After Two Years With an AI Scribe, a DPC Doctor Could See the Work No Metric Captures

Dr. Cory Annis runs 45-minute visits at Unorthodoc, her DPC practice in Carrboro, North Carolina. A patient comes in for knee pain. The conversation moves through insomnia, gardening, a family reunion, medication side effects, grief over a recently deceased parent, blood pressure and a new puppy. Two years ago, Annis turned on an AI scribe and let it listen.

The scribe organized every visit into a clean SOAP note. Knee pain landed in the chief complaint. Medication side effects went into the review of systems, filed correctly, sourced to the right part of the conversation. The gardening, the grief and the puppy got cut.

That gap between the note and the visit taught Annis something she hadn’t been able to articulate in four decades of practice.

What the AI Captured

The scribe did what it was built to do. It listened to 45 minutes of winding conversation and produced structured clinical documentation without being trained on Annis’s style. It recognized what was clinically relevant, organized it and excluded the rest.

“I could see the residue of decades of cognitive work that no quality metric will ever measure,” Annis wrote in a recent essay for DPC News. During every visit, she’s questioning, refining, sorting information, tracking timelines, recognizing patterns, weighing significance and filtering distractions, all at the same time. The AI made that invisible labor visible by performing a simplified version of it.

What Got Filtered Out

The dog stories. The vacation plans. The grief over a parent. From a documentation standpoint, the AI was right to cut them. A SOAP note doesn’t need to know about the puppy.

Annis sees those conversations differently. “Those conversations were not irrelevant,” she wrote. “They were doing some of the most important work of the visit.”

In a DPC practice with 30 to 60 minute appointments and panels of 400 to 800 patients, those conversations build the trust that makes the clinical work land. A patient who talks about her garden is a patient who will also talk about the lump she found last week. That kind of disclosure doesn’t happen in a 7-minute slot with a doctor she sees once a year.

Why DPC Practices Have a Different Relationship With This Technology

AI scribes are spreading across healthcare. A 2025 rapid review in JMIR found they reduce self-reported documentation time and improve physician engagement. Multiple health systems have deployed them in outpatient settings. Elation Health acquired a voice AI startup earlier this month to build ambient documentation into its EHR.

The tool works differently in a DPC setting than in a fee-for-service clinic. In a 7-minute visit, there isn’t much conversation to capture. The note is the visit. In a 45-minute DPC visit, the note is a fraction of what happened. The AI produces a cleaner record, and it also surfaces what Annis calls the invisible cognitive work: the simultaneous processing that makes a physician visit something more than a checklist.

Annis’s conclusion reaches something DPC advocates have argued for years, from an unexpected direction. “Medicine is not just information processing,” she wrote. “It is interpretation and judgment and relationship.” The AI made that visible by showing what a visit looks like when you strip it down to the documented parts.

What This Means

For DPC doctors considering an AI scribe: the technology works. It handles the documentation burden that eats into the time you built your practice to protect.

For physicians still in fee-for-service and thinking about whether DPC is the right move: this is what a 45-minute visit looks like from the inside. The AI scribe showed Annis what she’d been doing all along: cognitive work that only exists when you have enough time with each patient to do it.

Annis has been in practice for more than 40 years, with over a decade in DPC. She started Unorthodoc in 2015 after a $20,000 insurance clawback pushed her out of the insurance model. Her memberships start at $54 a month. Two years with an AI scribe gave her new language for work she’s done her entire career.