The Problem: 「Language Matching Takes 5 Minutes Per Email」
A small team running a B2B business receives 20+ buyer inquiries daily. Korean traders, American importers, Chinese retailers. Everyone writes in different languages.
Here's what used to happen:
1. Open the email
2. Notice it's in Mandarin
3. Navigate to the Google Drive 「Mandarin Templates」folder
4. Find the matching topic template
5. Copy, paste, customize
For 20 emails at 5 minutes each, that's 100 minutes gone. A quarter of the entire workday.
The Solution: Claude Code Identifies Language in 3 Seconds, Auto-Selects Template
Using Claude Code on a Windows environment, we built a lightweight automation:
The Flow
New email arrives
↓
Claude Code reads email body
↓
Language detection (Korean / English / Mandarin)
↓
Search matching language folder for topic-specific template
↓
Auto-select template + auto-fill basic info
↓
Upload draft to Google Sheet 「Buyer Response Queue」
Before/After
| Metric | Before | After |
|--------|--------|-------|
| **Time per Email** | 5 minutes | 20 seconds |
| **Language Mistakes** | 2~3 per month (Korean reply to English) | 0 |
| **Daily Capacity** | 15 emails (bottlenecked) | 50 emails (headroom) |
| **Template Not Found** | 1~2 times per week | Almost never |
Three Surprising Lessons
1. Grammar Patterns Beat Character Sets for Language Detection
We initially thought 「Chinese = Traditional Characters」, but buyers often mix languages due to time pressure. Claude Code judges by sentence structure, punctuation, special symbols (like 「」), making errors nearly impossible.
2. Same Language, Different Tones
English emails vary widely: 「relaxed importer's warm tone」vs. 「urgent wholesaler's blunt tone」. Right now we use a baseline template, but we're planning tone-aware template selection next.
3. Korean Honorifics Matter More Than Expected
A friendly tone perfect for US buyers felt off to Korean counterparts. Feedback led us to auto-select 「formal + professional」templates for Korean responses.
Next Experiments
• Auto-select templates by buyer country and industry
• Learn past transaction history to generate personalized greetings
• Auto-classify inquiries (inventory / sample request / complaint) and prioritize