Data-Informed implementation of ambient scribe training: Tracking adoption and perceptions in family medicine residency

Authors

Jessica Sefen, Advocate Health - MidwestFollow
Kevin Lee, Advocate Health - MidwestFollow
Alonzo Jalan, Advocate Health - MidwestFollow
Meinuo Chen Baca, Advocate Health - MidwestFollow
Bonnie Bobot, Advocate Health - MidwestFollow
Samantha Ladin, Advocate Health - MidwestFollow
Stephanie Malusic, Advocate Health - MidwestFollow
Collin Spencer, Advocate Health - MidwestFollow
Will Lehmann, Advocate Health - MidwestFollow
Keyonna Taylor-Coleman, Advocate Health - MidwestFollow
Deborah Simpson, Advocate Health - MidwestFollow

Recommended Citation

Sefen J, Lee K, Jalan A, et al. Data-Informed implementation of ambient scribe training: tracking adoption and perceptions in family medicine residency. Poster presented at: Alliance of Independent Academic Medical Centers (AIAMC) – National Initiative X: Meeting #2; April 17-18, 2026; Carlsbad, CA.

Affiliations

Aurora Health Care

Presentation Notes

Poster presented at: Alliance of Independent Academic Medical Centers (AIAMC) – National Initiative X: Meeting #2; April 17-18, 2026; Carlsbad, CA.

Abstract

Introduction: AI is rapidly being implemented within clinical medicine and is utilized by medical students, residents, and faculty alike. However, with the exponential development in modern technology, a training curriculum must quickly evolve to keep up with correct-use education, clearing misconceptions, and addressing concerns. We focused primarily on the ambient scribe tool (Nuance DAX Copilot) as it is the most used within our organization.

Aims: To provide faculty and residents within our residency program with structured training and education for effective use of AI-assisted documentation tools to improve patient care through strategic leveraging of clinical AI use.

Methods: To address knowledge gaps in understanding how AI tools work, we utilized teachable moments, incorporated formal lectures into our didactics, and held open discussions at resident faculty meetings. SlicerDicer data from Epic was used to evaluate the impact of ambient use by analyzing time spent writing notes. Participants completed an AI self-assessment tool drawn from the program’s AI Milestone to guide subsequent educational offerings at the midpoint of the project. Semi-structured group interviews framed by the Technology Adoption Model (TAM) constructs were conducted with 5 resident/faculty groups. Small group facilitator notes were analyzed using ChatGPT 5.2 to identify barriers and inform targeted responses.

Results: 35 AI milestone self-assessments with TAM elements were completed (Residents =18; Faculty = 17) with >50% of respondents rating themselves at advanced, beginner, or lower on Dreyfus scale on most items. On the DAX Copilot usage survey, 44 respondents participated in June 2025 (Residents n=24; Faculty n=20) and 34 in February 2026 (Residents n=18; Faculty n=16). Percentage of residents reporting low/no use (Never+Rarely; < 10% of encounters) was unchanged at 50% from June 2025 (12/24) baseline and midpoint February 2026 (9/18). Faculty reporting low/no use increased from 30% in June 2025 (6/20) to 37.5% in February 2026 (6/16). PGY-level patterns varied across timepoints: PGY2 residents reported higher usage in February 2026, whereas PGY1 and PGY3 residents reported higher usage in June 2025. PGY2 residents were strong users (tools were available since the beginning of their residency) and reported stronger influence from peers to use these tools on TAM items. Similar use was reported in PGY1 and PGY3 residents. The highest users generally had the highest concerns regarding AI use. Faculty had the widest range of responses across TAM elements: usage, perceived usefulness and ease or difficulty of use. Using data from Epic residents and faculty were grouped into either high (~80%), medium (30%) or low (1%) users for ambient scribe. The high user group reduced their time spent writing notes on average by 15 minutes, while the other two groups had no sustained improvement.

Discussion: Residents and faculty with the highest ambient scribe use demonstrated a significant reduction in time spent writing notes when analyzed across multiple years. In this sample, survey-based adoption did not increase over time; however, mixed-methods findings identified actionable barriers and opportunities to inform curriculum refinement and implementation. As these tools evolve, more formal longitudinal training is needed to promote appropriate and consistent use. Faculty champions are being identified to support critical analysis of tool use and facilitate adoption among their colleagues. Together, we aim to anticipate emerging challenges and refine educational strategies that keep pace with AI development and its integration into clinical practice.

Type

Poster