Automating prior authorisation for a 22-hospital health system
A document AI pipeline cut prior-authorisation turnaround from 4.5 days to under 6 hours, with clinician review preserved.
- Turnaround time
- 4.5d → 5.8h
- Documentation denials
- 19% → 4%
- Reviewer throughput
- 3.1×
- Function attrition
- 31% → 9%
The challenge
Cascade's prior-authorisation team handled 4,300 requests a week across 22 hospitals, each requiring clinical documentation to be located across three record systems and assembled into a payer-specific packet. Turnaround averaged 4.5 days, 19% of requests were denied for incomplete documentation, and the team had 31% annual attrition in a role nobody wanted.
Our solution
We built a retrieval and extraction pipeline that assembles the clinical packet automatically, checked against each payer's current documentation requirements. Critically, we designed it as decision support rather than decision automation: every packet is presented to a clinical reviewer with source citations and confidence flags, and the reviewer approves or amends. We built the offline evaluation suite before selecting a model, and shipped model cards and an audit trail that satisfied the system's compliance committee and external counsel.
The results
Median turnaround fell from 4.5 days to 5.8 hours. Documentation-related denials dropped from 19% to 4%. Reviewer capacity per FTE tripled, and the team redeployed six roles to appeals rather than making redundancies. Attrition in the function fell to 9%. The evaluation harness runs nightly and has caught two payer requirement changes before they caused denials.
Other engagements
- InsuranceCutting claims cycle time by 41% at a regional insurerA 90-year-old claims operation replaced its mainframe workflow without a single day of processing downtime.
- ManufacturingRecovering 6.2 points of gross margin across 14 plantsProcess mining across four ERPs found $31m of recoverable margin that plant-level reporting had been netting out.
Could this be your programme?
We will tell you where the analogy holds and where it does not — before you spend anything.
Or call +971 58 6044 510