By: Nathan Baggett, MD
My patient encounters begin a little differently today than they did just a few months ago.
After the familiar pleasantries I’ve used to greet patients for years, I now add a new scripted line:
If it’s okay with you, I’m going to have my phone listen to our conversation so it can help write my note behind the scenes.
For today’s trainees, this introduction may feel less disruptive, but for those of us who trained even just a few years ago in an era before artificial intelligence (AI), it represents a remarkable shift in our clinical workflow.
Prior to ambient dictation, I had never worked with a human scribe. My notes were my notes – written in my voice and in the style I had honed from the full “SOAP” note taught in the pre-clinical years of medical school to my department’s streamlined ED-specific note.[1] So when I was asked to be one of the early adopters of the ambient dictation software being rolled out in my department, I was skeptical.
Despite this skepticism, I dutifully reviewed the training documents my health system provided and decided to give it a try. At the start of my next shift, I pulled out my phone, activated the ambient software and placed it on the counter to listen to my patient encounter. The required disclosure line felt clunky, and I felt awkward as I tried to verbalize my physical exam findings in a way that the ambient scribe might understand (“Your lungs are nice and clear and I see that your work of breathing is normal”). After wrapping up my visit, I returned to my desk to review the note generated by my “AI scribe.”
And I hated it.
This AI generated note was long. While it seems to have captured most of the HPI, some key details were missed and in the complex ED environment, elements were misattributed as the AI struggled to navigate cross talk between patient, family members, the nurse, and an EMS crew. The physical exam section missed most of the exam elements I had so awkwardly tried to verbalize during the encounter. But most consequentially, the MDM section included a lot of words but lacked the substance that could capture the complex decision making required to navigate an ED visit.
After trying the ambient scribe for a few patients, my phone returned to my pocket for the rest of the shift. The ambient scribe felt intrusive and its output was underwhelming. I returned to my tried-and-true notes written with my dictation microphone and a keyboard.
Looking back, what bothered me most in my first foray into ambient dictation wasn’t that the note was imperfect or wholly unacceptable. It was that it didn’t feel like my note. After years of authoring my own documentation, I had developed my own voice and workflow which represented how I thought about my patient care. The ambient generated note felt foreign.
Still, I kept trying it. Partly because I had agreed to be one of the early adopters, I felt I had to try and figure it out and partly because I knew some of my discomfort was just a lack of familiarity with a new approach to documentation. With time, the awkwardness of disclosing its use faded and I started to see how I could make this new technology work with my workflow.
I started to see that the ambient software was pretty good at documenting the history of present illness. It maybe didn’t sound exactly like how I would document the visit, but it was sufficient and accurate enough that I could start to trust it more. However, I found that I preferred to write my own physical exam and medical decision-making sections for all but the simplest of patient encounters. This hybrid balance of use helped me to maintain ownership of the parts of the documentation that mattered most – my thinking and thought process.
As I figured out this hybrid workflow, the most surprising change I discovered had little to do with the note itself. After weeks of use, I started to notice I felt less rushed in a patient’s room. I didn’t feel this pull to hurry back to my desk to document the visit while the details were still fresh in my head. I didn’t get caught up trying to remember every part of a patient’s meandering timeline leading to their presentation. I found myself spending more time in the patient’s room. Knowing I had the ambient scribe capturing the details of the history that mattered, I felt an ease with just being present. Rather than listening so that I could document every detail correctly, I could simply listen to ensure that the patient was heard.
Yet my experience as an attending physician is only part of the story.
As educators, we cannot assume that our experience with emerging technologies like ambient dictation or AI in general will match our learners’ experience. Attending physicians may see documentation as a necessary task to complete to facilitate care. But for learners, documentation is more than just a task. Writing is a process of thinking. The act of writing the note helps learners synthesize information, formulate an assessment, and justify a plan.[2] For many learners, the process of writing the note serves as a mechanism for developing clinical reasoning. But with ambient technology that can not only document the HPI but also document the medical decision making, learners can easily offload these tasks.
This raises the question: how do our learners develop their clinical reasoning when systems like ambient dictation or AI-assistants are able to generate their note?
Perhaps these technologies will allow more time for our learners to be present with patients like it has for me. Perhaps this additional space will help to reinforce the humanistic side of medicine that will matter more with the rise of artificial intelligence. But it is also possible that by prematurely offloading cognitive tasks like documentation, we will inadvertently lose a mechanism that builds clinical reasoning and helps learners to think like physicians.
All clinical educators will need to grapple with the implications of AI and tools like ambient dictation in the clinical learning environment. As we do so, some questions to consider include:
- How can we prepare learners to function in an AI-infused future while also ensuring they master the foundational skills of practice?
- Should we be limiting access to these tools for learners? Should access be limited to more senior learners?
- Which aspects of documentation contribute meaningfully to learning and professional development, and which are primarily administrative?
- How can clinical educators assess a learners’ clinical reasoning when learners use AI?
- Will use of tools like ambient dictation create more opportunities for the humanistic side of medicine?
- If learners rely on AI-generated documentation, what educational experiences can replace the cognitive work traditionally performed through note-writing?
The challenge before us is not whether learners will use AI, but how we can help them to use it in ways that will allow them to develop independent clinical reasoning skills while also fostering the humanism that lies at the heart of the patient-physician relationship. the hard work of becoming clinicians who can think, decide, and act responsibly when it matters most.
About the Author
Nathan Baggett, MD is an emergency physician and educator at HealthPartners in St. Paul, Minnesota and an Assistant Professor of Emergency Medicine at the University of Minnesota Medical School. As Director of Artificial Intelligence for his emergency department, he leads initiatives to integrate AI into clinical practice, resident education, and assessment systems. He completed a Medical Education Fellowship in 2025 and is currently completing a Master of Academic Medicine at the University of Southern California, where he explores how AI can transform clinical reasoning, feedback, and training in health professions education. Dr. Baggett earned his MD from the University of Wisconsin School of Medicine and Public Health in 2017.
Photo courtesy of ISTOCK
The views and opinions expressed in this post are those of the author(s) and do not necessarily reflect the official policy or position of The University of Ottawa. For more details on our site disclaimers, please see our ‘About’ page