> ## Documentation Index
> Fetch the complete documentation index at: https://notiky-mintlify-b5788c9c.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Decision to task

> PM workflow — use Conversations to think, then assign an agent to ship.

Goal: **turn a product decision into assigned work** without losing context between chat and the board.

<Frame>
  <img src="https://mintcdn.com/notiky-mintlify-b5788c9c/kOTGsdF2jDE8PXYg/images/recipes/recipe-decision-plan.png?fit=max&auto=format&n=kOTGsdF2jDE8PXYg&q=85&s=2b3feeff78a22a0263256b53d1a852a1" alt="Conversation in Plan mode shaping a product decision" width="1176" height="828" data-path="images/recipes/recipe-decision-plan.png" />
</Frame>

***

## Scenario

You're deciding whether to ship feature X this sprint. You want to:

1. Explore tradeoffs with AI that pushes back (not yes-man chat)
2. Break the decision into tasks with enough context
3. Assign implementation to an agent on a connected runtime

***

## Steps

<Steps>
  <Step title="Open a conversation in Think mode">
    From your workspace, open **Conversations**. Select **Think** in the mode pills (**Ask**, **Think**, **Plan**, **Do**).

    Prompt example:

    > We're considering shipping SSO for enterprise. What are the risks we haven't written down? Challenge our assumptions.
  </Step>

  <Step title="Capture the decision">
    When you're aligned, switch to **Plan**. Ask Notiky to draft tasks with titles, descriptions, and suggested priority — still without mutating the board.
  </Step>

  <Step title="Create tasks on the board">
    Switch to **Do** if you want the agent to create tasks directly, or copy the plan into **New Task** manually for more control.
  </Step>

  <Step title="Assign implementation">
    Pick one task with clear acceptance criteria. Assign it to an [agent](/agents) that has access to your repo via a connected runtime.

    <Frame>
      <img src="https://mintcdn.com/notiky-mintlify-b5788c9c/kOTGsdF2jDE8PXYg/images/guides/tasks-assign-agent.png?fit=max&auto=format&n=kOTGsdF2jDE8PXYg&q=85&s=563214b2d764ce144fe83bd3c13a4bd6" alt="Task detail with Builder assigned in Properties" width="1176" height="828" data-path="images/guides/tasks-assign-agent.png" />
    </Frame>
  </Step>

  <Step title="Track with the team">
    Use [Tasks](/tasks) board columns — **Backlog**, **Todo**, **In Progress** — so everyone sees status. Link back to the conversation in a task comment for audit trail.
  </Step>
</Steps>

***

## Mode cheat sheet

| Mode  | Use when                                       |
| ----- | ---------------------------------------------- |
| Ask   | Quick factual questions from workspace context |
| Think | Exploring decisions — AI can argue back        |
| Plan  | Drafting tasks without executing               |
| Do    | Creating/updating tasks and artifacts          |

Full reference: [Conversations](/conversations).

***

## Tips for PMs

* Keep one decision per conversation thread when possible.
* Task descriptions should include **why**, not just **what** — agents use the same context humans do.
* Use [Projects](/projects) if the work spans multiple tasks for one initiative.

***

<AccordionGroup>
  <Accordion title="Conversation got too long">
    Start a new thread for a new decision. Link the old conversation URL in the first task comment for traceability.
  </Accordion>

  <Accordion title="Plan mode output is too vague">
    Ask for acceptance criteria per task before moving to **Do** or manual task creation.
  </Accordion>
</AccordionGroup>

***

## Related

<CardGroup cols={2}>
  <Card title="Conversations" href="/conversations" />

  <Card title="Tasks" href="/tasks" />

  <Card title="First agent task" href="/example-workflows/first-agent-task" />

  <Card title="Spec review loop" href="/example-workflows/spec-review-loop" />

  <Card title="FAQ" href="/faq">
    Common questions about agents and Conversations.
  </Card>
</CardGroup>