AI Clinical Documentation Copilot
Back to AI ServicesHealthTech · AI Copilot · 2025

AI Clinical Documentation Copilot

How we saved physicians 3+ hours a day with ambient AI.

MediAssist Health · 3 Hospitals · India + UAE

70%

Less Manual Charting

50%

Faster Documentation

3

Hospitals Deployed

6 Wks

To Production

The Challenge

MediAssist Health approached Palpx in Q2 2025 with a critical problem: physicians were spending 35–40% of their working day on manual clinical documentation — SOAP notes, discharge summaries, referral letters — all typed by hand into a legacy EMR.

This wasn't just a productivity issue. Burned-out physicians made more documentation errors, patient throughput suffered, and physician retention was declining. The CTO had evaluated 3 off-the-shelf solutions, but none integrated cleanly with Meditech Expanse or met Indian healthcare data residency requirements.

MediAssist clinicians collaborating
MediAssist physicians before the Palpx AI Copilot.

Our Solution

We built a custom AI copilot that operates in ambient listening mode during patient consultations — transcribing, understanding medical context, and auto-generating structured SOAP notes, ICD codes, and referral letters in real time, directly inside the physician's existing EMR workflow.

Architecture

Consultation Audio
   ↓
Whisper STT (medical vocab)
   ↓
Medical NLP Layer
   ↓
GPT-4o (fine-tuned on de-identified consultations)
   ↓
SOAP Note Generator   →   Meditech EMR (HL7 FHIR)
                      →   Physician Review UI
OpenAI WhisperGPT-4o Fine-tunedPython · FastAPIReact NativeAzure (India)HL7 FHIRMeditech API

Process Timeline

  1. Week 1

    Discovery & Compliance Audit

    Mapped physician workflows across 3 hospitals. Identified data residency requirements. Signed BAA and DISHA-compliant NDAs.

  2. Week 2

    Medical AI Architecture

    Designed ambient listening pipeline. Selected Whisper for STT with medical vocab fine-tuning. Defined EMR integration points.

  3. Weeks 3–4

    Core Build & Training

    Built transcription pipeline. Fine-tuned GPT-4o on 50K+ de-identified consultations. Built SOAP note generation engine.

  4. Week 5

    EMR Integration & Clinical Pilot

    Integrated with Meditech Expanse via HL7 FHIR. Ran a 2-week pilot with 8 physicians across 2 departments.

  5. Week 6

    Full Rollout & Training

    Deployed across 3 hospitals. Trained 60 physicians. Stood up monitoring dashboard for model performance.

Results

In the first month, our physicians reclaimed 3 hours a day. Documentation errors dropped 60%. Physician satisfaction hit the highest we've recorded in 5 years.

Dr. Rohit Sharma, CMO, MediAssist Health

70%

Reduction in documentation time

60%

Fewer documentation errors

+42

NPS among physician users

Gallery

Ambient listening

Live status pane visible to clinician.

Auto-generated SOAP

AI-inserted fields highlighted for review.

Review dashboard

Accept / edit / reject for every block.

UI design by Palpx — details obscured for client privacy.

Want a similar AI solution?

Tell us about your idea — we'll respond with a tailored scope.

Start Free AI Consultation