Skip to content
All projects

E-commerce · AI Customer Support · Human-in-the-loop

E-commerce AI Support Desk

Working MVP

A support desk that classifies customer requests, retrieves verified order context, drafts a grounded response, and requires human review before approval.

Support inbox listing incoming requests with status, priority, and intent

Problem

Customer-support teams need faster handling of order-related inquiries without allowing an AI system to invent order details or send unsafe responses automatically.

System Design

A support desk that classifies incoming requests, pulls verified order context read-only, drafts a grounded response, and routes it through an approve / edit-and-approve / reject step before anything reaches a customer.

Workflow Stages

  1. 01Customer request
  2. 02AI classification
  3. 03Shopify / order context lookup
  4. 04Verified context
  5. 05AI draft
  6. 06Human approve / edit / reject
  7. 07Audit log

Tools Used

  • Next.js
  • TypeScript
  • Supabase
  • Shopify integration structure
  • Anthropic/AI integration structure
  • Vercel

Safety & Human Review

  • Human approval before external sending
  • Grounded responses based on retrieved context
  • Audit logging
  • Mock mode for safe testing
  • Secrets kept server-side

Implementation Evidence

Real screenshots of the working system — collapsed by default so this page stays readable.

What's Working

Demonstrates a production-oriented support workflow with human review and traceable actions before any response reaches a customer. No performance numbers are reported because none have been independently verified.

Limitations & Confidentiality

Client kept anonymous unless explicitly approved for disclosure. No live Shopify or customer data is connected in the demonstrated build.

Have a similar workflow to automate?

I'll tell you honestly whether automation is the right fix.