Abstract
Letter to the editor
The recent brief communication by Alpert et al. 1 offers a timely exploration of AI scribe uptake among primary care physicians, yet it raises useful questions from a nursing education and simulation perspective. Alpert et al. report limited short-term quantitative change after four weeks of AI scribe use, with the only significant pre-post difference being fewer manually typed characters, down 15,398 (p=0.03). They also show a pattern simulation educators recognize. High utilizers reported lower perceived mental demand (p=0.01) and physical demand (p=0.03) than low utilizers, while most other outcomes did not shift. 1 This aligns with simulation evidence where affective responses and early adaptation often precede stable performance gains. This pattern mirrors what we routinely observe in simulation-based learning: early interactions with a new tool often produce a performance dip while cognitive load reorganizes. Short exposure windows mask the gradual gains that emerge as users internalize technology-supported workflows.
The implementation barriers described, including difficulty structuring notes, challenges with multispeaker encounters, and lack of EHR integration, resemble fidelity disruptions familiar to simulation educators. 1 When system behaviour fails to align with user expectations, cognitive effort increases and trust decreases. In simulation, this is treated not as user error but as a predictable response to low functional fidelity. The AI scribe in this study behaved much like a novice performer, capturing everything with enthusiasm but struggling to distinguish priorities in complex cases. Understanding this behaviour through a simulation lens helps explain why physicians spent additional time reorganizing extraneous content.
A key gap is stakeholder scope. The study centers on physicians, yet documentation burden and communication demands span roles. Nurses document across shifts, coordinate with families and teams, and manage multi speaker, interruption-heavy encounters. Those conditions stress the limits the authors describe, so nurse-inclusive and interprofessional evaluations are needed. Nurses spend large portions of work time in the EHR and documentation. 2 Qualitative evidence also shows nurse concerns about ethical issues and loss of human connection in AI-supported care, which links to empathy and dehumanization worries. 3 Simulation debriefing research shows role, workflow, and cognitive load shape performance under pressure, so AI scribes may behave differently in nurse–patient interactions than in physician visits. Nurse-focused evidence syntheses also highlight workload, ethics, and human-centered care as recurring themes in frontline nurse views of AI adoption.4,5
The strong optimism expressed by participants despite modest short-term gains aligns with evidence in simulation pedagogy, where affective responses often precede measurable performance improvement. A longer implementation period is essential. As the authors note, the current lack of longitudinal studies makes it difficult to judge the technology’s ultimate effectiveness beyond the “early adopter” phase. Iterative testing, and structured user feedback cycles would bring this research closer to established simulation integration strategies. Extending future studies beyond one month and across clinical roles will provide more reliable insights into how AI scribes influence documentation efficiency, cognitive load, and patient engagement.
Alpert et al. provide an important early snapshot of real-world AI scribe use. Integrating simulation science principles into future evaluations would strengthen methodological rigour and improve understanding of human–technology interaction in documentation work.
