Case studyLive in production

Operational intelligence from unstructured data

Agents that read claims files, reports and emails, then turn them into faster claims and new warranty revenue.

Visual for operational intelligence from unstructured dataAn automotive insurance company in the GCC
Problem statement

What the client was up against

Most of the insurer's operational knowledge sat in claims PDFs, police reports, repair estimates, technician notes and WhatsApp threads. BI tools only saw the structured 40%. Claims took weeks and warranty leads went unnoticed.

Our approach

How we approached it

  1. Mapped the claims journey and every document that touches it
  2. Combined unstructured files with ERP, MES and KPI data
  3. Added cross check agents so every finding is validated before use
  4. Rolled out in shadow mode, then human in the loop, then auto approval
Solution we created

What we built

An EXTRCT multi agent system that reads every claims document, benchmarks repair estimates, finds root causes and spots warranty opportunities. Each output links back to the source evidence.

Agents in the system

The multi agent team behind it

SourcesClaims system, ERP, document stores, email, WhatsApp
Document agentReads PDFs, scans, photos and handwriting in Arabic and English
Estimate agentBenchmarks repair estimates against history
Root cause agentExplains why claims happen, with sources
Warranty agentSpots customers ready for a warranty offer
Cross check agentValidates other agents before anything goes live
People decideReview, approve and act
Document agentReads PDFs, scans, photos and handwriting in Arabic and English
Estimate agentBenchmarks repair estimates against history
Root cause agentExplains why claims happen, with sources
Warranty agentSpots customers ready for a warranty offer
Cross check agentValidates other agents before anything goes live
Claims systemERPdocument storesemailWhatsApp
Results

What changed in the numbers

70 to 80%less time to process claims
4 to 6xhigher warranty conversions
15 to 25%saved on every payout

Measured in the client's live deployment.

Client names, detailed case studies, demos and references can be shared on request.

Looking for a similar solution?

Tell us your version of this problem. We will walk you through what we built and what a first pilot would look like for you.

Let's talk