# Working a task end to end

> How to take a GetIntel task from To do to Done, what each part of the brief means, and how to confirm the fix worked.

Source: https://getintel.ai/docs/academy/take-action/working-a-task/

A task is only useful once it ships. This is the full loop, from picking a task to seeing it move your numbers.

<Steps>

1. **Pick the task.** Take it from **Weekly plan**, or the top of any layer under **Actions**.

2. **Read the brief.** **Why this matters** gives the evidence: the answers and topics behind the task. **What to do** is the checklist. **Your work** holds anything already assembled, such as the facts AI compares or the pages it cites for a rival.

3. **Start it.** Choose **Start** so the task moves to **In progress**, and tick steps off as you go.

4. **Do the work, or hand it to your agent.** On-page fixes can go straight to [Claude Code, Cursor or Codex CLI](/blog/ai-visibility-tools-mcp-claude-code/): ask it to read the task with `get_task` and make the change in your repo. See [Ship fixes with your coding agent](/docs/academy/playbooks/ship-fixes-with-your-coding-agent/).

5. **Mark it done.** Choose **Mark done**.

6. **Check the effect.** Look at the topics the task moves after a full 7 days of new answers.

</Steps>

<Aside>
If GetIntel stops finding the problem before you mark the task done, it shows as **Detected as done**. That is confirmation the fix is visible to its scans.
</Aside>

## When a task isn't right

- **Not relevant**: dismiss it, for example a comparison against a brand you don't compete with.
- **Snooze**: hide it for 2 weeks when the timing is wrong.

## How long results take

Technical fixes can show in the next technical scan (or run **Re-check now**). Content and off-page work take longer: engines need to find the change, and topic scores average over 7 days. Give content changes a few weeks to a month before judging them.