# Send a daily report

<TLDR>
In **Agents**, select **Browse templates** and pick **Daily summary of a table**. Let the AI fit it to your table, select **Try it**, then **Go live**. Every morning at 8 it summarises what changed.
</TLDR>

## What you will build [#what-you-will-build]

A live <Term id="agent">agent</Term> that reads the newest rows of one of your <Term id="table">tables</Term> every morning and summarises what changed, in plain words. You start from a template, fit it, check one <Term id="run">run</Term> and take it live.

## Before you start [#before-you-start]

- A project you can edit, and a table that gets new rows regularly (for example orders loaded each night). See [Turn a spreadsheet into a table](/docs/guides/spreadsheet-to-table) or [Connect your first data source](/docs/guides/connect-data).
- To take it live: the Owner or Admin role. A live agent spends the workspace's <Term id="credits">credits</Term>.

<Callout type="info" title="Tip">
Describe your table's columns first. The agent reads them, and the summary gets sharper. See [Tables](/docs/data/tables).
</Callout>

## Steps [#steps]

<Steps>
<Step>
### Open the templates [#open-the-templates]
Go to <AppLink to="agents" /> and select **Browse templates**.
</Step>
<Step>
### Pick Daily summary of a table [#pick-daily-summary-of-a-table]
It runs every morning at 8.

- Select **Use** to let the AI fit it to your data. Tell it which table to read, for example "use the orders table and focus on totals by region".
- Or select **Add as a draft, as it is** and change the steps yourself.
</Step>
<Step>
### Check the schedule [#check-the-schedule]
Open the step that **Starts** the agent and change the time or days if needed. The time is shown with its time zone.
</Step>
<Step>
### Try it once [#try-it-once]
Select **Try it**. The draft runs once, now; nothing goes live.
</Step>
<Step>
### Read the summary [#read-the-summary]
Open the run from **Runs** and look at the **Result**. If it misses what matters, change the step's instruction (for example "start with the total, then the three largest changes") and try again.

<Screenshot src="/docs/_media/demp-bakery-business-data-agent-flow-page.png" alt="A run of the daily summary agent with its Result" />
</Step>
<Step>
### Take it live [#take-it-live]
A workspace admin selects **Go live** and confirms. It now runs every morning on its own.
</Step>
</Steps>

<Screenshot src="/docs/_media/agents-overview.png" alt="The agent templates, with Daily summary of a table" />

## Check that it worked [#check-that-it-worked]

- The agent's page shows it as live, starting daily at your time and time zone.
- The next morning, a new run appears under **Runs** with its summary and its cost in credits.

## Common variations [#common-variations]

- **A weekly report as a file.** The **Weekly report as a file** template writes a downloadable report on your tables and how they grew, every Monday at 9.
- **Score each new row.** The **Score each new item with AI** template scores each new row from 1 to 5, asks you, then saves the scores.
- **Copy a web API into a table.** The **Copy a web API into a table** template calls a web address that returns a list and adds the results to a table, keeping a history.
- **Run it on demand.** Change the start to run only when started, and select **Run** when you want a summary.

## If something goes wrong [#if-something-goes-wrong]

<Accordions>
<Accordion title="The summary says nothing changed">
No new rows since the last run. Check that loading into the table still works.
</Accordion>
<Accordion title="The run failed">
Open the failed step and read its log. Fix it, then choose **Run again from here** to reuse the earlier steps.
</Accordion>
<Accordion title="It did not run this morning">
Check that the agent is live and not paused, and that your workspace has credits in <AppLink to="settings.usage" />.
</Accordion>
</Accordions>

## Next steps [#next-steps]

- [Templates](/docs/agents/templates): every ready-made agent.
- [Runs](/docs/agents/runs): read and re-run any run.
- [Build your first agent](/docs/guides/first-agent): describe an agent from scratch.