Before: Data Classification by Hand
Every morning, the same task repeated itself.
Is this email a real buyer? Spam? Just someone looking for information?
Every day, 50+ incoming messages were read one by one, then manually typed into Excel with classification tags. Three hours was the baseline, plus plenty of errors. Duplicate buyer records showed up repeatedly, and important messages got lost.
After: Claude Code Becomes the "Filter"
The new approach was straightforward.
1. Google Form auto-logs all inquiries to Google Sheets
2. Claude Code auto-analyzes email content (extracts company name, interested products, deal urgency)
3. AI auto-assigns 3-tier priority tags ("Immediate Response", "Follow-up", "Info-Only")
4. Team reviews sorted list once a week
Results:
• Processing time: 3 hours → 10 minutes (1x weekly review)
• Buyer response rate: +35% average
• Duplicate records: 99% eliminated
• Team morale: Infinite improvement (laughing emoji)
Small Failures and Recovery
The first attempt didn't work. Claude Code parsed email text fine, but priority classification fell apart with mixed languages (English + Chinese + Korean).
So we rewrote the prompt.
Extract the following from each inquiry:
1. Company size hints ("100+ employees" → high priority)
2. First-time buyer flag (check specific keywords)
3. Response deadline ("within 3 days", "urgent" → immediate response)
With this specificity, accuracy jumped above 85%.
Dream Team's Small Discovery
Some messages tagged as "Info-Only" turned out to be pre-deal customer education questions. We didn't realize this until later. So we added a monthly review session with the team to audit classification rules.
Automation is a utility tool, not a replacement for judgment. This balance was the heart of the experiment.
What's Next
We're experimenting with a "Deal Cycle Predictor" bot by buyer. Connect past transaction history with current inquiries to spot patterns like "This buyer typically reorders in 6 months."
Non-developer automation doesn't need to be perfect. As long as it works, learns, and improves through repetition, it changes everything.