Case Study 04
Erixa: WhatsApp commerce with receipt-forensics checks.
A per-client isolated commerce bot connecting WhatsApp conversations to Shopify orders, OCR, screenshot fraud checks, and Claude Vision review.
The brief
Commerce automation needs isolation, integrations, and fraud awareness.
Each client needed its own WhatsApp commerce behavior, order flow, and data boundary. The system also needed to inspect payment screenshots instead of trusting manual review.
The hard part was joining commerce state, WhatsApp conversation state, Shopify orders, image forensics, OCR, and vision-model judgment without making the bot brittle.
Tenant
per-client bot behavior, data scope, and commerce rules
Forensics
ELA, TruFor, CAT-Net, GOT-OCR2.0, and Claude Vision checks
Shopify
order sync from WhatsApp conversation actions
Isolation
Per-client commerce brain
Client behavior, WhatsApp rules, order settings, and data scope stayed isolated so one bot configuration could not leak into another client.
Integration
WhatsApp to Shopify state
Customer conversation actions had to map cleanly into order creation, merchant review, payment state, and follow-up messaging.
Forensics
Receipt fraud pipeline
ELA, TruFor, CAT-Net, OCR, and Claude Vision signals were combined to flag forged or suspicious payment screenshots before fulfillment.
Stack