The Problem
A trading company's AI dream team of bots was monitoring dozens of buyer emails daily. After a few days, something strange emerged.
Emails were arriving, but responses weren't being prepared. Buyers complained about missing replies. Investigation revealed that legitimate business emails were being caught by spam filters.
Before
• Missing 5～7 critical buyer emails per week on average
• Buyer reply rate around 15～20%
• Normal emails sometimes flagged as spam, requiring manual recovery
• Bots attempting to process deleted emails caused errors to spike
• Wasted time on false alerts
Root cause analysis showed that incomplete sender information and inadequate domain verification were common culprits.
The Solution: Three-Line Validation Rules
We created a simple email pre-filter using Claude Code.
# Sender address validation rules
1. Check if domain is a legitimate corporate domain (@company.com format)
2. Scan email body for company name, product mention, specific inquiry content
3. Filter attachments to ensure standard business document formats
These three checks alone reduced false positives by 95%.
Measurable Results
After
• Buyer email miss rate: 1～2 per week (essentially eliminated)
• Buyer reply rate: 45～50% (tripled)
• Spam vs. legitimate email accuracy: 98%
• Manual intervention needed: 1～2 cases per month
The breakthrough was letting the bot automatically signal "this email is trustworthy," then prioritize accordingly.
The Small Lesson
We initially wanted to build "the perfect filter," overcomplicating the design. We even considered domain reputation scoring, DKIM/SPF verification, and machine learning models.
But what we actually needed was three simple lines of logic. Higher complexity breeds harder maintenance and more bot-to-bot conflicts.
Email filtering taught our small team that quick iteration beats perfection every time.