A laptop with stacks of sticky-tabbed paperwork — the messy inputs GRABBER processes.

GRABBER

GRABBER is the AI agentic workflow for capturing fraud, waste and abuse. It monitors claims, payments and related documents to identify risky patterns, anomalies and policy violations. It combines automated data extraction, risk scoring and configurable rules to flag suspicious cases for review while helping standard cases move through faster.

Morning briefing: the claims handler sees which new claims are waiting for a decision and which are running automatically.
All new claims, grouped by status and by intake channel: web portal, email, phone, assistance and fleet.
The execution trace of a claim processed automatically, with every step recorded.
Fields extracted from damage photos, each with a confidence score.
A phone claim: the voice agent's conversation transcript and the fields extracted from it.
Completeness check: missing fields are found and a data request to the customer is drafted.
Duplicate screening: a new claim matched against an earlier one, dimension by dimension.
Geolocation check: the reported accident location conflicts with the captured coordinates.
The handler's decision is recorded and the claim file is created.
A completed claim with its full audit trail.

The prototypes are based on synthetic data.

  • Fraud detection and validation: extracted values are checked against tariffs, policies, business rules, QR codes and other official sources — not just read.
  • Exceptions route to humans automatically; everything else flows straight through.
  • Cuts leakage, and removes the most error-prone manual work.