Actor supports a minimal training to teach how should emails be considered. This option takes priority over Custom Filters and Default Labelling.
How it works:
- currently Actor users the rules you set to create a very small unique model for your account (not this is applied per account, so if you have multiple accounts, you need to recreate the rules)
- next step will be for Actor to detect rules based on your behaviour.
- rules are like a text based notebook, with a specific structure/format
What is an AI Rule?
A rule is like an instruction you give Actor Assistant how to filter emails to do what you want.
When you see an email from sales@example.com move it to Promotion folder.
In order for the system to understand and learn from you, the rule has a specific structure and format.
- Start each rule with “## Rule: <short name>”
- Describe conditions under “When:” (sender, subject, body, attachments, time, etc.)
- Add action “Then label:” and an account label (check Account Data tab for your labels)
- Add action “Then move:” and only of: Inbox, Archive, Trash, Spam or other folders
- Active: true or false (you can easily deactivate a rule, without deleting or commenting all lines)
- (Optional) Examples: bullet list of email subjects or snippets
IMPORTANT:
- Lines starting with # are comments and ignored.
- Blank lines are ignored.
- You can have multiple labels/actions by repeating “Then …” lines.
- Match: any|all (default any) to combine multiple When conditions
- If overlapping actions (Then), the last one applies.
Example AI Rules
## Rule: Invoices to Finance
When: sender contains invoices@ or domain contains invoices.com; subject contains invoice OR receipt
Then label: Finance
Active: yes
Examples:
– Subject: “Invoice MEN-2025-0041”
– Subject: “Your monthly receipt”
Some advice on conditions
- better to write multiple When lines and combine them with Match (any or all) instead of having a very long When line.
Because this section is going to be a long-term feature of Actor, you can either contact us on how to create rules and suggest ideas/feedback.