> ## Documentation Index
> Fetch the complete documentation index at: https://docs.marketer.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Tasks and activity
> Follow a task from approval through live steps, input requests, outputs, and the final owning-product state.
A task tracks work that continues beyond one response. Its detail tells you what is happening now, which step or specialist owns the work, what output was produced, and whether a human decision is still required.
Do not use elapsed time or a quiet chat as the status. Use the state and latest step shown on the task.
## Open the same task from anywhere
You can reach a task from several starting points:
* Select its card in the conversation that started it.
* Select its row under **Currently Running** on **Home**.
* Open **Chats and tasks**, select **Tasks**, and choose the task row.
* Open **Activity** from Quick actions to find active or historical work.
* Open **Agents** from the right edge of a conversation, then select the run.
* Follow a task or run link from **Notifications**.
These entry points should lead back to the same work. Confirm the task title, workspace, originating conversation, and created time before acting on it.
## Read live progress
The in-conversation activity disclosure has three levels of information:
| Level | What it answers | How to use it |
| ---------------- | -------------------------------------------- | ------------------------------------------------------------------------------------- |
| **Working for…** | Is the current turn still active? | Treat it as elapsed activity, not an estimated completion time. |
| **Searched** | Which visible sources have been consulted? | Open or compare them with the evidence boundary in your request. |
| Step timeline | What is complete, current, or still pending? | A green check is complete, the dark marker is current, and muted markers are pending. |
For a longer task, its result block can also show the current step count and a progress control. For coordinated work, open **Agents** to see the specialist or pipeline behind each step.
## Follow a running task
Match the task to the originating conversation and intended workspace. If two runs have similar titles, use the created time and attached context to distinguish them.
Use the state on the selected run as its current truth. Compact run rows use **Running**, **Waiting for you**, **Done**, **Failed**, or **Canceled**. Activity can show broader family states such as **Queued**, **Working**, **Waiting**, **Completed**, or **Partially completed**.
Identify the latest active or unresolved step. If an agent owns it, select that agent in the workbench to inspect its goal, progress, and available sources.
When the task is **Waiting**, answer the requested question or decide the linked approval. Return to the same run after the decision; do not start a replacement task while it is resuming.
At a terminal state, open the result or artifact itself. Compare it with the original request and follow any output link to the product surface that owns it.
## Respond to each lifecycle state
| State | What changed visibly | What you should do |
| ---------------------------------- | --------------------------------------------------------------- | --------------------------------------------------------------------------------- |
| **Queued** | The run exists, but no work step is active yet | Keep the original run; check prerequisites if it remains queued unexpectedly. |
| **Running** or **Working** | A step or specialist is active and progress can update | Open the active step when you need detail; otherwise let the run continue. |
| **Waiting for you** or **Waiting** | Progress stops on a question, approval, or missing prerequisite | Supply the requested input or open **Awaiting action**. |
| **Done** or **Completed** | Planned work reached its end and a result may be available | Open the result; verify completeness, evidence, and the owning product state. |
| **Partially completed** | Some output is available alongside unfinished work | Use only the supported portion and inspect the incomplete branch before retrying. |
| **Failed** | The run ended on an error | Read the last successful step, error, connection state, and any preserved output. |
| **Canceled** | Work stopped before the planned end | Decide whether to restart from the original request or revise the scope first. |
### Waiting on a question
Return to the originating conversation and answer with the missing fact or constraint. Keep the reply narrow so it resolves the pause without changing the task's objective.
### Waiting on a decision
Open **Awaiting action** and review the exact proposal. An approval can release the current run or allow a recorded change; a rejection can stop that branch. Return to the task afterward and confirm its new state.
### Failed or partially completed
Do not immediately resubmit the entire request. First determine:
1. Which step last completed successfully.
2. Whether an output was preserved.
3. Whether the failure came from a disconnected source, missing permission, invalid destination, or the task itself.
4. Whether retrying would create duplicate creative, workflow, or campaign work.
Retry only after the cause is resolved. If usable output exists, continue from that output instead of repeating finished work.
## Review a completed task
The completed example shows three distinct layers:
1. The response explains the outcome in the conversation.
2. The result block shows `3/3 steps` and marks each step complete.
3. Suggested follow-ups offer separate next requests, such as writing copy or adapting the launch.
Review those layers in order. A suggested follow-up is not work that already happened.
For the output itself, verify:
* It answers the original objective and preserves the stated boundary.
* Required records, products, or connected sources are the intended ones.
* Decision-driving figures identify their dates and measurement context.
* Missing data, failed steps, and assumptions are visible.
* Generated creative or copy matches approved product facts and review criteria.
* Any recommended account change is still a proposal unless a separate result confirms otherwise.
## Use Activity for the workspace-wide record
Open **Activity** when you do not know which conversation owns the work or when you need history across the workspace.
The page separates current execution from broader history:
* **Agent work (*n*) · last 7 days** contains task and run records.
* **Currently active (*n*)** groups work that has not reached a terminal state.
* **Recent (*n*) · last 24h** shows recent workspace events.
* **Full history (*n*)** supports source, agent, and date filters; **Search activity…**; grouping **By time**, **By agent**, or **By category**; and paging.
Select a run to open its detail. Native agent detail can expose the live run body, its current question or approval, **Stop** while the run is non-terminal, and an **Outputs** section. Other expanded Activity entries can show an **Error**, the **Trigger**, fields that **Changed** as Before → After, a **Linked** record, and the originating **Chat**. Use the open-detail control from the Agents panel when you want a reload-safe full-page record.
### Activity is an audit trail, not the final object
Activity can prove that Marketer recorded a run or event. It does not replace the current campaign, creative, brief, workflow, or provider state.
| Activity says… | Verify next in… |
| ------------------------------ | ---------------------------------------------------------- |
| A campaign proposal completed | **Awaiting action**, then **Campaigns** after any approval |
| A creative run completed | The run output, then **Studio** or **Assets** |
| A brief was created | **Briefs** |
| A recurring run executed | **Workflow** and its run history |
| A report was produced | The linked report output and its source settings |
| A connected-account change ran | The owning Marketer surface and, when needed, the provider |
## Avoid duplicate work
Before starting a second task with the same objective:
1. Check **Currently Running** on Home.
2. Search **Activity** for the task title or agent.
3. Check **Awaiting action** for a paused approval.
4. Open the original conversation for a waiting question.
5. Confirm that the first run reached a terminal state.
Starting again while the first run is queued, working, or resuming can create overlapping analysis or duplicate artifacts.
## Next step
Resume waiting work, stop an active run, recover a failure, and verify the output.
Reopen the correct conversation or recent task from the combined history.
Move from the task summary to the specialist, output, and source behind a step.
Review the recorded scope before approving or rejecting.
Follow receipts back to the conversation, task, or product record that explains them.
Apply a source, time, scope, calculation, and action check.
## Related topics
- [Drafts and launches](/product-guides/campaigns-drafts-and-launches.md)
- [Chats and tasks](/working-with-marketer/chats-and-tasks.md)
- [Task control and recovery](/working-with-marketer/task-control-and-recovery.md)
- [Answers, tasks, and proposals](/working-with-marketer/answers-tasks-and-proposals.md)
- [Run your first task](/get-started/first-task.md)