps health systems see that capacity more clearly and act on it faster. Clinical decision support may help determine the right care, while access and capacity intelligence can help patients reach that care through a functioning operational pathway. The technology is therefore part of a broader access-management discipline, not a stand-alone answer to the workforce shortage.
Bottom line
The future of healthcare AI is not just about helping clinicians think faster; it is also about helping the healthcare system move patients through care more intelligently. Organizations that combine demand forecasting, smarter scheduling, automated waitlist backfill, targeted reminders, and closed-loop referrals should measure the gains rather than assume them. One 2024 study at a hospital in Türkiye reported a roughly 10% increase in appointment attendance and a 6% increase in capacity utilization after implementation of an AI-based appointment system. Those results are encouraging, but they are site-specific and should not be treated as a guaranteed outcome for every health system.
A separate 2019 study in the United Kingdom showed that machine-learning models could stratify the risk of nonattendance, supporting the idea that predictive tools can help target interventions rather than treating every scheduled patient the same way. Taken together, the studies illustrate the more defensible promise of operational AI: better information can help health systems make more precise decisions about how to use finite capacity. At scale, even modest improvements in attendance, scheduling, referral completion, and slot utilization can translate into meaningful access gains without pretending that technology alone can solve the underlying workforce problem.
The best medical insight in the world still needs an open slot, a routed referral, a prepared patient, and an operational workflow capable of connecting those pieces. Access and capacity management may therefore prove to be one of AI’s more consequential healthcare applications outside direct clinical decision-making. The opportunity is not to automate away care-delivery complexity, but to make it more visible and manageable, so it is less likely to stand between patients and the care they need.
Jake McCarley is CEO of Alluvium.