Engagement snapshot
- Mandate
- Remove repetitive intake work and increase referral throughput without letting incomplete cases slip forward.
- Timeline
- 10 weeks from intake mapping to production deployment.
- Team shape
- Referral ops manager, 2 automation engineers, and 4 intake coordinators.
The problem
Referral teams retyped faxed and emailed referrals, then manually chased missing authorizations, demographics, and visit prerequisites before scheduling could start.
What we built
Implemented intake capture, completeness checks, missing-item outreach, and scheduling-ready routing with human review for ambiguous referrals.
Operating context
The network had enough referral demand that manual intake became the bottleneck. Schedulers were often waiting on information that intake teams were still collecting, but the handoff points were not explicit.
Key constraints
- Referrals arrived through mixed channels and in inconsistent formats.
- Readiness rules differed by specialty and payer.
- The workflow had to separate clearly complete referrals from truly ambiguous cases.
What we built
Mixed-channel referral capture
Normalized faxed, emailed, and portal referrals into one structured intake workflow.
Readiness validation engine
Checked prerequisite information before a referral entered scheduling-ready status.
Missing-item outreach flow
Triggered requests and reminders for missing information while keeping coordinators in control of the edge cases.
Delivery path
Referral class mapping
Grouped referral types by readiness requirements and missing-data patterns.
Coordinator pilot
Tested the workflow against live intake volume and tuned exception queues for real coordinators.
Specialty expansion
Rolled out across specialties with configurable readiness rules and dashboard reporting.
Why it mattered
The automation improved throughput because referrals were no longer pushed downstream as half-complete work. Intake and scheduling gained a cleaner boundary, which reduced rework and helped more referrals make it to appointment.
Implementation notes
- Referral automation needs readiness logic first and extraction second.
- Mixed-channel intake is manageable once all paths normalize into one state model.
- Schedulers trust the workflow only when ready means truly ready.