The Problem, and One Key Insight
Our small team introduces 15 new products to global buyers every week. The challenge: manually translating Korean product descriptions into English, Simplified Chinese, Traditional Chinese, Japanese, Spanish, French, German, Portuguese, Russian, and Arabic.
That's roughly 60 hours per month on pure repetition. Then one idea struck: "What if Claude Code could take one Korean description and generate 10 language versions at once, each tuned to how local buyers actually talk?"
The Workflow We Built
Step 1: Set Up a Google Sheets Template
We created a simple Google Sheet with these columns:
Product_Name_KO | Korean_Description | English | Traditional_Chinese | Simplified_Chinese | Japanese | Spanish | French | German | Portuguese | Russian | Arabic
We input only the product name and Korean description. Claude Code automatically fills the rest.
Step 2: Write the Claude Code Script
We gave Claude Code this instruction:
"Read the Korean product description. Translate it to each language, but adjust the tone for each market. English buyers want professional detail. Simplified Chinese buyers prefer direct benefit statements. Arabic buyers expect respectful, quality-focused language. Keep each version 40-60 words."
Core script logic:
Loop through each target language.
Make an API call for each one.
Automatically write the result back to the matching cell.
Log all actions in a separate sheet for error tracking.
Step 3: Run It on Your Windows Device
Open Claude Code on Windows, execute the script. Less than 2 minutes later, all 10 versions are complete. Zero coding experience required.
What Surprised Us
Surprise 1: Each Market Has Different "Hot Buttons"
Same product, completely different sales angles by language.
• English buyers: specs, certifications, lead time
• Simplified Chinese buyers: value-for-money, reorder potential
• Arabic buyers: safety, child-safe materials
Claude Code automatically reflected these differences. Our buyer response rate went up 15%.
Surprise 2: Word Count Consistency Matters More Than Expected
Early attempts produced wildly different lengths per language. Adding "40-60 words" to the prompt standardized everything. Consistency was key to looking professional.
Surprise 3: Review Time Was Shockingly Short
Manual translation of 10 languages would take 4 hours, plus 2 hours of review. The AI-generated output needed only 15 minutes of proofreading. The quality was already excellent.
A 3-Step Checklist for Non-Developers
1. Prepare Your Google Sheet
• Create a new sheet
• Add one column for your source language, one for each target language
• Add 2-3 sample rows of data (for testing)
2. Write Your Claude Code Script
• Visit Claude.ai and enable Claude Code
• Say: "Read [cell range], translate each to these languages, write results to [those cells]"
• Copy the generated code
3. Run It on Windows
• Authenticate with your Google account (grants sheet access)
• Execute the script
• Check your sheet for results
How Our Week Changed
Monday morning: write 10 product descriptions in Korean. Before lunch, run Claude Code. By 2pm, every language version is ready. Meanwhile, our team's automation bots handle other tasks (organizing buyer responses, collecting tracking codes).
60 hours per month became 5 hours. We reinvested the 55 freed-up hours into deeper buyer relationships and new product exploration.
That's what true automation value looks like. Not the technology itself, but time given back.