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

# Ask the AutoAdy Agent

> Give the Agent a clear account, period, question, and boundary for any proposed change.

Use **Agent** to investigate connected data and prepare work. A precise question makes its answer easier to check and its proposed actions easier to review.

## Start with a bounded question

1. Confirm the Meta account in the sidebar, then open [Agent](https://app.autoady.io/agent).
2. State the question and period. For example: “Review this account's last 14 days. Which ads need investigation? Show the metric, period, and evidence for each finding.”
3. Add your actual objective or target if it matters. Ask the Agent to identify missing information instead of assuming a lead is qualified or an order is paid.
4. Inspect the answer's account, date window, entities, sources, and limitations. Open the relevant table when you need to verify the underlying data.
5. If you want a change, ask for a specific proposal and inspect its approval preview. Confirm the exact action only after checking the affected account, entity, and amount.

## Useful prompts

> Compare campaign performance for the last 30 days. Keep campaigns with different result definitions separate. Tell me if coverage or attribution limits the comparison.

> Prepare a budget-change proposal for this ad set. Show its current budget, budget type, proposed amount, and currency before asking me to approve anything.

> Draft a client summary from the report. Explain observations and uncertainties without claiming that a change caused an improvement.

## Confirm what happened

An answer describing an action is different from a confirmed write result. After approving a change, read the tool result or receipt and check current provider state. Follow the offered recovery for an unknown outcome before making a fresh request.

If a plan, role, credit, or connection gate blocks the work, resolve the displayed reason. Rephrasing the request does not provide missing account permissions or activate an unavailable source.

The built-in Agent uses approval previews for supported writes. An external agent has different approval controls: read [Connect your own agent](/agent/connect) before sharing credentials.

Next: [Review delivery changes](/meta-ads/delivery) or turn a recurring read into a [scheduled task](/automation/scheduled-tasks).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.