AI-Powered Patient Intake for Meridian Health
How Shujal cut manual document processing by 68% with a RAG-based intake pipeline for a regional healthcare provider.
- Manual processing reduced
- 68%
- Avg. intake time
- 9 min → 2 min
- Documents/day automated
- 3,400+
The challenge
Meridian Health processed thousands of referral and intake documents every day — faxes, PDFs and scanned forms — by hand. Three full-time analysts spent their days reading documents, extracting fields and routing cases. The backlog delayed patient care and the work was error-prone.
What we built
We designed an AI document-processing pipeline that ingests incoming documents, classifies them, extracts structured fields and routes them to the right team. A Retrieval-Augmented Generation layer lets staff ask natural-language questions against a patient's document history with cited sources.
- Ingestion & OCR for faxes, PDFs and scans
- Classification & extraction with an LLM grounded in Meridian's own templates
- Human-in-the-loop review for low-confidence cases
- RAG search over patient document history with access controls
# Simplified extraction step with confidence-based routing
result = extractor.run(document)
if result.confidence < THRESHOLD:
route_to_human_review(result)
else:
commit_to_ehr(result)
The outcome
Within ten weeks the pipeline was handling over 3,400 documents a day. Manual processing dropped 68%, average intake time fell from nine minutes to two, and the analysts moved to higher-value clinical coordination work.
Shujal embedded with our team and shipped in ten weeks. It now handles work that used to take three full-time analysts. — VP of Operations, Meridian Health