The Hidden Risk in Organizing 3,500 Business Partner Records
One afternoon, while organizing Google Sheets data on a Windows device, something unexpected appeared. Beyond simple name sorting, the same company was stored in multiple formats.
「ABC Trading」, 「abc trading」, 「ABC TRADING」, 「Abc Trading Limited」
Identical company, entirely different entry methods.
Before: The Manual Verification Era
• Comparing rows one by one in Excel
• Simultaneously reviewing company name, email, phone across multiple fields
• Taking approximately 20 hours, with about 15% error probability
• Duplicate entries discovered repeatedly even after initial sorting
After: Claude Code Agent Steps In
A simple prompt entered on the Windows device.
Review the Google Sheets business partner list:
1. Detect company name case variations
2. Compare email address similarities (typos, outdated addresses)
3. Standardize phone number formats and check for duplicates
4. Organize discovered duplicates into a separate sheet
The results were remarkable.
Duplicates Found: 247 records
• Exact duplicates: 89 records
• Format-variant duplicates: 158 records
• Items missed by agent: 0 records
The Unexpected Discovery
The agent went beyond simple deduplication and found patterns.
• Same contact person registered with multiple email addresses
• Businesses where old and new company names were mixed together
• Similar phone numbers registered as different departments
These subtle differences were nearly impossible for humans to spot while manually reviewing 3,500 entries.
Real Business Impact
After deduplication, email automation was executed. Previously, the same person received multiple messages. Now, contact reaches them exactly once.
「Thanks to the agent's sorting, our email bounce rate dropped from 3% to 0.2%. Business partner response speed also improved.」
Why Automation Cannot Completely Replace Human Work
Here's where it gets interesting. After the agent identified duplicates, humans still made the final decisions.
• Some flagged duplicates actually represent different business partners
• Certain older information needed preservation
• Merge versus separate decisions required business context
Non-Developer Team's Insight: Agents execute rules perfectly, but experience and judgment remain fundamentally human territory.
The Road Ahead
Automatic duplicate checks now run whenever new business partners enter the system. A single simple command each week maintains data quality on the Windows device.
Automation eliminates repetitive tasks; humans focus on judgment. That's the small transformation our team experienced.