The Problem: Same Questions Every Day, Same Answers Every Day
A small B2B business run by a non-developer faced a persistent challenge: repetitive buyer emails. The same questions appeared over and over.
• 「What's the current inventory status?」
• 「When can you ship samples?」
• 「What's the minimum order quantity?」
• 「What payment methods do you accept?」
Over 100 emails arrived daily, with more than half being variations of these four questions. Response time alone consumed 2+ hours per day.
Before: Manual Classification and Copy-Paste
1. Email arrives → Manual reading and comprehension
2. Manual entry into Google Sheets
3. Find and copy corresponding template
4. Compose reply email
5. Verify and send
Time per email: Average 2 minutes
Error rate: Approximately 8% (missed questions or repeated answers)
Stress level: Mental shutdown every afternoon after 3 PM
After: Google Forms + Claude Code Auto-Classification
Stage 1: Restructure Google Forms as an "Answer Repository"
Set up real-time collection of all buyer-submitted Google Forms responses.
Google Forms → Webhook → Google Sheets (Real-time Sync)
Stage 2: Claude Code Automatically Categorizes Questions
A simple script running on Windows environment detects new emails every 5 minutes.
Classification criteria:
• Inventory-related: 「Current stock」, 「Shipping timeline」
• Payment-related: 「Payment methods」, 「Price inquiry」
• Sample-related: 「Sample shipping」, 「Sample costs」
• Other: Questions not fitting the above three categories
Claude Code analyzes each question's keywords and automatically assigns them to corresponding columns in Google Sheets. Priority levels are simultaneously assigned.
• 🔴 Urgent: First-time buyer, clear intent for large order
• 🟡 Normal: Existing buyer, routine inquiry
• 🟢 Reference: Auction site automated emails, suspected spam
Stage 3: Auto-Generate Reply Drafts
Upon classification completion, response templates are automatically generated.
Question: 「Do you currently have blue size M in stock? I need 100 units.」
Auto-generated draft:
Hello,
Thank you for your inquiry.
We currently have [inventory quantity] units of blue size M available.
Estimated shipping date for 100 units: [auto-calculated date].
Please refer to the following details:
• Unit price: [auto-reflected]
• Sample shipping availability: [Yes/No]
Thank you.
The staff member need only review this draft in 1 second and hit send.
Results: Changes After 3 Weeks
| Metric | Before | After | Improvement |
|--------|--------|-------|-------------|
| Time per email | 2 minutes | 15 seconds | 87.5% reduction |
| Error rate | 8% | 0.3% | 96% decrease |
| Daily working hours | 2 hrs 20 min | 20 min | 85% reduction |
| Avg buyer response time | 4 hours | 12 minutes | 95% reduction |
Unexpected Discoveries
1. **Question patterns vary by season**: When Claude Code analyzed the data, it revealed that buyer inquiry types differ between summer and winter. This insight enables proactive outreach emails ahead of season changes.
2. **New FAQ patterns emerged**: As the automation system aggregated similar questions, previously invisible patterns became clear. New FAQ items like 「Shipping insurance」 and 「Bulk discount conditions」 were identified.
3. **Individual buyer preferences surfaced**: Some buyers prefer chat over email, others want detailed explanations. These preferences naturally emerged during the auto-classification process.
Key Learnings
The true power of automation isn't saving time, it's revealing hidden patterns. Data flows that remained invisible during manual processing suddenly become apparent once automation takes over.
The next phase plans to refine customer management strategy based on these insights.