Tasks

Task extraction

What alfred_ counts as a task when reading email, and how to tune the rules.

When triage runs, alfred_ scans every email for commitments and asks. Anything that looks like work-to-track gets pulled into your task list automatically — linked back to the source thread.

What counts as a task

alfred_ looks for two signal types:

Commitments you made

Phrases like:

“I’ll send the proposal Tuesday”
“Let me get back to you tomorrow”
“I’ll have it for you by EOD Friday”
“Will follow up next week”

When you write these in a draft (and especially when you send), alfred_ creates a task with the implied deadline.

Asks made of you

Phrases like:

“Can you send me the deck by Friday?”
“Need your sign-off on the proposal”
“Could you review the contract today?”
“Please confirm the budget”

When you receive an email containing a clear ask with a deadline, alfred_ creates a task you can act on.

What doesn’t get extracted

alfred_ skips:

Vague language without a deliverable (“we should chat sometime”)
FYI statements with no action (“just letting you know we shipped”)
Newsletters, automated alerts, transactional email
Anything from a sender on your noise list

Tuning extraction

Settings → Tasks → Extraction. You can:

Disable auto-extraction — capture only via explicit chat asks
Limit to VIPs — only auto-extract from VIP senders
Limit to specific senders / domains — only auto-extract from your top clients
Require explicit deadline — skip extraction when no date is mentioned

Reviewing what was extracted

The Daily Brief surfaces newly extracted tasks under “Tasks alfred_ created from email” — you can review and dismiss any that aren’t real tasks.

If alfred_ extracted something that wasn’t a task, mark it “not a task” — that signal feeds back into your personal model.

Deadlines — how alfred_ infers due dates
Follow-ups — promises tracked in the email itself, not as separate tasks