Written and maintained by CASRAI Editorial Board
Last updated
Clinical informaticists, digital health architects, and ambulatory practice leaders evaluating Nabla Copilot (often searched as Nabla AI scribe, Nabla medical scribe, or Nabla ambient AI) require a clear technical breakdown of how ambient clinical documentation processes audio, maps clinical entities into Electronic Health Record (EHR) fields, and enforces privacy boundaries. Developed by Nabla (founded in 2018 by Alex Lebrun, Delphine Groll, and Martin Raison), Nabla Copilot is an ambient clinical intelligence platform that listens to patient–clinician encounters and generates structured clinical documentation across more than 55 medical specialties and 35 languages.
What Is Nabla Copilot and How Does Its Ambient Pipeline Work?
Unlike traditional speech-to-text dictation software (Nabla dictation mode, where a clinician speaks verbatim commands into a microphone after an encounter), Nabla Copilot operates passively during an in-person or telehealth consultation via a three-stage natural language processing pipeline:
- Multi-Speaker Diarization and Automatic Speech Recognition (ASR): The audio stream captured via the desktop web interface, Chrome extension, mobile Nabla app, or embedded telehealth SDK is segmented by speaker role (distinguishing the clinician, patient, family members, and medical interpreter) and transcribed in real time across 35+ supported languages (with automatic cross-language translation into English or French clinical notes).
- Clinical Entity Extraction and Structuring: Specialized clinical language models filter out conversational small talk and extract medically pertinent facts—chief complaints, history of present illness (HPI) chronology, review of systems, physical examination findings, pertinent negatives, assessment, and treatment plans—formatting them into a structured SOAP note, History and Physical (H&P), consultation Nabla letter, or specialty-specific template.
- Clinician-in-the-Loop Verification and Coding Assist: Before any text enters the legal medical record, the attending clinician reviews the drafted note alongside suggested ICD-10-CM diagnosis codes, CPT procedure codes, and plain-language patient instructions, editing or overriding any section prior to final EHR attestation.
How Nabla Integrates with EHR Systems and Nabla Connect
In enterprise health systems and digital health platforms, ambient scribes are evaluated on how seamlessly they exchange structured data without forcing clinicians to manually copy and paste blocks of text. Nabla supports three integration patterns:
| Integration Pattern | Technical Mechanism | Supported Systems & Data Flow |
|---|---|---|
| Bi-Directional Enterprise EHR Integration | HL7 FHIR Release 4 (R4) APIs and SMART on FHIR launch context | Integrates with Epic (via Epic Showroom / Haiku / Canto / Hyperspace), Oracle Health (Cerner), Athenahealth, NextGen, and eClinicalWorks. Reads patient demographics, problem lists, and active medications; writes signed note sections, discrete vitals, and ICD-10 problems back to the encounter chart. |
| Nabla Connect (Embedded SDK / API) | JavaScript Web SDK and REST/Webhook API for custom EHRs and telehealth platforms | Allows programmable EHRs (such as Canvas Medical) and proprietary virtual-care platforms to embed Nabla’s ambient recording and note-generation UI natively inside their own clinician workspace. |
| Standalone Web & Browser Extension | Chrome extension and standalone desktop/mobile web app | Provides one-click structured section transfer and smart-phrase export for clinics operating legacy or web-based EHRs without custom API provisioning. |
Security, HIPAA Compliance, and Zero-Audio-Retention Architecture
When institutional privacy officers evaluate using AI with Protected Health Information (PHI) under a HIPAA BAA, audio recording storage and model training policies are the primary risk vectors. Nabla’s security architecture enforces four enterprise controls:
- Regulatory Certifications and BAA Execution: Nabla operates as a HIPAA Business Associate (executing formal Business Associate Agreements with covered entities), maintains independent SOC 2 Type II attestation, holds ISO/IEC 27001 certification for its Information Security Management System, and complies with the EU General Data Protection Regulation (GDPR).
- Zero Persistent Audio Storage: Encounter audio is streamed over TLS 1.2+ encryption and processed transiently in memory to generate the transcript and structured note; raw audio recordings are never stored persistently on disk by default and are immediately discarded once transcription finishes.
- No Foundation Model Training on Customer PHI: Patient encounters, transcripts, and finalized clinical notes from enterprise healthcare customers are contractually excluded from training or fine-tuning external foundation models.
- Configurable Transcript Retention and Access Control: Health systems can configure automated purge windows for generated text drafts (for example, automatically deleting temporary notes after EHR export or after a defined 1-to-30-day review window) alongside SAML 2.0 / OIDC Single Sign-On (SSO) and role-based access control.
Evaluating Ambient AI Scribes in Clinical Informatics Research
Health services researchers studying ambient documentation tools like Nabla Copilot measure outcomes across three standardized informatics endpoints:
- EHR Active Time Metrics: Reduction in “pajama time” (after-hours EHR documentation outside scheduled clinic hours) and total time spent in Notes per 8 hours of scheduled patient visits, extracted from EHR audit logs (such as Epic Signal).
- Documentation Quality (PDQI-9): Blinded evaluation of ambient-generated notes using the validated 9-item Physician Documentation Quality Instrument (PDQI-9)—scoring notes on whether they are up-to-date, accurate, thorough, useful, organized, comprehensible, succinct, synthesized, and internally consistent—plus explicit tracking of omission rates and unsupported hallucinations.
- Clinician Cognitive Load and Burnout: Pre- and post-deployment changes measured via the NASA Task Load Index (NASA-TLX), the Stanford Professional Fulfillment Index (PFI), or the Maslach Burnout Inventory (MBI).
