GENZAX AI E-commerce Operations & Order Management System

PROBLEM

Selling online is the easy half. The operational half is a long tail of small jobs that all have to happen and none of which are interesting: checking an order is actually complete and valid, spotting the customer who double-submitted at checkout, deciding whether an order looks risky enough to hold, working out where it should be fulfilled from, telling the customer something useful, and then — the part that quietly costs the most — noticing days later that a parcel has stopped moving. Most of that work only gets done when someone remembers, and the failures are invisible until a customer complains.

APPROACH

Split the order lifecycle into two halves and automate both. The first is the intake path: everything from an order arriving to a validated, classified, routed record. The second is the post-dispatch path, which runs on its own schedule and asks a different question — has anything gone wrong since. Two rules shape the build: orders that fall outside normal parameters stop for a person rather than being auto-processed, and the system never invents order, customer, inventory or shipping information it does not have.

SOLUTION

Built in n8n. An order intake webhook (or a manual trigger with sample orders for testing) normalises the order and validates it; anything invalid is written as an invalid record rather than dropped. Existing orders are loaded and checked for duplicates, and a duplicate is recorded as such instead of being processed twice. Valid, unique orders pass into an AI understanding, risk and routing stage that reads the order and customer context, classifies risk, and decides fulfilment routing. An approval gate then splits the flow: orders needing review are held for human approval, the rest are auto-approved — which in this build is explicitly a test-mode path — and an order record is written either way. Two separate schedules run the second half: delivery monitoring detects delays and issues and decides whether escalation is required, and a daily analytics schedule summarises operations. A follow-up engine and order status management close the loop, updating the same order record.

WHAT IT DEMONSTRATES

• A lifecycle split into an event-driven intake path and two scheduled monitoring paths, rather than one linear run
• Invalid and duplicate orders recorded as outcomes instead of being silently discarded
• Risk classification driving a real approval gate, not just a label on a record
• Human approval holding the orders that fall outside normal parameters
• Post-dispatch monitoring — delay and issue detection with a conditional escalation branch
• Test-mode paths marked as such in the workflow itself, so a demo run can never be mistaken for a live one

WHERE THIS STANDS

This is a working automation prototype built by GENZAX as a portfolio project, running on sample order data. No store platform, payment provider, carrier, inventory system or mailbox is connected, and the auto-approval and escalation notice steps are explicitly test-mode paths in the workflow. No real order, customer record or payment has passed through it. There are therefore no clients, no order volumes, no revenue, no delivery statistics, no ROI and no business results to report — and none are claimed anywhere on this page.