I Built an AI Reconciliation System 3 Ways — Here's What Actually Works

Every week I spent 3–4 hours manually matching e-commerce orders with payment records. Now it takes 5 minutes.

In this video I walk through 3 real approaches I built — what worked, what failed, and the expensive lessons I learned along the way.

── APPROACHES COVERED ──
00:43 Approach 1: Custom Web App (Vue.js + Fastify)
02:02 Approach 2: n8n Visual Automation
03:27 Approach 3: Claude Code + MCP (what I still use today)
05:47 3 Mistakes That Cost Me Time & Money
07:56 The Stack & Honest Takeaway

── STACK USED ──
• Claude Code (MCP orchestrator)
• Fastify MCP Server (Node.js ESM)
• Supabase (PostgreSQL + Storage)
• Gmail API (OAuth 2.0)
• Python + openpyxl (reconciliation logic)

── IF YOU'RE A BUSINESS OWNER ──
The same mistakes will cost you more — not just money, but weeks of debugging.
If you need a production-ready automation system built right from day one, I'm available on Upwork:
👉 [your Upwork profile link]

Видео I Built an AI Reconciliation System 3 Ways — Here's What Actually Works канала Jay C