Fax to AI: Automating Healthcare's Most Stubborn Workflow

The Problem with Healthcare Fax

Inbound faxes arrive as images. Staff print them, read them, manually re-enter data into EHRs, and route them to appropriate clinicians. This takes 5–15 minutes per fax.

The AI Pipeline

Inbound fax (via eFax API) → Azure Document Intelligence (OCR) → GPT-4 for clinical entity extraction → document classification → routing rules engine → EHR integration. The whole pipeline runs in under 90 seconds.

Document Classification at Scale

Healthcare faxes include: referrals, lab results, prior authorisations, prescription refill requests, medical records, and insurance correspondence — each requiring different downstream handling. Current accuracy is 94%.

Human-in-the-Loop Design

We never route fully autonomously for clinical decisions. The AI extracts and classifies; humans confirm and act. Volume capacity per staff member increased 8x.