Table 3. Clinician-identified human factors for ambient artificial intelligence scribe implementation in intensive care unit settings.

Theme	Key insight	Representative quote
Lack of education and training	Clinicians feel unprepared to work with AIa technologies	“AI is coming for all of our futures … we’re going to be using AI technology, but no one is teaching us anything about it.” (ICUb trainee 3) “The health literacy or the literacy around AI is so poor right now… there’s misinformation and not much familiarity.” (ICU trainee 2)
Need for consent and transparency	Clinicians emphasize the importance of informing the team before using ambient AI scribes	"It should be like a standard practice- you would say at the beginning: ‘We’re going to be using the [ambient AI scribe]. Is everyone OK with that?’" (ICU nurse 3) “Whoever is using it should notify everybody that it’s on and how it’s going to be used.” (ICU nurse 4) “There’s consent fatigue … but there’s still a need to present it elegantly and transparently.” (ICU trainee 4)
Concern for surveillance and psychological safety	Clinicians voiced concerns about surveillance stemming from a lack of clarity regarding how ambient AI scribe data would be used	“What’s actually being kept? Is it an audio recording, a transcript? Is it a summary? I don’t understand what is being kept from the scribe.” (ICU RTc 1) “You’re being watched… and that changes your behavior. It’s like the panopticon.” (ICU trainee 1) “Providers should be protected from surveillance.” (ICU trainee 5)
Communication adaptation in response to ambient AI scribe	Staff may shift language to accommodate the ambient AI scribe	“I think we’re going to adapt how we speak during rounds so that the [ambient AI scribe] can do its best job.” (ICU attending 12) “It will increase professional descriptions and decrease off-the-cuff ridiculous remarks. People actually need to use more formal medical language. Instead of ‘AKId,’ the patient has ‘prerenal acute kidney injury.’” (ICU attending 1) “People would be more cautious. they might consider their word choice more carefully.” (ICU nurse 5) “We might structure our communication a little more differently … if we want to make it a very structured note, I bet we would adhere to that even more.” (ICU attending 15) “It could affect the frankness of how someone describes a case.” (ICU RT 2)
Perceived reduction in cognitive load	Ambient AI scribe may free cognitive resources during rounds	“It’s going to help me just pay attention to the conversation without having to frantically jot down notes at the same time.” (ICU APPe 1)
Preference for personalized output and the need for review	Clinicians want editable summaries that are tailored to specific discussions	"The most important would be accurate transcription … the second most important feature would be that it then organizes that information into some type of paragraph or systems-based note template that I could then review and edit.” (ICU attending 7) “There’s no way I would want it to automatically put in a note without me reviewing it.” (ICU attending 16)
Opportunity to capture structured communication exchanges that are inconsistently or not documented at all	Ambient AI scribes could document key conversations often missed in the chart	“I do think using it for the goals-of-care conversation would be nice. Sometimes the notes on these discussions are sparse and you don’t know why patients went from Full Code to DNRf.” (ICU nurse 6) “I don’t think we really know how people are using the ORg-to-ICU handoff checklist. If you have to place a note now, people will have to use it.” (ICU trainee 2)

^aAI: artificial intelligence.

^bICU: intensive care unit.

^cRT: respiratory therapist.

^dAKI: acute kidney injury.

^eAPP: advanced practice practitioner.

^fDNR: do not resuscitate.

^gOR: operating room.