Trainable Email Categorization

An AI-powered email management system that allows users to customize and train the AI to categorize emails based on context rather than just keywords.

Key Points:

• Problem:

  • Current email categorization tools like Sanebox filter emails but lack customization options
  • No good email manager software allows users to customize AI prompts and train the system

• Solution:

  • Build an email categorization system that understands content context
  • Allow users to train the AI with custom instructions (e.g., "put emails related to my kids in this folder")
  • Create separate inboxes for different contexts in a user's life

• Current Implementation:

  • Siqi built a Zapier automation that uses GPT-4 to categorize emails into different labels
  • Uses a very long prompt in the automation to guide the AI

• Market Gap:

  • At least 12 companies have attempted this but none have done a "really great job"
  • Missing feature is the ability to train the system to understand personal context

• Potential Expansion:

  • Start with inbox organization
  • Could expand to understand and organize all content about a user's life