Medical dictation is older than the EHR. Physicians have dictated notes — narrating the clinical encounter into a recording device, then having that recording transcribed — for decades. The workflow has evolved from physical tape cassettes to phone-based dictation lines to real-time speech-to-text software, but the underlying mechanic has remained the same: the physician narrates, and the narration becomes the note.
Ambient AI scribes represent a different mechanic entirely. Instead of asking the physician to narrate documentation, the system listens to the clinical encounter as a natural conversation and extracts structured clinical content from it. The physician doesn't speak to a documentation system — they speak to the patient, and the documentation happens in parallel.
That difference sounds simple, but it has meaningful implications for how clinicians experience each workflow, what errors each approach produces, and which clinical contexts each fits best. A direct comparison is useful for any clinician evaluating which approach to adopt, or whether to use both.
The Dictation Workflow: Explicit Narration, Explicit Control
Traditional dictation requires a conscious shift in attention. The physician finishes the encounter — or pauses within it — and narrates the note, directing their speech toward documentation rather than toward the patient. For experienced dictators, this is a practiced skill: they've learned to structure their verbal narration to produce the note sections they need, in the order they need them, with the clinical specificity required for billing and medicolegal completeness.
The strength of this approach is explicit control. The physician is the author in real time. Every sentence in the note is something the physician chose to say. There is no downstream inference about what the physician meant; the transcript reflects what was actually narrated. In complex cases — ambiguous diagnoses, multi-layered assessments, notes that require careful clinical reasoning to be legible — dictation gives the physician direct authorship of every nuanced sentence.
Modern speech-to-text dictation software has largely eliminated the transcription delay that characterized older dictation systems. Dragon Medical One and similar medical speech recognition tools can render typed text in near real time, allowing dictation directly into EHR note fields. For physicians who have been using these tools for years, the workflow is fast, reliable, and deeply familiar.
The limitation is that dictation is a parallel task to patient care, not an embedded one. In a 20-minute appointment, dictating a note means choosing between dictating in front of the patient (which some patients experience as impersonal), dictating after the patient leaves (which delays the note and adds time to the schedule), or dictating between patients (which reduces the buffer time that experienced schedulers know is essential). Each of these trade-offs has real consequences in a high-volume clinic.
The Ambient Scribe Workflow: Passive Capture, Active Review
Ambient documentation removes the narration step. The system captures the encounter conversation — both clinician and patient speech — without any additional input from the physician during the encounter. After the visit ends, a draft note is generated and presented for physician review.
For physicians who find dictation disruptive to patient interaction — who prefer to maintain full eye contact, move freely in the exam room, and let the conversation develop naturally — ambient capture is genuinely different. The visit runs as it always would. There is no documentation mode to enter and exit. The cognitive overhead of narrating documentation while simultaneously providing clinical care is eliminated.
Consider a plausible outpatient scenario: an urgent care physician in a busy New England walk-in clinic, managing a high-volume shift with a mix of respiratory infections, musculoskeletal complaints, and a few complex multi-system presentations. Under a dictation workflow, documentation between patients consumes the 3–5 minute room-turnover window, compressing the schedule when encounters run long. With ambient capture, the note draft is ready when the physician leaves the room — review takes 60–90 seconds for an uncomplicated visit, and she's not accumulating a documentation backlog as the shift progresses.
The trade-off is that the physician is no longer the explicit author at the moment of capture. The draft is an interpretation of the conversation, not a direct transcription of the physician's narrative intent. Complex clinical reasoning that wasn't explicitly verbalized may not appear in the draft at all, or may appear imprecisely. That requires a different review standard than reviewing a dictated note — not just confirming that the transcription is accurate, but actively verifying that the note represents the clinical reasoning that actually occurred.
Accuracy Profiles: Different Failure Modes
Dictation and ambient capture fail in different ways, and understanding those failure modes helps with review.
Dictation errors are typically transcription errors: the speech recognition system misheard a word. "Lisinopril 10mg" becomes "lisinopril 10mm." "No shortness of breath" becomes "no shortness of breadth." These errors are usually visually detectable on review — they read wrong, sound wrong, or don't match clinical convention. They're also bounded: the error is within a sentence the physician actually said, not in content the physician didn't say at all.
Ambient AI errors include transcription errors of the same type, but also add errors at the interpretation layer: content attributed to the wrong speaker, clinical information from the patient's account misclassified as physician-stated findings, or — in systems that use generative approaches for the Assessment and Plan — content that is clinically plausible but was not actually the physician's conclusion. This last category is the most important to watch for, because it can appear fluent and reasonable without being accurate.
We're not saying dictation has better accuracy than ambient AI in absolute terms — that depends heavily on the specific tools being compared and the clinical context. The point is that they have different error signatures that require different review strategies.
Consent and Patient Experience
Dictation, in its classic form, doesn't require patient consent — the physician is speaking about the patient after or between encounters, not during them. When dictation happens in the patient's presence, patients are generally aware that documentation is occurring, but the conversation is between the physician and the documentation system, not involving the patient directly.
Ambient capture is different. The patient's voice is being recorded and processed as part of the documentation workflow. Informed consent — or at minimum, clear notice — is both an ethical standard and, in many jurisdictions, a regulatory one. Practices using ambient documentation should have a clear consent process: verbal explanation and notation in the chart, a written notice in the patient intake materials, or both. Patients who decline ambient recording should be accommodated without difficulty, and the practice workflow should not penalize the physician for those encounters by creating a documentation gap.
Most patients, in practice, are accommodating once the purpose is clearly explained. A brief statement — "I'm using a clinical documentation assistant that will help me generate your visit note; your conversation will be processed for that purpose" — is sufficient for most encounters. The conversation is short but important, and it distinguishes practices using ambient tools responsibly from those that don't address consent at all.
Which Makes Sense for Which Setting
For clinicians who are highly proficient dictators, have invested in medical speech recognition software, and have developed a dictation style that integrates naturally into their workflow, switching to ambient capture is not an obvious improvement. Dictation gives them direct authorship and a familiar review process. The ambient scribe's advantage — eliminating the narration step — doesn't save them much if they've already made dictation nearly invisible.
Ambient capture makes more sense for clinicians who find dictation disruptive to patient interaction, who run high-volume clinics where any per-encounter overhead compounds across the schedule, or who haven't developed a systematic dictation practice and are currently documenting from memory after hours. It's also better suited to complex, multi-party encounters — patient visits with a caregiver present, pediatric visits with a parent, encounters with an interpreter — where dictating in the room is particularly awkward.
Some clinicians use both approaches situationally: ambient capture for straightforward visits, dictation for complex cases where they want explicit control over the Assessment narrative. There's nothing wrong with that approach — the choice of documentation method per encounter is a workflow decision, not a commitment. What matters is that the note that results is accurate, reviewable, and clinically sound. The route to that outcome is secondary to the quality of the destination.