Platform Overview

How AdjudiCLAIMS Works

From document upload to compliant AI response — every step observable, every output traceable.

Step 1

Document Ingestion Pipeline

Every uploaded file is processed through a four-stage AI pipeline before it's available in the claim workspace.

1
Upload & OCR
Google Document AI

PDF, DOCX, images — all extracted to searchable text via Document AI with layout-aware OCR.

2
Classification
Gemini Flash

12 document categories, 150+ subtypes. Medical reports, legal filings, employer records, DWC forms — automatically labeled.

3
Field Extraction
Claude Sonnet

Key fields pulled: diagnosis codes, WPI percentages, AWE values, dates, body parts, physician names.

4
Embedding & Index
Voyage Large 4

Full-text and semantic vectors stored in pgvector. Documents instantly searchable by content and meaning.

Step 2

The 12-Tab Claim Workspace

Every claim opens into a structured workspace. Each tab surfaces a different lens on the claim — all driven by the documents uploaded.

Tab 01
Overview
Claim summary, key dates, parties, injury description, status.
Tab 02
Documents
All uploaded files, classified and searchable. AI-extracted fields surfaced inline.
Tab 03
Deadlines
Full statutory deadline calendar — every missed deadline shows its legal consequence.
Tab 04
Benefits
TD rate calculator, payment schedule, late-payment penalty tracker.
Tab 05
Medical
Medical billing overview, WPI summary, MTUS guideline matching.
Tab 06
Coverage
AOE/COE determination tracking per body part — factual, GREEN zone only.
Tab 07
Investigation
10-item investigation checklist, timestamp tracking, completion status.
Tab 08
Liens
Lien register, payment status, lien resolution tracking.
Tab 09
Chat
UPL-filtered AI assistant. Factual answers, GREEN/YELLOW responses, RED blocked.
Tab 10
Education
Contextual regulatory education. 57 always-present entries. 86 dismissable terms.
Tab 11
Audit Log
Every action logged with timestamp, user, and reason. Immutable for 7 years.
Tab 12
Compliance
Claim-level compliance score, outstanding items, supervisor review queue.
Step 3

The 3-Stage UPL Enforcement Pipeline

Every AI chat interaction passes through three independent enforcement layers before a response is shown.

🔬
Stage 1: Query Classifier
Pre-chat gate

Lightweight LLM call classifies the query as GREEN, YELLOW, or RED before any document retrieval begins. RED queries are blocked immediately.

📋
Stage 2: System Prompt
Role-specific guardrails

UPL-filtered system prompt enforces zone boundaries. Claims examiners receive examiner-specific prompts that prohibit legal analysis, case outcomes, and settlement recommendations.

🛡
Stage 3: Output Validator
Post-generation scan

24+ prohibited language patterns checked against every response. "Should settle for," "you should," "coverage is," and similar legal-adjacent phrases trigger a block even if the classifier passed.

Step 4

Contextual Education — Always Present

Every decision point in the workflow surfaces regulatory context automatically.

Tier 1

Dismissable Basics

86 foundational terms — AWE, WPI, QME, UR, IMR. New examiners see these by default and dismiss permanently once learned. Shown contextually next to each relevant field.

Tier 2

Always-Present Regulatory Core

57 regulatory education entries covering every legally mandated duty. These are NEVER hidden. "LC 4650 requires TD payment within 14 days because failure triggers an automatic 10% penalty." Glass Box foundation.

Workflows

20 Decision Workflows

Step-by-step guides for every major claims decision — from initial three-point contact through QME dispute resolution. Triggered automatically when the examiner reaches that decision point.

Training

4 Mandatory Training Modules

Onboarding → Foundation → Compliance → Advanced. Quarterly refreshers. Monthly compliance reviews. Training completion gated by assessment scores. All tracked in the audit log.

See it with a real claim.

Request a demo and we'll walk your team through a live claim — upload through compliance dashboard.

Request Demo
AdjudiCLAIMS provides factual information only — not legal advice. All substantive claim decisions require licensed attorney involvement for legal issues. Cal. Bus. & Prof. Code § 6125.