Dr. Vivek Singh talks with Dr. Timothy Judson, Chief Population Health Officer at UCSF, about how AI can act as a force multiplier against some of healthcare's greatest challenges, including access and affordability. One in three Americans carries medical debt, and one in three lacks a primary care provider. Access, affordability, and quality have long behaved like an iron triangle, where improving one comes at the expense of another. Dr. Judson believes AI is the first real opportunity to break that tradeoff, but only under the right conditions. Here are three takeaways from the conversation.
Most physicians still work in fee-for-service systems, where efficiency gains translate directly into seeing more patients rather than driving down cost or improving outcomes. Without a shift toward value-based care, AI becomes additive instead of transformative. UCSF has already applied AI to risk adjustment, the process of accounting for how sick a patient population actually is, so a health plan isn't unfairly rewarded for enrolling only the healthiest patients. This has long been a core pain point of value-based care, and Dr. Judson sees an even larger opportunity in aligning payment structures with the efficiency AI can create.
The traditional model relied on care managers who called patients repeatedly to close a single gap, like an overdue A1C check. But closing that one gap is not the same as staying with a patient over time. Disease changes, and a patient who looks stable today can look very different in six months, so the harder job is keeping people on a longitudinal follow-up schedule that catches those changes as they happen, not just checking a box once. This is where platforms built for longitudinal tracking, rather than a single point-in-time read, start to matter. RapidAI's own aneurysm solution, Rapid Aortic, and third-party oncology tools are built around this kind of ongoing monitoring, following a finding over time rather than closing the loop after a single scan.
Predictive AI has already improved how health systems identify the roughly five percent of patients driving half of total costs. The next step, already underway at UCSF, is agentic AI handling initial outreach at scale while a single care manager supervises a dozen AI agents and steps in only when a case needs escalation. A chart summarization tool built by UCSF's own informatics team now does in seconds what used to take a care manager up to twenty percent of their time, reviewing a chart to confirm whether a patient qualifies for a complex care program.
Vendors regularly bring UCSF strong tools, but the harder problem is almost always whether a tool fits into existing clinical workflows without requiring teams to leave the systems they already use. Trust builds in two ways: transparency into why a model reached its conclusion, such as seeing the top reasons a patient was flagged high risk, and direct experience watching a tool perform over time. Dr. Judson also points to a parallel in radiology, where incidental findings like a lung nodule can slip through the cracks unless a discrete diagnosis triggers a tracked care pathway. Closing that kind of gap is where population health and radiology increasingly overlap.
The throughline across the conversation is that AI cannot manufacture more physicians or more appointments, and none of these tools work in isolation. What they can do is help a system deliver more consistent, better coordinated care to far more people than a manual, one-to-one model ever could, provided the workflow, the incentives, and the trust are all built to support it.
Interested in seeing how an enterprise platform spanning radiology, care coordination, and patient management can enable better outcomes and access to care? Explore the Rapid Enterprise Platform.
Want to hear the full conversation? Listen to Season 2, Episode 5 of Radiology Rewired with Dr. Timothy Judson, available now at [PODCAST URL PLACEHOLDER].
AI can help health systems care for more patients by reducing manual work, extending clinical capabilities into underserved communities, and connecting care teams. Supporting earlier intervention and fewer unnecessary delays or transfers can also help lower the cost of care.
AI is helping care teams identify at-risk patients earlier, track disease progression, and prioritize follow-up across larger populations. By turning imaging and patient data into actionable insights, it can support more proactive care and help fewer patients fall through the cracks.
The technology itself is rarely the limiting factor. The bigger challenges are whether a tool integrates into existing clinical workflows without disrupting them, and whether clinicians trust its outputs, which is often built through transparency into how a model reached its conclusion, such as through deep clinical AI outputs, and direct experience watching it perform over time.