Smart Rules
A smart rule is just a normal rule that uses one or more AI signals as conditions. It lets you automate things keywords can't express — like "is this actually urgent?" or "is this a cold sales pitch?"
Smart rules need AI features enabled. See AI, Explained and AI privacy & consent.
The AI fields you can match on
Every signal MailPrism detects is available as a condition. The values are the same ones listed in Classification signals and the Conditions reference:
| Field | Values |
|---|---|
| AI Category | Urgent · Important · Personal · Work · Financial/Billing · Newsletter · Promotional · Cold Email · Spam · Social · Notification · Transactional · System · Other |
| AI Urgency | High Medium Low |
| AI Sentiment | Positive Neutral Negative |
| AI: Is Automated Email | true / false |
Spam and cold outreach are matched through AI Category: pick Spam or Cold Email. Cold Email comes from the AI's cold-outreach detector.
Building one
It works exactly like any rule — you just pick an AI field as the condition:
- Rules → Create Rule, then choose Rule builder.
- Add a condition and choose an AI field — AI Category, AI Urgency, AI Sentiment, AI: Is Automated Email, and more.
- Add your actions.
- Save and enable.
Examples worth copying
Surface urgent mail
When AI Urgency is High then Star · Mark important · Notify me by email
Quiet the cold pitches
When AI Category is Cold Email AND Sender History is First-time sender then Apply label
Cold outreach· Archive
Catch unhappy customers fast
When AI Sentiment is Negative AND AI Category is Work then Star · Track response needed
Tidy the noise
When AI: Is Automated Email is
trueAND AI Category is Notification then Apply labelNotifications· Mark read · Archive
Make smart rules dependable
- Pair AI with a plain condition. Combining urgency with sender relationship is more reliable than urgency alone.
- Start with safe actions — label, star, notify — before letting a smart rule archive or reply.
- Use confidence where it helps. Response-tracking conditions expose an AI
confidence value (
0.0–1.0) so you can require the AI to be sure. You can also raise the global AI confidence threshold in Settings → Rule Defaults & Safety so AI conditions only fire when the model is sure enough. See Conditions reference. - Tune the signals themselves. If a category or urgency call feels consistently off, adjust how MailPrism reads your mail in Preferences & tuning.
After enabling a smart rule, review the Rule Logs to see what it matched — the log captures the AI's reasoning, so you can tell why it fired.
→ Next: Preferences & tuning