Invaritech Claims Recovery Agent

Invaritech Claims Recovery Agent helps Australian grocery suppliers recover money lost through unsupported retailer deductions. A single shipment can create a purchase order, invoice, delivery docket, remittance advice, and promotional agreement. Teams often reconcile these documents by hand. Small deductions are easy to miss. Many go unchallenged because the investigation costs more than the claim. We built a document intelligence workflow that extracts structured fields, connects related documents into one case, and tests each deduction against delivery evidence, agreed terms, and the Food and Grocery Code. Deterministic application logic calculates the recoverable amount. An open model then drafts the claim letter using those verified findings. The dashboard keeps the operator in control. Users can track uploads through processing, inspect document relationships, attach or remove evidence, review each deduction verdict, and manage the resulting claim. The recovery ledger tracks claimed, recovered, and outstanding amounts. The current demo uses Fireworks-hosted open models for classification, extraction, vision OCR, and claim drafting. Document linking, arithmetic, deduction verdicts, and ledger state remain inspectable application logic. The open-model layer can also run behind a self-hosted, OpenAI-compatible endpoint on AMD GPUs. This supports full local document intelligence when privacy, control, or data residency requires it. In the five-document demo case, the system identifies AUD 2,352 in recoverable deductions across an unsupported short-delivery claim, prohibited shrinkage, and promotional funding above the agreed cap.