How a Regional Health System Cut Referral Intake Time by 45%
An Appian case-management workflow with AI document classification, taken from proof of concept to production across 24 clinics, with a human reviewer on every decision.
Taking AI Referral Intake from Pilot to Production
The client operates 6 hospitals and 24 clinics across the Midwest. Patient referrals arrived by fax, email and portal, and intake staff keyed each one into scheduling and EHR systems by hand.
The Challenge
Referral backlogs delayed patient scheduling by 9 days on average. An earlier AI pilot had worked in a demo but stalled before production, blocked by EHR integration and data quality. Leadership would approve a new attempt only if it started small, kept staff in the loop and showed a measurable result.
Four workstreams, one phased rollout
Proof of Concept
- Scoped one referral type at one clinic, with an agreed success measure
- Delivered a working Appian workflow in 8 weeks
- Reported baseline and result to leadership before any wider commitment
AI Document Classification
- Classified incoming referrals and extracted key fields automatically
- Routed low-confidence cases to a human reviewer by default
- Logged every AI suggestion and every reviewer decision for audit
Appian Case Management and Integration
- Built intake, triage and scheduling hand-off as Appian case workflows
- Integrated with the client's existing EHR and scheduling systems
- Tested integrations end to end with automated QA
Rollout and Embedded Support
- Phased go-live clinic by clinic, with a change-management plan for staff
- Ran weekly accuracy reviews with the intake leads
- Kept the build team on as the embedded support team after launch
Faster Referrals. Staff in Control. AI in Production.
After 7 months, referral intake runs on an Appian workflow with AI classification at every site, and a staff member confirms each decision. Intake time dropped by 45%, and the backlog cleared within 6 weeks of full rollout. The proof of concept that started it cost less than 10% of the full program.
See what AI in production could do for your intake teams
Tell us about your workflow, your systems and your constraints. We'll scope a proof of concept small enough to measure.