Discovery That Started From a Simple Sort Request
Last Monday, I decided to clean up the inventory status spreadsheet that had been accumulating on my Windows device. Item codes, quantities, receiving dates, and allocations by buyer were all mixed together. I wanted to sort the data by SKU order, which should have been straightforward.
Request: 「Can you sort all rows of my Google Sheet by Column A (SKU),
and also flag any duplicate SKUs you find?」
An Unexpected Result
The bot came back 2 minutes later with the sorted spreadsheet and a note:
「Duplicate SKUs found: 12 items are registered 2 or more times. For example: SKU-4521 appears twice (once from March receiving, once from April), and SKU-7834 appears exactly twice with identical quantities.」
This wasn't just sorting, it was data validation.
Before vs. After
Before
• Manual sorting: 15 minutes required
• Finding duplicates: Manual scanning caught only 3～4
• Data reliability: Unstable
After
• Bot sort + validation: 2 minutes (fully automated)
• Finding duplicates: All 12 identified
• Data reliability: 100% assured
• Bonus discovery: Found 5 items with same SKU but different receiving dates (potential accounting errors)
Real Impact
Of the duplicates found, 3 were actual receiving errors, while 9 were data entry mistakes. After correcting these, we finally solved the mystery of why inventory totals didn't match expectations.
What was particularly interesting is that the bot didn't just say 「these are duplicates」 and stop. It compared receiving dates and quantities further, distinguishing which duplicates were 「true errors」 versus 「timing differences」.
Key Takeaway
Adding even simple validation logic to automation tools can catch massive data errors that manual work typically misses. What's truly remarkable is that a non-developer can build this level of validation through Claude Code without any programming knowledge.