Clinical Documentation Copilot
Ambient AI scribe that drafts structured notes, ICD-10 codes, and prior-auth letters in real time.
- Client
- Northwind Health
- Year
- 2025
- Duration
- 14 weeks
- Team
- 4 engineers, 1 designer, 1 clinical SME
Northwind Health runs 31 ambulatory clinics where clinicians were spending close to two hours on documentation for every hour of patient care. We built an ambient copilot that listens to the encounter, drafts a structured note against Northwind's own templates, proposes ICD-10 codes with supporting chart evidence, and assembles prior-authorization packets — all inside Epic, with nothing auto-filed without a clinician signature.
The brief
Reduce documentation burden without introducing a second system, a second login, or any pathway for PHI to leave the compliance boundary.
- Healthcare
- LLM
- HIPAA
The decisions that mattered.
Every engagement turns on a handful of calls. These are the ones that decided whether this system reached production.
Deployed inside the EHR, not beside it
The entire experience renders in an Epic SMART on FHIR panel. Clinicians never leave the chart, which is why adoption held above 80% past the six-month mark instead of collapsing after the pilot.
Nothing files without a signature
Every artifact is a draft. Confidence thresholds route uncertain extractions into a review queue, and the sign-off action is a single keystroke from the note body.
Payer rules as versioned configuration
Prior-auth requirements differ per payer and change often. We modeled them as versioned config rather than code, so the revenue-cycle team updates rules without a release.
PHI never left the boundary
Inference runs inside Northwind's VPC under a signed BAA. Trace capture redacts PHI before storage, and the audit log is exportable for their annual HITRUST assessment.
What shipped, and what changed.
Outcomes
- Documentation time per encounter down from 16 minutes to 5
- 14 clinician hours reclaimed per week, per provider
- Prior-auth cycle time reduced 71%
- Coding accuracy up 12 points against the internal audit sample
“The first tool we've deployed where clinicians asked us to expand it faster, not slow it down. That has never happened here before.”
What we built it with.
- Claude
- Python
- FHIR
- Postgres
- AWS
- Next.js
Other projects from the shelf.
Equity Research Terminal
Multi-modal RAG over 10 years of filings, transcripts, and broker research with citation-grounded answers.
Read morePredictive Maintenance Platform
Time-series + vision models predicting equipment failure 72 hours in advance across 14 plants.
Read moreAI Concierge for Retail
End-to-end conversational agent resolving 62% of post-purchase requests without human escalation.
Read moreTell us about the problem.
A 30-minute call with the senior team. We will tell you what is realistic, what it would cost, and whether we are the right people for it.