AI Scribes Help Reduce Burnout, With Lee Schwamm, MD

Many clinicians are turning to artificial intelligence (AI)–powered documentation assistants to reduce administrative burden and improve efficiency, yet not all physicians are immediately convinced. Many are comfortable with long-established electronic health record (EHR) workflows or prefer to maintain full control over documentation.

According to Lee Schwamm, MD, chief digital health officer at Yale New Haven Health System, such hesitation is understandable but may be short-lived once clinicians see the impact these tools can have on both workflow and patient engagement.

“The technology is not for everyone,” Schwamm said. “There are some providers who are extremely facile with the electronic health record [and] have structured their notes. They use lots of drop-down lists or checklists. They've really optimized their use of the electronic health record, or they don't see that many patients in a given week, or the encounter itself is really focused on the exam and not as much on the conversational parts. So, for those physicians, it might not be the right choice, but I think for a skeptical physician, the first thing to do is to get them to talk to their peers who are using it and have found it useful.”

AI scribes, such as those powered by platforms like Abridge, use natural language processing to record the clinician–patient conversation and automatically generate structured notes for clinician review. A new multicenter study showed this tool reduced physician burnout from 51.9% to 38.8% after 1 month of use (P <.001).1,2

To make the most of the tool, some physicians may need to subtly adjust their communication style. The notes capture what’s said aloud, meaning that clinicians who tend to write thoughts silently after the visit may need to verbalize more of their reasoning in real time.

Schwamm noted this shift reflects a broader evolution in the patient–physician dynamic. In the past, medical notes served primarily as a record for colleagues or billing, not as a communication tool for patients. But as patient access to records has increased, transparency during visits—explaining diagnoses, treatment options, and care plans aloud—has become integral to trust and engagement.

Earlier AI documentation tools were largely limited to transcription, but newer iterations extract structured clinical data and can integrate directly with the EHR. Schwamm said this capability opens new opportunities for real-time clinical support, from generating differential diagnoses to preparing medication orders and coding recommendations.

Concerns about safety and accuracy remain central to conversations about AI, but Schwamm emphasized that AI scribes operate with a “human-in-the-loop” model, or augmented intelligence, where physicians maintain complete oversight of the generated note before it becomes part of the record. Transparency tools, such as the ability to trace each phrase to the original transcript or audio, further reinforce trust and accountability.

At Yale, Schwamm and colleagues are exploring how AI scribes can enhance medical education, allowing students to learn note-taking fundamentals while integrating digital tools into their workflow. He said the team is testing hybrid approaches in which trainees write their own notes, then review AI-generated versions to combine accuracy with efficiency.

“We have to be really focused on [trainees writing their own notes] so we don't lose important skills,” Schwamm said. “The other thing is that we should really be trying to segregate the things that human brains do really well, like reasoning, from the things that AI does really well, like recognizing and matching patterns, and so AI should be the best possible assistant that your doctor could have, whispering in their ear all the time about what might be happening or what they should consider or how they might choose between several options, but fundamentally respecting the ability of the physician to engage with patients, draw out from the patients the important information, and translate that information into medical terminology to enable them to make the right diagnosis.”

A relevant disclosure for Schwamm includes serving as a volunteer member of a research advisory committee for Abridge to facilitate the use of ambient data for clinical researchers.

References

1. Olson KD, Meeker D, Troup M, et al. Use of Ambient AI Scribes to Reduce Administrative Burden and Professional Burnout. JAMA Netw Open. 2025;8(10):e2534976. doi:10.1001/jamanetworkopen.2025.34976

2. Schwamm L. AI Scribes Reduce Physician Burnout Within 30 Days, With Lee Schwamm, MD. HCPLive. October 27, 2025. https://www.hcplive.com/view/ai-scribes-reduce-physician-burnout-within-30-days-lee-schwamm-md. Accessed October 28, 2025.