AI Refinement

AI Refinement helps you transform rough backlog items into well-specified, ready-to-develop tickets. It analyses your item and suggests acceptance criteria, edge cases, and validation steps based on your personas and best practices.

Starting a Refinement Session

1. Open a backlog item detail page.
2. Click the Refine with AI button (or the Sparkles icon).
3. The AI Refinement panel opens alongside the item content.

What the AI Analyses

The AI considers:

• The item title and description
• The item type (feature, bug, spike, chore)
• Project personas on the roster
• Best practices assigned to the project from the Library
• Existing acceptance criteria (to avoid duplication)

Refinement Chat

The refinement panel is a chat interface. You can:

• Ask “What edge cases should I consider?”
• Ask “Write acceptance criteria for the user authentication flow”
• Ask “What deployment steps will this require?”
• Challenge a suggestion: “That AC doesn’t make sense because…”

The AI remembers context within the session so you can refine iteratively.

Accepting Suggestions

When the AI proposes acceptance criteria, edge cases, or other content:

1. Review each suggestion in the chat.
2. Click Apply next to a suggestion to add it to the item’s structured fields.
3. Edited accepted suggestions update the underlying field immediately.

Quality Score

Every time you add or edit content, the item’s quality score recalculates. Target 80+ to move an item to “Ready” status. The quality indicator in the top-right of the item detail shows the current score and which fields need attention.

Clearing Session History

Close the refinement panel to end the session. Click the panel title and select Clear history to start a fresh session on the same item.