Clinical documentation support that clinicians actually trust
Structured note drafting deployed entirely inside the hospital VPC, cutting documentation time 41% with every output clinician-reviewed.
Where they started
Clinicians at a multi-site provider were spending close to two hours a day on documentation. Previous AI pilots had been rejected by the clinical safety board because outputs could not be traced to source and no PHI could leave the network.
What we did
- 1
Deployed open-weight models on hospital-controlled GPU infrastructure so no PHI left the VPC.
- 2
Built retrieval over internal protocols and the patient's own record, with every generated statement linked to its source.
- 3
Designed the interface around review rather than acceptance: nothing is committed without explicit clinician sign-off.
- 4
Ran a six-week shadow deployment where output was generated but never shown, and scored against clinician-written notes.
- 5
Established an evaluation harness tracking factual accuracy and omission rate per release.
The outcome
Approved by the clinical safety board and rolled out across three sites. Documentation time fell 41%. No PHI leaves the hospital network.
“It is the first tool we have trialled where the safety board's questions had straightforward answers.”
Other case studies
Cutting support volume by half with grounded retrieval
A national operator's support assistant answered 19% of queries correctly. Rebuilding the retrieval layer took it to 54% without changing the model.
FintechAutomating KYC review without losing the audit trail
Document extraction and verification cut manual KYC review 62% while making every decision reproducible for regulators.
EcommerceSemantic search across a 240,000-SKU catalogue
Replacing keyword search with hybrid semantic retrieval lifted search conversion 23% and cut zero-result searches by four fifths.
Let's talk about what you're building
Tell us the problem. We'll tell you honestly whether AI is the right tool, and what it would take.