Before: The Manual Grind
Every Monday morning, I spent 30 minutes at my Windows desktop doing the same repetitive task.
Pasting newly collected buyer contacts into a Google Sheet, cross-referencing with the existing list, manually flagging duplicates, and filtering out malformed email addresses.
Wading through 500+ rows each time:
• Same email address but different company names (typo? merger?)
• Entries with valid format but suspicious domains
• Rows packed with empty fields
Doing this week after week, focus would wane, and I'd miss 1 or 2 contacts every month.
After: The Bot's Precision Touch
I gave Claude Code simple instructions:
Read columns A (email) and B (company name) from the sheet:
1. Remove rows where emails are exactly identical
2. Flag rows with empty email or missing @ symbol
3. Separate entries where domain is not .com, .co.kr, etc. into a separate list
4. Organize results into a new sheet
The bot finished processing in 4 hours. The results were striking.
3 Patterns Discovered
1. Identical Email, Different Company Names (47 cases)
At first I thought it was an error, but upon review:
• Multiple regional distributors
• Parent company and subsidiaries
• Shared email addresses (family-run businesses)
Some of these turned out to be the same buyer. Manual work would have missed this entirely.
2. Suspicious Domains (23 cases)
Personal Gmail and Yahoo accounts were mixed in. Separating them allowed the sales team to adjust contact priority.
3. Blank Fields with Real Meaning (12 cases)
Rows with an email but completely blank company name. These weren't simple omissions, they were unverifiable contacts from our perspective.
Post-Automation Changes
| Metric | Before | After |
|--------|--------|-------|
| Processing time | 30 min/week | 4 hours/month (once) |
| Missed duplicates | 1~2/month | 0 |
| Validation method | Manual review | Rule-based auto-filter |
| Suspicious item classification | Manual | Auto-separated |
Tip: The Bot's Limitations
Of course, it isn't perfect.
The domain validator sometimes flagged legitimate startup domains as "suspicious." Manual verification is still needed for edge cases.
Now the workflow is: bot categorizes, team spot-checks quickly. Takes 5 minutes instead of 30.
---
Pro tip: The more explicit your bot's rules, the higher the result's reliability. Instead of vague instructions like "remove duplicates," specify: "delete if emails are identical, but flag separately if only the domain differs."