Nabla Copilot Review

Clinicians spend a substantial part of every day writing notes, and almost none of that time is spent on the patient in front of them.

The documentation burden is not a minor irritation. It is one of the most consistently cited drivers of clinical burnout, it extends the working day past the last appointment, and it turns a conversation into a typing exercise conducted over a screen.

Ambient documentation is the most convincing application of language models in medicine precisely because it does not practise medicine. It listens, and it writes down what was said — which is the part of the job nobody trained for and nobody wanted.

What Nabla Ships Today

• Ambient documentation — it listens to the consultation and produces a structured clinical note without the clinician typing.
• Dictation and coding too: clinical-grade speech recognition inside the EHR, plus E/M and ICD-10 suggestions aligned with AMA guidelines.
• Deep EHR integration with Epic, athenahealth, Oracle Health, NextGen, Arya EHR and Greenway Health.
• Deployment at real scale: 130+ health organisations, 85,000+ clinicians, 20M+ patient encounters annually, with CVS Health, Carle Health and Denver Health named.
• HIPAA, SOC 2 Type 2, ISO 27001 and GDPR, with offices in Brooklyn and Paris.

What Does Ambient Documentation Actually Change?

Nabla's ambient documentation captures the patient-clinician conversation and generates a structured clinical note from it automatically.

The change is in where the clinician's attention goes. Typing during a consultation splits it — the clinician is listening, deciding and transcribing at once, and the patient is talking to someone looking at a monitor. Removing the transcription lets the consultation be a conversation again.

That is a quality-of-care argument as much as an efficiency one, and it is the more durable of the two. Patients disclose more to someone looking at them, and a clinician not composing a sentence while listening catches more of what was said.

The efficiency case is real too and is usually stated as time reclaimed after hours. Documentation that would otherwise be finished in the evening — the phenomenon clinicians call pyjama time — is largely done by the end of the appointment.

The critical distinction is what the tool does not do. It does not diagnose, does not decide, and does not carry any clinical responsibility. It records a conversation and structures it, and every word of the resulting note remains the clinician's — which is why review before signing is not optional.

The distinction from general transcription is worth holding. A meeting tool like Otter.ai produces a record of a conversation; a clinical note is a legal document with a required structure.

Ambient documentation is convincing precisely because it does not practise medicine. It writes down what was said, which is the part nobody trained for.

Coding Suggestions And Where The Money Is

Nabla also offers coding: E/M and ICD-10 suggestions aligned with AMA guidelines, and this is where the platform touches revenue rather than time.

Coding decides what a visit is reimbursed at, and clinicians are systematically bad at it — not through carelessness but because the guidelines are complex, they change, and nobody became a doctor to study billing taxonomy.

The characteristic error is under-coding. A clinician who did the work of a higher-level visit but documented it at a lower level loses the difference, and does so quietly and repeatedly. Suggestions drawn from what the note actually contains correct that asymmetry.

The risk is the mirror image and needs saying plainly: coding above what the documentation supports is not an administrative error but a compliance problem, and in some contexts a serious one. A suggestion is a prompt for a human decision, never a justification.

The honest framing is that this feature is where ambient AI most obviously pays for itself, and also where it needs the most governance. A clinic deploying it should decide in advance who reviews coding suggestions and against what standard.

EHR Integration Is The Whole Ballgame

Nabla lists integrations with Epic, athenahealth, Oracle Health, NextGen, Arya EHR and Greenway Health, and this list matters more than any accuracy claim.

The reason is what happens without it. A documentation tool that produces excellent notes in its own interface has created a copy-and-paste task, and a clinician who must move a note manually between systems has saved very little and gained a second place for things to go wrong.

Integration is also brutally hard in healthcare. Epic in particular is a substantial engineering and commercial undertaking, and the presence of a working integration is itself evidence that a vendor has cleared reviews most cannot.

The list's composition tells you who Nabla is built for. Epic and Oracle Health cover large health systems; athenahealth, NextGen and Greenway cover ambulatory and independent practices. That is deliberate breadth rather than a single-segment bet.

The practical question for any specific buyer is not whether an integration exists but how deep it goes — whether the note lands in the right encounter, in the right template, with the right metadata, or whether it arrives as a block of text somebody has to file.

Anyone tempted to build this themselves on a speech API such as Deepgram or AssemblyAI should price the EHR integration first, because that is where the effort actually lives.

• Confirm not just that your EHR integrates but how deeply the note lands in it.
• Decide in advance who reviews coding suggestions and against what standard.
• Treat every generated note as a draft the clinician is legally responsible for.
• Check language coverage against your actual patient population.
• Establish patient consent practice for recording before the first pilot session.

Reading 130 Organisations And 85,000 Clinicians

Nabla states 130+ health organisations deployed, 85,000+ clinicians using the platform, and 20M+ patient encounters annually, naming CVS Health, Carle Health, Denver Health, Tia Health and the University of Toledo Health.

These figures are more informative than most vendor numbers, because healthcare procurement is exceptionally slow and exceptionally thorough. A hospital system does not deploy a tool that touches the clinical record without security review, legal review, clinical governance and a pilot.

The encounter figure is the most useful of the three. Organisations and clinicians can be counted generously; 20 million encounters a year is a usage measure, and usage is what distinguishes a deployment from a purchase that nobody opened.

Denver Health is a notable name for a specific reason: it is a safety-net institution serving a complex, diverse patient population, which is a harder environment for speech recognition than a well-resourced private practice.

None of this is audited, and all of it is vendor-stated. But the combination of named customers, an encounter-volume figure and integrations with Epic is a considerably stronger evidence base than a percentage claim, and it is the right thing to weight.

The patient-facing half of the same encounter is a separate product category, served by tools such as Soula, and it carries entirely different risks.

Piloting Ambient AI In A Clinic

Ambient documentation succeeds or fails on details that never appear in a demo: accents, noisy rooms, interruptions, and the clinician who talks in fragments.

A pilot that runs in a quiet office with a cooperative patient proves nothing about a Tuesday afternoon.

1. Settle patient consent practice for recording first, with your legal team, before a single session.
2. Pilot with clinicians who document differently from each other, including the fastest and the most thorough.
3. Include your genuinely hard consultations — strong accents, background noise, interpreters, patients who interrupt.
4. Have each clinician compare the generated note against what they would have written, and record what was missing rather than what was right.
5. Verify the note lands in the correct encounter and template in your EHR without manual filing.
6. Audit a sample of coding suggestions against your compliance standard before letting anyone accept them routinely.

Nabla: Deployment And Compliance

Should Your Clinic Deploy Nabla?

A fit for your clinic when

• Documentation is measurably extending your clinicians' working day.
• You run Epic, athenahealth, Oracle Health, NextGen or Greenway Health.
• Your consultations are conducted in English or French.
• Under-coding is costing you revenue you can demonstrate.
• You have the governance capacity to review notes and coding suggestions properly.

Not yet when

• Your patient population speaks languages the platform does not support.
• Your EHR is not on the integration list, which turns notes into a filing task.
• You cannot establish clear patient consent practice for recording consultations.
• Nobody will own the review standard for generated notes and codes.
• Your documentation burden is administrative rather than clinical, which this does not touch.