An AI-powered customer support agent that uses tool-calling to look up orders, answer policy questions, and handle refund requests — built to replace a first-line support rep.
Customers expect instant answers about their orders, but manually checking order status, answering repetitive policy questions, and processing refund requests takes up significant support team time.
This agent uses an AI model that decides which tool to call based on the customer's message — looking up real order data, answering from a knowledge base, or flagging a refund for human approval — instead of following a fixed conversation script.
- 🔍 Real-time order lookup by order number
- 🤖 AI decides which tool to use based on the question (true tool-calling, not a scripted flow)
- 📚 FAQ answers pulled from a knowledge base (RAG)
- 💰 Refund eligibility checks with human-approval gate for sensitive actions
- 💬 Built-in chat interface for easy testing/demo
| Layer | Tool |
|---|---|
| Automation | n8n |
| Database | Supabase (Postgres) |
| AI Model | OpenAI (via n8n AI Agent) |
| Knowledge Base | Supabase Vector Store |
Chat Message → AI Agent Look Up Order (Supabase) FAQ Tool (Vector Store) Refund Tool (Supabase + human approval) The AI Agent reads the customer's message and decides which tool (if any) it needs to call to answer accurately — no fixed script, no manual routing.
✅ Core system complete — order lookup, FAQ answers, and refund handling with human-approval gate all working end-to-end.
- Order lookup tool
- FAQ tool (RAG)
- Refund tool with approval gate
- Full conversation testing
- Add architecture diagram
- Add demo screenshots/GIF
Coming soon — full setup guide will be added as this project progresses.
MIT