# Use the Agent with your project work

Ask questions, inspect sources, and review proposed ideas, insights, and experiments.

The **Agent** is an AI-assisted experience for working with team context. It is available when the product offers it, your team has not opted out, and the account's access and AI usage limits allow it.

## Start a conversation

1. Choose the intended team and project context.
2. Open **Agent** or an AI entry point on the item you are reviewing.
3. Ask a question that names the work or outcome you care about.
4. Inspect the response, referenced entities, and visible tool activity.
5. Open linked work to verify important claims against the source records.

For example: “Which observations support our checkout ideas?” is more focused than “What should we test?” Include the relevant project or item when several have similar names.

## Review proposed records

The Agent can present proposals for ideas, insights, and experiments. A proposal card lets you inspect the name, description, project, and related fields before creating the item. Edit inaccurate wording or relationships, then choose the create action. Discard proposals you do not want to turn into work.

A proposed record is not saved merely because text describing it appears in chat. Confirm the card's created state and open the resulting item after accepting it. Creation still respects your role and plan limits.

## Return to a conversation

Use the Agent's conversation list to reopen your own threads. AI Conversations belong to the individual user within a team and are private by default; they are not a shared team chat history.

Use response feedback controls when an answer is helpful or problematic. If a response is incomplete or a tool fails, preserve the specific error and retry a focused request after checking the source work.

If Agent disappears, check [AI availability and privacy](https://conversionlab.app/docs/ai/privacy). The core research and experiment workflows remain available according to the team's ordinary access.