Beacon Health
An AI second read that catches what the first one misses.
- Client
- Beacon Health
- Sector
- Provider ops · Bengaluru
- Engagement
- 5 weeks
- Type
- Client engagement
The problem
Beacon's billing team manually filed prior authorizations from faxed charts. Incomplete submissions drove a wave of denials, and every denied prior was revenue that aged out before anyone noticed.
What we built
- 01
Intake that reads the chart
Inbound faxes are OCR'd and structured. The agent extracts the procedure, diagnosis, and supporting evidence, then matches it to the payer's published policy reference.
- 02
A HIPAA-scoped tool budget
Every tool the agent can call is sandboxed and logged. PHI never leaves the boundary, and the immutable audit log — not the model — is the source of truth for compliance.
- 03
A human queue for the hard ones
When confidence is low or policy is ambiguous, the agent routes the case to a human with a one-paragraph summary and a link to the evidence. No silent guesses.
How we measured it
Share of priors fully automated, denial rate for incomplete submissions, and revenue per ops FTE — reviewed weekly with Beacon's RCM lead.