Autonomous learning from resolved tickets
When learning is enabled, Atlas extracts Q&A from eligible resolved tickets, auto-approves high-confidence answers, and queues uncertain ones for staff review.
Prefer to watch? These videos show it step by step.
How learning works
With learning enabled, Atlas watches for eligible resolved tickets - conversations that reached a clear answer. From each one it extracts a question-and-answer pair that could help the next member with the same problem.
Auto-approval and the confidence score
With learning enabled, Atlas turns eligible staff answers from closed tickets into reusable Q&A. Add automatically publishes confident entries within daily limits; uncertain entries and entries flagged for possible personal data wait for review. Ask me first holds every new entry for approval. New knowledge bases default to Add automatically; saved Off settings stay off.
- High confidence - approved automatically and marked as auto-approved, so you can always tell them apart from hand-approved entries.
- Uncertain - held as pending until a staff member approves, edits, or rejects it. Nothing uncertain becomes an answer on its own.
Where learned answers live
Approved pairs join your knowledge base alongside the sources you added yourself, and Atlas cites them like any other material. See knowledge-base sources for how the rest of the base is managed.
Plans and learning choices
Learning requires a plan with AI access and the relevant Atlas settings enabled. Free and Basic provide ticketing without Atlas. New knowledge bases default to Add automatically; an existing Off setting stays off. Choose Ask me first to review every new entry.
To manage what has been learned, see reviewing learned Q&A from Discord.