Ambient Listening in Healthcare: What It Is, How It Works and What to Ask Before You Buy (2026)

Updated October 2026

Ambient listening in healthcare is AI that records a conversation between a clinician and a patient, with the patient's agreement, and turns it into a draft clinical note for the clinician to review and sign. It is also called ambient clinical intelligence, ambient AI or an AI scribe, and it is now built into Epic, Oracle Health and athenahealth as well as sold by dozens of specialist vendors.

This guide explains ambient listening for the people who decide whether to use it: physician and nursing leaders, practice owners, CIOs and the product teams at digital health companies. It covers how the technology works step by step, where it is used, what the evidence shows, the risks, how notes reach the EHR, the main vendors, and the questions to ask before you sign. If you are adding ambient listening to your own product, see our healthcare AI solutions.

What is ambient listening in healthcare?

Ambient listening is software that listens in the background of a clinical visit and writes the first draft of the documentation. The clinician talks with the patient normally, without dictating, and afterwards reviews a structured note, such as a SOAP note, plus any suggested orders or codes.

"Ambient" means the tool works in the background. The clinician does not give commands or narrate the exam. That is the difference from the speech recognition many clinicians already use, which types what the clinician says word for word.

You will see several names for the same idea. Ambient clinical intelligence is the term Nuance (now Microsoft) used for its DAX product. Ambient clinical documentation and ambient AI scribe are the most common general terms. Epic calls its version Chart with Art.

How ambient listening works, step by step

An ambient listening system runs six steps: consent, capture, transcription with speaker separation, clinical understanding, drafting in the clinician's template, and clinician review before the note enters the EHR. Each step has its own failure modes, which is why clinician review never goes away.

1. Consent. The clinician tells the patient the visit will be recorded and asks for agreement. The VA, for example, uses the Veteran's verbal consent and says the tool is always optional.
2. Capture. A phone, tablet or room microphone records the visit. Telehealth tools capture both sides of the call.
3. Transcription and speaker separation. Speech recognition tuned for medical terms turns audio into text, and the system labels who is speaking: clinician, patient, family member or interpreter.
4. Clinical understanding. A language model picks out symptoms, history, exam findings, medications, the assessment and the plan, and ignores small talk.
5. Drafting. The system writes the note in the clinician's template and style, and may suggest orders, diagnosis codes and patient instructions.
6. Review and filing. The clinician edits and signs. The note is filed in the EHR, either natively or through an integration.

Where ambient listening is used today

Ambient listening started in outpatient clinic visits and has spread to emergency departments, hospital wards, nursing, home health, therapy and telehealth. EHR vendors now list ambient drafting for several of these settings, and large public systems such as the VA are rolling it out.

How widely is ambient listening used?

Ambient listening is one of the fastest-adopted technologies in healthcare, but most organizations still use it in a minority of visits. MGMA polls show 28% of medical groups using ambient AI for notes in May 2024, and 71% of practice leaders reporting some AI in patient visits by August 2025.

• May 2024: an MGMA Stat poll of 424 respondents found 28% of medical groups used ambient AI to transcribe speech or draft notes; 69% did not. Many non-users were waiting for their EHR to add it.
• August 2025: a later MGMA poll of 244 respondents found 71% reported some AI use in patient visits, but 47% used it in 25% of visits or fewer. Among AI users, 39% said it reduced workload and 44% said it had not.
• Vendors: the Peterson Health Technology Institute counted about 60 AI scribe tools in the market in its March 2025 report.

Since then, Epic, Oracle Health and athenahealth have all added their own ambient tools, which is likely to push adoption up further among groups that were waiting for their EHR.

Does ambient listening work? What the evidence shows

The evidence is consistent on burnout and mixed on time. Surveys and trials show clinicians feel less burdened, while measured time savings are modest, vary by product and specialty, and depend heavily on how often clinicians actually use the tool.

Two lessons follow. Measure your own baseline before you start, because published results vary widely. And treat adoption as a result in its own right: a tool used in a third of visits saves a third as much as one used in every visit.

Accuracy and safety risks of ambient listening

The main risks are notes that leave out something important, add a finding that was never said, attribute words to the wrong person, or flip a negative into a positive. Clinician review catches most of them, so the design goal is a draft that is quick to check.

In the UCLA trial, clinicians reported clinically significant inaccuracies "occasionally" with both scribes. That is why no responsible deployment lets an ambient draft reach the chart without a clinician's signature.

Consent and privacy for ambient listening

Ambient listening records protected health information, so it needs patient consent, a business associate agreement with the vendor, and clear rules on how long audio is kept and whether it trains AI models. State recording laws in places like California, Florida and Pennsylvania require every party in the room to agree.

• Ask every patient. Even where the law allows one-party consent, asking builds trust. Record the answer in the chart and make declining easy.
• Telehealth counts. Phone and video visits follow electronic recording rules, which are stricter in some states.
• Retention and training. Get the audio retention period and the model-training terms in writing. Epic's first AI Charting customer states that recordings are kept for a limited time and are not used to train AI models.
• Disclosure laws. Texas now requires practitioners to disclose some uses of AI in care. Check your state.

Our ambient AI scribe implementation guide has a patient consent form template you can copy and a state-by-state recording-consent table. For the engineering controls, see what a HIPAA compliant AI medical scribe needs.

How ambient listening notes get into the EHR

Ambient notes reach the EHR in one of three ways: the EHR vendor's own built-in tool, a partner scribe connected through the vendor's program, or a custom integration using FHIR or HL7 v2 interfaces. The route decides whether the note lands in your template, as a draft, on the right visit.

• Built in. Epic Chart with Art, Oracle Health's Clinical AI Agent and athenahealth's own scribe draft notes inside the EHR. See our guide to the Epic AI scribe.
• Partner programs. Abridge, Microsoft Dragon Copilot, Nabla and others connect through Epic's partner programs; athenahealth's Ambient Notes program lets practices pick an engine such as Suki.
• Custom integration. A scribe in your own product writes the note with a FHIR DocumentReference or an HL7 v2 document message. Formats and draft rules differ by EHR, as our DocumentReference guide shows. Our EHR integration services team builds these connections for Epic, Oracle Health and athenahealth.

Ambient listening vendors in 2026

The market splits into four groups: EHR vendors' own tools, enterprise scribes sold to health systems, specialty and self-serve scribes for practices, and platforms that software companies use to build their own. Most prices are not public outside the self-serve group.

For a side-by-side view of cost, EHR support and specialty fit, see our AI medical scribe comparison.

Questions to ask before you buy ambient listening

Ask about specialty fit, EHR write-back, audio retention, model training, consent support, accuracy reporting and total first-year cost. Get the answers in writing, and test the hard cases in your own pilot rather than relying on a demo.

What does ambient listening cost?

Enterprise ambient listening is sold on negotiated contracts without public prices, and EHR vendors have not published prices for their own tools. Self-serve scribes publish monthly plans: Freed, for example, lists plans from $39 to $119 per clinician per month as of October 2026.

Count the full cost: licenses, EHR build work, devices, training and support. Set it against your own pilot results. The Peterson Health Technology Institute found that early adopters saw burnout benefits but no significant financial gains yet, and warned that higher coding levels could raise costs for payers and patients. Treat time saved as capacity, not cash, unless it becomes more visits or less overtime.

Where ambient listening is heading

Ambient tools are moving from notes to the work around the note: suggested orders, diagnosis and visit-level codes, patient instructions, nursing documentation and home health assessments. Each new output needs its own review step and its own accuracy checks.

Epic's Art page already lists order queuing, code suggestions and patient instruction drafting. Oracle Health added order creation to its agent in the US in February 2026. The practical effect for buyers is that a scribe decision is becoming a decision about which AI helps with the whole visit, so EHR fit matters more each year.

How to get started with ambient listening

Start with a small pilot: two or three specialties, clinicians who want to try it, a baseline from your EHR data, and a consent process your counsel has approved. Judge it on eight weeks of use, then decide whether to scale, switch tools or stop.

1. Pick the visits. Conversational outpatient visits are the easiest place to start. Procedures, crowded rooms and visits with interpreters are harder; add them once the basics work.
2. Check your EHR first. If your EHR vendor now offers its own tool, include it in the pilot alongside one other product, so you compare fit rather than features on a slide.
3. Set up consent. Write the script, add a chart field for the answer, and tell patients ahead of the visit. Make declining quick for both patient and clinician.
4. Measure before and after. Track time-in-note, after-hours charting, how often clinicians use the tool, edit effort and errors. Clinician surveys are useful, but pair them with EHR data.
5. Review errors every week. Name a clinical owner, collect every reported error, and feed patterns back to the vendor.

Our ambient AI scribe implementation guide covers each of these steps in detail, including security controls and EHR write-back for Epic, Oracle Health and athenahealth.

How Nirmitee helps with ambient listening

We build ambient documentation for digital health companies and provider groups, and connect it to Epic, Oracle Health and athenahealth. We have built an AI documentation assistant inside a behavioral health EHR, with consent capture, drafts tied to the transcript, clinician sign-off and prompt version control. We are ISO 27001:2022 certified and sign BAAs. We do not resell scribe products, so if an off-the-shelf tool fits your needs, we will tell you.

Related: healthcare AI solutions, our voice to clinical notes case study, and what it takes to build an AI clinical scribe beyond the prototype.