Decompose
Break an outcome into suggested work cards with effort estimates.
Decompose is the AI agent's most powerful action. Give it an outcome with a clear goal, and it generates a complete set of work cards — each with a name, goal description, effort estimate, and dependency mapping. Instead of spending 30 minutes manually listing tasks, you get a draft breakdown in seconds.
How Decompose works
The agent reads your outcome's:
- Goal — What you're trying to achieve
- Context — Background information and current state
- Directives — Constraints and guidance
- Success signals — What done looks like
- Existing cards — Cards already created (to avoid duplication)
It then generates a set of suggested work cards that collectively accomplish the goal.
Running Decompose
Open the outcome
Navigate to the outcome you want to decompose.
Click AI → Break into work cards
Click the AI button in the header and select Break into work cards from the dropdown.
Wait for the agent
The AI drawer opens with a reasoning trace showing the agent's thinking process in real time. After a few seconds, the results appear.
Reviewing suggestions
Each suggested card shows:
- Name — A concise task title
- Goal — What completing this card achieves
- Estimated effort — How long the work should take (e.g., "2 hours," "1 day," "3 days")
- Dependencies — Which other suggested cards this one depends on
- Suggested assignee — An optional recommendation based on team member skills (if member context is available)
Accepting cards
You have full control over which suggestions become real cards:
- Accept — Click the accept button on individual cards to create them
- Accept All — Click Accept All to add every suggestion at once
- Preview — Click Preview to see the full card detail before deciding
- Dismiss — Click Dismiss to clear all suggestions
Accepted cards are created immediately under the outcome in Backlog status.
Don't accept blindly — review each card's goal and effort estimate. The AI is a starting point, not the final plan. Adjust names, combine cards that overlap, or split cards that feel too large.
Getting better results
The quality of Decompose output depends directly on the quality of your outcome:
- Detailed goals → More specific and actionable cards
- Context → Helps the AI understand what already exists and what needs to change
- Directives → Ensures suggested cards respect constraints like "mobile-first" or "must use existing API"
- Success signals → Gives the AI clear completion criteria to work toward
If the initial decomposition doesn't quite hit the mark, refine your outcome's goal and context, then run Decompose again. The agent will incorporate your updates and any existing cards to avoid duplication.
Next Steps
- AI Agent Overview — Return to the full overview of AI features
- Creating Outcomes — Write better outcomes for better decomposition
- Creating Work Cards — Other ways to create cards