Turns clinical conversations into signed notes, automatically.
Physician burnout is the healthcare system's most documented unsolved problem. A 2024 AMA survey found that 62% of physicians report burnout symptoms, with administrative burden — specifically documentation — cited as the leading cause. Physicians spend an average of 1.84 hours on EHR documentation for every hour of direct patient care (NEJM Catalyst, 2023 analysis of Epic data).
The problem compounds in primary care and urgent care settings where patient volume is high and appointment times are short. A primary care physician seeing 20 patients per day spends 2.5–3 hours per day on documentation after clinic hours ("pajama time"). That documentation burden is directly linked to turnover: a 2023 Stanford Medicine survey found physicians with high documentation burden were 2.3x more likely to intend to leave their practice within two years.
The note-writing itself is the constraint. Clinicians know what happened in the visit; they are transcribing that knowledge into a structured format that the EHR requires. That transcription is not clinical judgment — it's data entry.
The Clinical Documentation Agent listens to the patient-clinician encounter (with patient consent), transcribes the conversation, and generates a structured SOAP note draft that the clinician reviews, edits, and signs — typically within 60–90 seconds of the encounter ending.
The agent captures the subjective (patient-reported symptoms, history), extracts objective findings mentioned by the clinician, generates an assessment based on the discussion, and drafts a plan section covering orders, referrals, and follow-up instructions. The output is formatted for the organization's EHR templates.
Critically, the clinician always reviews, edits, and signs the note. The agent produces a first draft; the clinician's signature is the attestation. This maintains clinical and legal responsibility with the clinician while eliminating the blank-page problem and the post-clinic documentation marathon.
Audio captured device-side or via browser is sent encrypted to AWS Transcribe Medical, which handles clinical speech, accents, and medical terminology better than general-purpose ASR. A custom clinical NLP layer built in Python segments the transcript into SOAP sections using a model fine-tuned on real encounter data. Anthropic Claude then generates the structured note draft using the transcript plus patient context pulled from the EHR via FHIR R4 APIs, with the full pipeline orchestrated in LangGraph. All compute runs in BAA-covered AWS environments with encryption at rest and in transit, immutable audit logs in PostgreSQL, and audio deleted after transcription — no long-term recording retention.
A primary care practice with 8 physicians sees approximately 160 patients per day. Current documentation time is approximately 2.5 hours per physician per day, largely completed after clinic hours. Physician satisfaction scores on "time for patient care" are consistently low.
After deploying the clinical documentation agent, physicians review and sign ambient-generated notes within 60–90 seconds of each encounter. Documentation is completed before the next patient is seen. Post-clinic documentation time drops to near zero.
These projections are informed by published outcomes from Nuance DAX Copilot deployments (Microsoft/Nuance, 2024 customer outcomes report) and Abridge's published data from health system partners including UPMC and Kaiser Permanente. Actual results vary by specialty, patient acuity, and EHR environment.
| Metric | Before | After |
|---|---|---|
| Time to complete SOAP note after encounter | 5–12 minutes of active typing / dictation per note | 60–90 seconds to review and sign ambient-generated draft |
| Post-clinic documentation hours ("pajama time") | 1.5–3 hours per physician per clinic day | Near zero (documentation completed in-visit) |
| Documentation completed before next patient | Minority of encounters; most deferred | Target: >90% of notes signed same-day, during clinic |
Records the clinical encounter via device microphone or web with patient consent. Handles background noise, multiple speakers, and regional accents without hallucinating unclear segments — uncertain or inaudible sections are flagged for clinician review rather than filled in.
Produces structured SOAP notes aligned to the organization's EHR templates, keeping subjective (patient-reported) and objective (clinician-observed) sections accurate and distinct. The plan section includes medications discussed, orders mentioned, and follow-up instructions — not a summary.
Pushes completed drafts directly into the clinician's EHR for review and signature — no copy-paste. Supports Epic, Cerner/Oracle Health, and Athenahealth via SMART on FHIR or direct API where available.
Note structure and extraction logic are configured per specialty — primary care, psychiatry, and surgical follow-up have meaningfully different documentation requirements. Templates are configurable at the practice and individual clinician level.
Every step is logged: what was captured, what was generated, and what the clinician edited before signing. Audio is deleted after transcription per configurable retention policy, with all infrastructure running under signed BAA agreements.
Anonymized work with hard metrics — NDA-bound, no client names.
41%
Faster Critical Response
91%
Alert Precision
31%
False Alarm Reduction
“The false alarm rate was a genuine patient safety issue — staff were silencing monitors as a coping mechanism. The AI layer changed the signal-to-noise ratio enough that nurses are paying attention to alerts again.”
— ICU Clinical Director, Regional Medical Centre
We custom-build each agent to fit your data, your rules, and your existing systems.
Free 30-min scoping call
Fordel Studios