> ## Documentation Index
> Fetch the complete documentation index at: https://docs.truestate.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Jobs and recurring work

> Give an agent reusable instructions and inspect each run.

A Job defines work an agent should perform on demand or on a schedule. Use it for tasks such as reviewing a dashboard, investigating changes, or preparing commentary.

## Choose the right tool

| Need                                                             | Use          |
| ---------------------------------------------------------------- | ------------ |
| Explore a question interactively                                 | Conversation |
| Repeat defined imports, transformations, training, or inference  | Pipeline     |
| Repeat instructions that require an agent to inspect and explain | Job          |
| Build a visual sequence of workflow actions                      | Automation   |

## Define the task

Open **Jobs** in the Agent sidebar. Prepare a clear instruction, choose the assignee agent and relevant context assets, and configure a schedule only when needed.

```text theme={null}
Review the weekly sales dashboard. Compare the most recent complete week
with the preceding week. Identify the three largest changes, check whether
data is complete, and draft commentary for the sales report. If the source
is stale or incomplete, explain that instead of drawing a firm conclusion.
```

State what may change, what the output should be, and how to handle missing information. Name the destination report or dashboard when the job should update one.

## Run once before scheduling

Run the job on demand and inspect its history and output. Check that it used the intended sources, respected the reporting period, and produced the expected result.

Then configure the schedule, checking the timezone and source-data availability. If the job depends on a refreshed dataset, arrange the timing so the data is ready first.

## Maintain a recurring job

Review failures and unexpected output in the run history. Update instructions when the business definition or attached assets change. Review the schedule when retiring a source, dataset, or report.

A Job may call a pipeline as part of its work, but the Job's instructions and the pipeline's processing logic serve different purposes. Keep both understandable so failures can be traced to the right step.