What Card Coaching Does
Card coaching is an AI-powered writing assistant that helps team members improve the quality of their retrospective cards during the reflect phase. When activated, it analyzes the text of a card and suggests a revised version that is clearer, more specific, and more actionable.
Many teams struggle with vague feedback. Cards like "deployments are bad" or "communication needs improvement" are common but hard to act on. Card coaching transforms these into concrete observations that give the team something tangible to discuss and address.
Card coaching works on individual cards and is available only during the reflect phase of a retrospective. It does not alter cards automatically — the card author always decides whether to accept the suggestion.
How to Trigger Coaching
To request AI coaching on a card you have written:
- Open a retrospective that is in the reflect phase.
- Write your card content in any column, or click on an existing card you authored.
- Click the sparkle icon on the card toolbar, or select "Get AI coaching" from the card's menu.
- Wait a moment while the AI analyzes your card. A coaching suggestion will appear below your original text.
The coaching request typically takes two to five seconds depending on the length of your card. You can continue writing other cards while waiting for the suggestion to appear.
Reviewing AI Suggestions
When the AI returns a coaching suggestion, you will see a draft panel beneath your card with the suggested revision. The panel includes:
- Suggested text — The AI's revised version of your card, typically more specific and action-oriented.
- What changed — A brief note explaining why the AI made the changes it did, such as "Added specific example" or "Clarified the impact."
- Accept — Replaces your card text with the AI suggestion.
- Reject — Dismisses the suggestion and keeps your original text.
If you accept the suggestion, your card text is updated immediately. You can still edit the card afterward to make further adjustments. Accepted suggestions become regular card content with no special status.
You can request coaching multiple times on the same card. If the first suggestion does not feel right, reject it, tweak your original wording, and try again. Each request generates a fresh analysis.
Examples of Coaching in Action
Here are some real-world examples of how card coaching transforms vague feedback into actionable observations:
Example 1: Vague frustration
- Original: "Deployments are a pain"
- Coached: "Our deployment process required three manual steps last sprint that caused two failed releases. Automating the database migration step would reduce deployment failures."
Example 2: Non-specific praise
- Original: "Great teamwork this sprint"
- Coached: "The pairing sessions between frontend and backend developers on the checkout feature cut our integration bugs in half compared to last sprint. We should continue cross-functional pairing for complex features."
Example 3: Broad complaint
- Original: "Communication needs to be better"
- Coached: "Design decisions made in the Monday sync were not shared with developers until Wednesday, which blocked the profile page work for two days. Posting sync summaries in the team channel within an hour of the meeting would prevent these delays."
What the AI Optimizes For
Card coaching applies several principles when suggesting revisions:
- Specificity — Replaces generalizations with concrete details where possible, encouraging the author to recall specific events or examples.
- Actionability — Frames observations in a way that points toward a potential action the team can take.
- Constructive tone — Preserves the author's intent while ensuring the language is constructive and forward-looking rather than blaming.
- Appropriate length — Keeps suggestions concise enough to be read quickly during discussion, while including enough detail to be meaningful.
Card coaching does not have access to your sprint data, ticket history, or other external context. Its suggestions are based solely on the text you write. If the AI adds a detail that is not accurate, reject the suggestion or edit the accepted text to correct it.
Privacy and Anonymity
Card coaching preserves the anonymous-by-default principle of Unpack. When the AI processes your card, only the card text is sent — no author information, no team member names, and no metadata about who wrote the card. The coaching interaction is visible only to you until you accept a suggestion, at which point the updated card text appears like any other card.
Other team members cannot see that you used card coaching, and there is no indicator on the card itself that AI was involved. The goal is to help every team member contribute their best thinking without judgment about whether they needed assistance.
When to Use Card Coaching
Card coaching is most valuable in these situations:
- When you know something was a problem but are struggling to articulate it clearly.
- When you want to ensure your feedback is constructive rather than purely critical.
- When you are new to retrospectives and want guidance on writing effective cards.
- When you want to move from observation to suggestion and need help framing a recommendation.
Card coaching is less useful when your card is already specific and actionable. If you have written a detailed observation with a clear suggestion, the AI may not have much to add. In those cases, the coaching suggestion will often closely mirror your original text.