Ito Health · Jamil Global

The problem

Policy PDFs were blocking reliable AI workflows due to inconsistent layouts and variations. Manual review was expensive and slow, and LLMs produced occasional hallucinations.

What was deployed

Vendor

Jamil Global

Products

pdf-parse, Python, OCR Service, Multi-LLM Strategy, JSON Schema Validation, PostgreSQL / JSON Store, Fill Engine

Technique

LLM

Build or buy

Built in-house or to order

Deployment

Custom build

Data used

Rhode Island-focused policy documents, including TANF, SNAP, Medicare, and Medicaid forms (PDFs)

Scale

processing policy forms at scale, across a full benefits document portfolio

What changed

Each row is quoted from the source. Figures we could not find on the page in those words are marked — they are kept, not deleted, so you can judge them.

Difficulties and limits

• Policy PDFs are often inconsistent in layout, with mixed form structures, nested headings, and dense conditionals
• Manual review was expensive and slow
• LLMs were useful for extraction but produced occasional hallucinations when uncertain

• Low-confidence or conflicting annotations are flagged for review, while high-confidence fields flow automatically to the annotation store.

Sources

• jamilglobal.com — vendor self-report
• jamilglobal.com — vendor self-report