When EHR vendors introduced structured note templates in the 2010s, they made a reasonable promise: if you constrain what clinicians enter, documentation becomes faster and more complete. A template pre-populates the review-of-systems checkboxes, reminds you to capture BMI, and prevents the blank-page paralysis that comes with a complex visit. For a decade, that was roughly the best available option.
Ambient AI scribes work from a different premise. Instead of shaping what you type, they remove typing from the equation almost entirely. The system listens to the patient encounter as it happens, then assembles a structured note draft — HPI narrative, ROS, assessment, and plan — while you're still in the room. The question clinicians reasonably ask is: which approach actually returns more time?
The honest answer is that these two tools are not quite competing for the same slot in the workflow. Understanding why helps clinicians decide where each belongs.
What EHR Templates Actually Do Well
Templates have real strengths that ambient AI doesn't replicate. A well-designed diabetes management template enforces documentation standards that payers and accreditation bodies expect — it ensures the A1C date is recorded, that foot exam findings appear in a discrete field, and that the statin conversation is captured in a structured way that supports quality reporting. In systems running HEDIS or MIPS quality measures, template compliance matters in ways that ambient-generated prose often does not.
Templates also perform predictably across staff. A medical assistant can complete a template-driven pre-visit workflow, populating vitals, medication reconciliation, and reason-for-visit in structured fields before the physician enters the room. That's work ambient AI cannot intercept because the conversation hasn't begun. The efficiency gain from templates in those upstream steps is real and durable.
Where templates fail is in the cognitive and time cost of operating them during a live encounter. A physician managing a 40-year-old with hypertension, knee pain, and incidental anxiety raised by the patient mid-visit faces a branching conversation that no template anticipated. Clicking through irrelevant review-of-systems fields, opening separate encounter sections for each complaint, and manually re-sequencing the note to reflect what actually happened — that's where template overhead accumulates. The documentation happens, but after the patient leaves, and it costs an hour or more per shift.
The Ambient AI Workflow: Where the Time Returns
Consider a plausible scenario from an outpatient internal medicine practice: a solo internist running 20 patients per day, with an average complexity of two to three active problems per visit. Under a template-plus-typing workflow, that clinician might spend 45–60 minutes on documentation after hours, reconciling her typed fragments with the template structure. With an ambient scribe, the note draft is available within minutes of the encounter ending — typically before she's finished the visit wrap-up with the patient.
The time returned is most pronounced in the HPI and the assessment and plan. These are the sections that require the most narrative judgment and the most active writing. A good ambient system captures the patient's words accurately, restructures them into clinical language, and positions them within the SOAP framework. The physician's remaining task is review, correction, and approval — not composition.
Objective findings (vital signs, physical exam measurements) still require manual entry or device integration. Ambient AI does not replace the discrete data fields that templates handle well. A physician who finds the template's structured exam section useful should keep it. The ambient layer complements that workflow; it doesn't supersede it.
The Hidden Cost of Template Dependency
There is a subtler issue with heavy template reliance that gets less attention: the note that emerges often doesn't reflect the clinical reasoning that actually occurred. Template notes can read as collections of filled checkboxes rather than clinical narratives — they satisfy billing documentation requirements while obscuring the physician's differential thinking. From a medicolegal standpoint, a note that reads "reviewed and updated" without a coherent narrative of the encounter provides weaker documentation than a prose account of what was found and why decisions were made.
Ambient-generated notes tend toward narrative, which more closely mirrors how physicians actually think through a case. Whether that narrative accurately captures the right clinical content depends on the quality of the system's medical language model and how well it handles specialty-specific terminology. This is a real limitation in current systems and one worth examining before deployment.
Where Ambient AI Falls Short — And Why That Matters
We're not saying EHR templates are a legacy mistake that ambient AI will eliminate. Templates remain the right tool for structured quality reporting, pre-visit workflows, and environments where discrete data capture drives downstream clinical logic (clinical decision support alerts, order sets, quality dashboards). A practice that has invested heavily in template-driven workflows for MIPS reporting should not abandon that infrastructure for ambient documentation.
Ambient AI is also not immune to accuracy errors. Background noise in a busy exam room, patient accents, rapid speech, and simultaneous conversation from multiple speakers all degrade transcription quality. Medical terminology remains a challenge for general-purpose speech models — a patient who says "I've been having trouble with my sugar" may or may not map cleanly to a note entry about glucose management depending on the system's clinical reasoning layer. Every ambient-generated note requires physician review before sign-off, and that review step cannot be treated as a formality.
Consent is a workflow consideration that templates never created and ambient scribes always will. Informing patients that an AI system is capturing their visit conversation is both an ethical obligation and, in many states, a regulatory one. That conversation adds ten to thirty seconds to each visit and requires a clear practice policy. Some patients will decline. The workflow must accommodate that without disrupting the encounter.
Thinking About This Practically
The most productive framing isn't "ambient AI versus EHR templates" — it's identifying which documentation tasks each approach handles well, then building a workflow that uses both appropriately. In practice, that often means: structured templates for pre-visit data collection, quality measure fields, and order sets; ambient AI for the HPI narrative, assessment, and plan sections that require active writing.
Clinicians evaluating ambient tools should specifically test note accuracy in their specialty context. A family physician managing chronic disease, behavioral health, and preventive care in the same visit has different documentation demands than an urgent care provider seeing undifferentiated acute complaints. The ambient system's performance should be evaluated against the actual case mix it will encounter, not a curated demo.
The time savings question is real — but the more clinically important question is whether the resulting note accurately represents what happened and supports continuity of care. Time saved on a note that requires extensive correction is not time saved at all. That's the standard worth holding ambient AI scribes to, and the standard that will separate tools worth adopting from tools that generate a different kind of documentation burden.