A Problem That's More Than Just a Problem
Every Friday at 3 PM, I'd sit in front of my Windows device and spend 30 minutes manually updating shipment statuses one by one.
「Order ID 001, Shipped from Korea, Current location: Singapore, Expected arrival: 72 hours」
「Order ID 002, Current location: Malaysia, Delivery delayed...」
Repetitive work. My fingers knew it. The process stayed the same.
A Shift in Thinking
I connected Claude Code with Google Forms.
Before (Pre-Automation):
• Receive shipping company email → Read manually → Type into spreadsheet
• Time investment: 3~4 hours per week
• Error rate: ~15% (location typos, missing dates)
• Can't respond to buyer inquiries instantly
After (Post-Automation):
• Receive shipping company email → Claude Code auto-parses content → Google Apps Script transfers data → Google Sheets auto-populates → Real-time dashboard updates
• Time investment: 0 minutes (fully automated)
• Error rate: <0.5% (system validation added)
• Can share real-time tracking links with buyers
Tech Stack (Non-Developer Friendly)
Shipping Company Email
↓
Claude Code (Interprets Content)
↓
Google Apps Script (Transmits Data)
↓
Google Sheets (Auto-Input & Categorization)
↓
Google Data Studio (Real-Time Dashboard)
The most surprising part was that Claude Code actually "understood" shipment status. It didn't just recognize the word "in transit," it detected delays from email context and automatically flagged them as "alerts."
Real Results
• **Time Saved**: 4 hours/week → 52 hours/quarter
• **Accuracy Improved**: 70% reduction in inventory errors
• **Buyer Satisfaction**: "I've never seen tracking this fast" feedback
• **Cost**: Nearly zero (Google Sheets free, Claude API under $5/month)
How the Dream Team Bots Reacted
The Hamsters team noticed something.
"Wait, the shipment data is already organized?"
Then the Puppies team took that data and built a buyer analytics bot with it. That's next week's story.
What I Learned
1. **Email is never just text.** Claude Code reads context. From a single phrase like "shipping delayed," it detects patterns.
2. **Start automation small.** Don't try to rebuild your entire order system. Start with "the most repetitive 10 minutes."
3. **Dream teams collaborate.** One bot creates automation. Another bot uses that data to build more automation.
At first I wondered if this even counted as "automation." It felt too simple. But after 48 hours, 4 weeks, 3 months, that tiny automation created changes that were anything but small.