Podcast Episode: Transforming Healthcare: Why the Real AI Revolution Isn’t in the Pilot—It’s in C

Pip: Welcome to the CDO TIMES podcast — where we track the gap between what technology promises and what actually survives contact with reality.

Mara: Today we're working through Carsten Krause's latest writing on healthcare AI — specifically the hard question of what it takes to move from a promising pilot into something clinicians actually trust and use at scale. Let's start with why that gap is so difficult to close.

Transforming Healthcare: From Pilots to Clinical Production

Pip: The central tension here is one that shows up across every regulated industry, but healthcare makes it unforgiving: an AI system can be technically impressive and still be clinically useless if it hasn't earned the trust of the people whose names are on the charts.

Mara: The piece draws on a panel discussion called "Transforming Healthcare: From AI Pilots to Clinical-Grade Systems," and Greg Kuehnen of UNC Health sets the frame directly: "We have been ruthless about value."

Pip: And ruthless is doing real work in that sentence. This isn't about ROI dashboards — Kuehnen's definition of value includes physician experience, patient experience, quality, community access, and financial sustainability all at once.

Mara: The most mature use case the panel surfaces is ambient clinical documentation. Aaron Caine, CIO at Community Health Center, describes the core problem plainly: "One of the biggest pain points for providers is transcribing the notes into the system." After deploying ambient scribes, encounter closure rates improved measurably.

Pip: Kuehnen's framing of the same technology is worth sitting with — "Clinical documentation, ambient scribes, fantastic game changer. That's not why anyone got into medicine in the first place."

Mara: So the upshot is that the first real wave of healthcare AI may not be about diagnosis at all. It's about reclaiming the hours physicians spend on administrative overhead — and that has direct downstream effects on burnout and retention.

Pip: The administrative angle runs wider than documentation. The panel lists scheduling, call center automation, prior authorization, revenue cycle, claims processing — and the American Hospital Association puts administrative costs at roughly twenty-five percent of U.S. healthcare spending.

Mara: One of the more striking examples is UNC Health's scheduling assistant. It was built to ease staffing pressure and reduce wait times. Then Kuehnen noticed something unexpected in the patient satisfaction data: "Patients who speak non-English were rated as much higher." The system supported sixteen languages by default.

Pip: A capacity tool quietly became a health equity tool. That kind of unintended upside is worth noting — it's not something you'd have discovered in a pilot that ran for ninety days.

Mara: Caine adds a leadership dimension that's easy to overlook. He personally immerses in emerging technologies before introducing them organizationally, and he recommends removing the word AI from many conversations entirely. His reasoning: "Remove AI from the conversation and talk about it merely as a tool." It reframes the discussion around workflow outcomes rather than technology fear.

Pip: And on the data side, Kuehnen flags that EMR vendors themselves are a bottleneck — delayed access, limited APIs, expensive licensing. AI readiness turns out to be a data architecture problem as much as a model problem.

Mara: The panel closes on trust as the non-negotiable. Kuehnen's standard for patient-facing AI: "We will commit to you that it's being done for your benefit." Governance, explainability, and human oversight aren't features — they're the foundation.

Pip: Which makes the next question almost inevitable: if trust is the foundation, what does the governance infrastructure actually look like when you try to build it?

Mara: The through-line across all of this is that the hardest part of healthcare AI isn't the algorithm — it's the institutional credibility required to deploy it.

Pip: Trust as infrastructure. We'll be back next episode with more from the CDO TIMES.