Self-Learning
When a human agent resolves a chat, a visitor gives a thumbs-down, or the bot stumbles and recovers, BubblaV drafts a new Q&A from what actually worked. You review the suggestion — or let strong answers publish automatically — and your bot gets smarter without touching the knowledge base.
What Self-Learning does
Suggested-answer review queue
Learned answers land in Knowledge → Q&A as drafts with a Suggested badge. Edit, approve, or discard — nothing goes live without you.
Auto-apply strong answers (Pro)
Turn on auto-apply to publish strong-evidence drafts immediately — typically your own agent's replies on resolved chats.
Conversation provenance
Every learned entry shows which conversation it came from, so you always know why your bot knows what it knows.
Weekly learning report
Your weekly report email counts how many new answers your bot learned from conversations.
How it works
From setup to a resolved conversation.
Your bot drafts
Resolved handoffs, bot self-corrections, and thumbs-down threads are mined nightly and drafted into Q&A suggestions.
You approve
Review suggestions in Knowledge → Q&A — tweak the wording, approve to publish, or discard.
Your bot improves
Approved answers join your knowledge base and are served by the same RAG pipeline. Or enable auto-apply for strong answers.
Self-Learning FAQ
How BubblaV self-learning works.
Will learned answers go live without my approval?
Will learned answers go live without my approval?
No. Drafts never go live unless you approve them — or unless you explicitly enable auto-apply, which only publishes strong-evidence answers (like your agent's replies on resolved chats).
What if the bot drafts a wrong answer?
What if the bot drafts a wrong answer?
Edit it before approving, or discard it. Answers are drafted only from your real conversation transcripts, and questions you dismissed in Content Gaps are never re-proposed.