Cutting prior-auth cycle time by 71%
Average cycle time fell from 4.2 days to 1.2 days. Denials dropped 38% in the first quarter post-launch.
- Industry
- Healthcare
- Size
- 31 ambulatory clinics
- Engagement
- 14 weeks
- Team
- 4 engineers, 1 designer, 1 clinical SME
- Cycle time reduction
- 71%Cycle time reduction
- Denial reduction
- 38%Denial reduction
- Hours saved / clinician / week
- 14Hours saved / clinician / week
Manual prior authorization consumed 6+ hours per case and delayed care for thousands of patients.
Prior authorization at Northwind consumed more than six hours of staff time per case and delayed care for thousands of patients each quarter. The work itself was mechanical — pull evidence from the chart, match it against a payer's current criteria, assemble a packet, submit, track. But payer criteria differ by plan and change frequently, so an earlier rules-engine attempt had been abandoned after the maintenance burden exceeded the savings.
The engagement
- Industry
- Healthcare
- Size
- 31 ambulatory clinics
- Engagement
- 14 weeks
- Team
- 4 engineers, 1 designer, 1 clinical SME
We built a HIPAA-compliant agent that drafts letters, attaches evidence, and routes to payers via existing APIs.
Four phases, each with a written definition of done.
Shadow the real process
Three weeks with the revenue-cycle team timing every step of an actual prior-auth case. The documented process and the real one diverged significantly, and the gap was where the volume sat.
Model payer rules as configuration
Rather than encoding criteria in code, we made payer policy versioned configuration owned by the revenue-cycle team — so a criteria change is an update, not a release.
Build inside Epic
The entire experience ships as a SMART on FHIR panel. Evidence assembly, letter drafting, and submission tracking all happen without leaving the chart.
Gate on confidence
Nothing files without a clinician signature, and low-confidence evidence matches route to a human reviewer with the source passages attached.
What changed.
Average cycle time fell from 4.2 days to 1.2 days within the first quarter, and denials dropped 38% — largely because packets now consistently include the evidence payers were rejecting cases for omitting. The revenue-cycle team has updated payer criteria 40+ times without engineering involvement, which is the result that made the system sustainable rather than merely impressive.
- Cycle time reduction
- 71%Cycle time reduction
- Denial reduction
- 38%Denial reduction
- Hours saved / clinician / week
- 14Hours saved / clinician / week
“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
The capabilities behind this engagement.
More of our work in healthcare.
Clinical documentation, triage, prior auth, and HIPAA-compliant copilots. We work inside your EHR — not around it.
Other engagements worth reading.
Grounding an AI research analyst on a decade of data
Analyst throughput doubled on coverage tasks and onboarding time for new hires dropped by half.
Read morePredicting equipment failure 72 hours out
False-positive rate dropped to 9%, downtime reduced 34%, and OEE improved by 6 percentage points.
Read moreLet's talk about what this would look like for you.
Thirty minutes with the senior team. Bring the problem and we will give you an honest read on scope, timeline, and whether we are the right fit.