Workers' Comp Automation

Outcome

5-stage pipeline: Intake → AI Extraction → Normalization → HITL Review → PDF

• 50-state DWC form mapping with state-specific field requirements
• Human-in-the-loop verification — AI suggests, human confirms
• State-compliant PDF generation with employer and employee data pre-fill
• Built for All Lines Claims Consultants (ALCC)

The Problem

Workers' comp wage statement preparation is labor-intensive, error-prone, and varies by state. Adjusters manually key data from pay stubs and employer statements into state-specific forms, with each state having different fields, calculation rules, and formatting requirements.

The Architecture

Built a 5-stage pipeline: document intake and classification, AI extraction with structured output, normalization to standard schema, human-in-the-loop review with side-by-side verification UI, and state-compliant PDF generation from 50-state DWC form templates.

The Solution

Adjusters upload wage documents, the AI extracts and normalizes earnings data, a human reviewer verifies accuracy in a split-pane UI, and the system generates the correct state-specific DWC form PDF. Reduces per-statement processing from 20+ minutes to under 5 with verification.

Stack & Role