Monday Morning 8 AM, Inbox Hell
A team leader at a trading firm repeated the same task every Monday morning. With 200 overseas buyer emails accumulated since Saturday, they would open each one and manually record the following items.
• Inquiry Type (Product inquiry, Sample request, Contract renewal, Other)
• Sender Location (Southeast Asia, Middle East, Europe, Africa)
• Response Priority (Urgent, Normal, Defer)
• Attachment presence
This task alone consumed 8 hours weekly.
The Accidental Discovery
One day, while opening Google Sheets on a Windows device, they accidentally opened Claude Code's function window. That moment sparked an idea.
"If Google Sheets' IMPORTRANGE can fetch email headers and body text, could Claude Code read and auto-classify them?"
Before: Manual Classification Process
Check inbox (15 minutes)
↓
Open emails one by one (2 hours 30 minutes)
↓
Read content and select category (3 hours)
↓
Enter into Google Sheets (2 hours)
↓
Review and correct (15 minutes)
Total time: 480 minutes (8 hours)
Error rate: 12-15%
After: Automated Classification System
After copying email text into Google Sheets, they input a simple text prompt to Claude Code.
[Google Sheets - Column A]
Sender email address
Email body text
Send date
[Claude Code Analysis]
Classify this email by:
1. Inquiry type
2. Sender region
3. Priority level
4. Keywords
The results were striking.
Time to classify 200 emails: 480 minutes → 2 minutes
Error rates improved unexpectedly. Manual work showed 12-15% errors, while Claude Code achieved 8%. (Particularly accurate with complex inquiries and multilingual emails.)
Unexpected Bonus Effects
1. **Faster Response Times**: Not only categorizing emails but also automatically suggesting the best team member to respond
2. **Objective Priority Recognition**: Mechanizing urgency scoring made team task order more objective
3. **Data Accumulation**: 200 weekly classified emails became a data resource, revealing buyer regional inquiry trends at a glance
Challenges and Solutions
Issue 1: Claude Code reported "overload" when email bodies were too long
Solution: Added Google Sheets function to limit email body length to 500 characters
=LEFT(A2, 500)
Issue 2: Classification accuracy dropped with emojis and special characters
Solution: Added instruction to Claude Code to "remove special characters before analysis"
Issue 3: Accuracy decreased when Korean, English, and Chinese emails were mixed
Solution: Separated Sheets by language (can consolidate later)
Team Reaction After 48 Hours
The team leader said in Wednesday's meeting: "We no longer spend time on email classification, so we can focus on actual responses." Buyer response time actually dropped from 2 hours to 30 minutes.
What We Learned
Automation isn't built only from sophisticated coding knowledge. It starts by precisely identifying small daily frustrations and experimenting with combinations of existing tools (Google Sheets, Claude Code).
For non-developers especially, this is an advantage. Because unconstrained by developer thinking patterns, you can ask the innocent question: "Is this possible?" And surprisingly often, the answer is yes.