Before(Previously): Manual Email Sorting Every Day
There was a ritual that started every morning. Open the mailbox, scan through 100+ emails, and mentally sort them into "Urgent", "Pending", and "Reference" categories. Repeated inquiries from the same buyer get one line, new trading partner proposals go into another notebook, shipping notices get pasted elsewhere.
The same content copied multiple times consumed 30 minutes to 1 hour each day.
Foreign buyer emails were especially challenging because their subjects often lacked clarity, requiring you to read the entire content to determine priority. Over time, important messages occasionally got missed.
After(Now): Bots Handle the Sorting
Step 1: Email Collection Automation
The first transformation came through Gmail API automatic retrieval. A simple request to Claude Code:
Requirements:
• Auto-collect unread Gmail emails from past 24 hours
• Extract sender address, subject, body preview
• Record to Google Sheets in real-time
The basic framework was complete in 3 days. No more manual email copying.
Step 2: Smart Classification Rules
Second was the classification logic. Claude Code learned buyer email patterns.
| Category | Criteria |
|----------|----------|
| "Urgent" | Contains keywords like order confirmation, quantity change, shipping issue |
| "Under Review" | New product inquiry, price negotiation, sample request |
| "Info" | Business courtesy, catalog request, general inquiry |
The bot now scans subject and body keywords, auto-assigning categories.
Step 3: Priority Scoring
Beyond simple classification, it began auto-assigning priority scores. Rules:
• Order modification from existing buyer: 9 points
• Large inquiry from new buyer: 8 points
• Repeated price negotiation: 6 points
• General greeting: 2 points
Results
Before: 30 minutes manual work daily
After: Auto-completed by 7 AM, just review the priority-sorted sheet
The most surprising part was when the bot started remembering specific buyer "response patterns." Trading partners who always inquire Monday morning, buyers expecting 2-day responses. All automatically detected and logged.
Lessons From Small Failures
Initially, all emails received equal weight, causing the bot to miss duplicate subjects. Adding the condition "reference past 72 hours from same sender" to Claude Code improved duplicate detection to 90%.
Current Workflow
1. Buyer Email → Gmail (auto-collect)
2. Apply classification rules → Calculate priority score (Claude Code)
3. Sort in Google Sheets (real-time)
4. Filter removed, summary report generated at 9 AM
The inbox is no longer a stress source. Important messages appear automatically at the top, remaining tasks are handled by priority sequence.