Taking an ambient AI scribe from pilot to 40 clinics
A primary care network had an ambient AI note-taking pilot that clinicians liked but leadership couldn't approve for rollout. CloudXChange designed the evaluation, privacy, and EHR integration work needed to scale it.
- Client
- A multi-site primary care network
- Industry
- Healthcare
- Engagement
- Monthly retainer
- Duration
- 9 months
The challenge
The pilot ran in three clinics using a vendor tool with no integration into the network's EHR. Physicians copied notes by hand, and nobody could say how accurate the drafts were.
Leadership had three blocking questions before any wider rollout: is it accurate enough, is patient data handled in line with PHIPA, and will it fit into how clinicians already work?
The approach
Measured quality before scaling
Built a review workflow where clinicians rated a sample of AI drafts each week, producing an accuracy baseline and a way to catch regressions after model updates.
Redesigned the data flow for privacy
Mapped where audio and transcripts went, removed unnecessary retention, added audit logging, and produced documentation for the privacy impact assessment.
Integrated with the EHR
Led the FHIR-based integration so drafts appeared directly in the encounter for physician review and sign-off.
Results
- clinics live
- 40
- documentation time saved per clinician per day
- 2 hrs
- privacy findings at review
- 0
The rollout was approved after the privacy review, and the network's own team now runs the evaluation workflow monthly without outside help.
“We went from a promising pilot nobody could sign off on to a system our physicians and our privacy officer both trust.”
Technology
- FHIR R4
- Speech-to-text
- Large language models
- Azure
- Audit logging
Have a technical decision you'd like a second opinion on?
Book a free 30-minute call. You'll leave with a clearer view of the problem, whether or not we work together.