How Classification Works
The AI reads sender, subject, and body, then assigns the most relevant label using:- Sender identity and your relationship
- Email content and whether it requires action
- Label descriptions you define
- Training examples you provide
- Feedback history from past corrections
Default Labels
| Label | Catches | Default On | Notifications |
|---|---|---|---|
| Important | Emails requiring response or action | Yes | On |
| Calendar | Meeting invites and calendar notifications | Yes | On |
| Billing | Payments, invoices, receipts | Yes | Off |
| Newsletter | Informational newsletters | Yes | Off |
| Hiring | Recruitment communications | No | Off |
| Investors | VC and investor communications | No | Off |
| To Respond | Emails needing personal reply | No | Off |
| To Delegate | Emails for someone else on your team | No | Off |
| FYI | Informational, no response needed | No | Off |
| Other | Everything else | Yes | Off |
Creating a Custom Label
Name and Describe
Give it a name and write a specific description: “Emails from current paying customers about product issues or feature requests. Excludes sales prospects.”
Training Examples
The most powerful accuracy tool. In Settings > Labels, click a label and add matched examples (emails that belong) and unmatched examples (emails that don’t). A few well-chosen examples outperform long descriptions.The Feedback Loop
Every time you move an email between labels, the AI treats it as a correction. Drag the email to the correct tab, or pressl/v to reassign. After a few corrections, the same misclassification rarely recurs.
Per-Label Settings
| Setting | What It Does |
|---|---|
| Notifications | Push notification on new email in this label |
| Auto-archive | Auto-archive emails (useful for receipts) |
| Description | AI classification criteria |
| Training examples | Specific emails that should/shouldn’t match |
AI Memories
Teach the AI about your preferences and contacts.
Why Primary Matters
How using Slashy full-time accelerates AI learning.