The North San Diego County medical group reports fewer missed appointments and sharply lower front-office costs after combining AI, human support and a broader operational rebuild—raising a practical question for physician groups about where the technology ends, and better management begins.

Escondido-based Graybill Medical Group reports that patient no-shows declined nearly 24% after it overhauled its patient-access operations with a combination of artificial intelligence, human support and redesigned workflows.

But the experience offers a caution for physician organizations considering similar technology: Graybill did not simply install an AI product and wait for results.

Before introducing AI, the medical group examined call wait times, abandonment rates, appointment access, no-shows, staffing costs, patient complaints and employee feedback, according to Healthcare IT News. Third Way Health, Graybill’s technology and operations partner, then learned the group’s existing processes, identified inefficiencies and introduced AI into selected high-volume administrative tasks.

The effort occurred as part of an even broader transformation. After Third Way Health separated from a managed-services partner in 2024, Graybill undertook a system-wide operational rebuild across nine locations that included changes to information technology infrastructure, telephony, patient-access tools and front-office workflows, according to a Third Way Health case study.

That wider restructuring makes it difficult to isolate exactly how much of Graybill’s improvement came from AI itself.

The reported results are nonetheless significant. Third Way Health says Graybill’s no-show rate declined from 5.1% in the fourth quarter of 2024 to 3.9% in the fourth quarter of 2025—a 1.2-percentage-point absolute decline and roughly a 24% relative reduction.

Healthcare IT News also reported that Graybill reduced certain front-office operating costs by approximately 50%. Third Way Health says the changes produced about $3 million in savings across call-center and medical-records staging operations, including benefits and overhead.

That figure should not be interpreted as independently audited net savings. Publicly available information does not disclose Third Way’s fees, implementation expenses, or other costs needed to calculate the intervention’s full return on investment.

Graybill’s model also was not based on eliminating human interaction. AI handled repetitive tasks such as appointment scheduling, common patient questions, and directing callers to appropriate resources. At the same time, human workers remained involved when conversations required judgment, empathy, or more complicated problem-solving.

For physician practices, the potential operational payoff is straightforward. Fewer missed appointments can improve schedule utilization and potentially expand patient access without requiring additional clinical capacity. Faster call handling and lower administrative expense could also ease some of the financial pressure associated with operating a multisite medical group.

But Graybill’s results do not establish that AI alone caused those improvements.

Workflow redesign, human outreach, new technology, changes in staffing and broader infrastructure improvements occurred together. No independently controlled study has determined how much of the improvement can be attributed to any one component.

Graybill’s experience therefore may offer a different lesson than the usual promise that AI itself will transform medical practices: the organizations that benefit most may be those that identify broken processes first and automate selectively afterward.

Publicly available information does not disclose the number of appointments underlying the no-show calculation, results by specialty or patient population, changes in staffing levels, vendor and implementation costs, or the intervention’s net return on investment.