# Agent chat

Open a module's agent, ask it questions or give it work, attach files and screenshots, follow what it does, and stop it or add to your request while it works.

Every module has an AI agent that knows the module and can do its work for you: look up records, explain what you're seeing, fill in and change records, and run the module's actions. You ask in plain words, as you would ask a colleague.

## Open the chat

1. Open the module from the sidebar.
2. Click **Open Agent Chat** at the top of the module's panel (the robot button).

The chat opens on the right. A new chat suggests a few things to try, such as **Show me the current status.** Click one, or type your own.

## Ask

1. Type in the box at the bottom (**Ask the agent…**). **Shift + Enter** starts a new line.
2. Press **Enter** or click **Send**.

Be specific, the way you would with a new colleague: "List the open work orders for the Riverside site, oldest first" works better than "show work orders".

## Follow what it does

While the agent works you see **Thinking…**, then a card for each step it takes, such as **Get desktop report**:

- **…** means the step is running, **✓** that it worked and **✗** that it failed.
- Click **▶** on a card to see exactly what it looked up or changed.

The answer follows the cards. If something isn't right, say so in your next message: "No, only the ones assigned to me."

## Attach files and screenshots

Give the agent something to work from: a photo, a PDF, a spreadsheet, an email or a screenshot.

- Click **Attach** (the paperclip) and pick one or more files, or
- drag files onto the chat box, or
- paste a screenshot with **Ctrl + V**.

Each file shows as a card above the box. Click **×** on a card to remove it before sending. Long pasted text becomes an attached text file automatically.

## Add to your request, or stop it

While the agent is working:

- **Type another message** and press **Enter**. It's marked **Queued** and goes when the agent finishes. Click **Send now** to give it to the agent straight away, for example "and include last month too".
- **Stop it:** click **Stop** at the top of the chat, or the square button in the box. It stops at its next step.

## Talk instead of typing

Click the mic button by the chat box to talk to the agent. See [Talk to an agent by voice](article:59f107f8-0eb9-4ff2-b8f0-913e78600916).

## Your conversation

- **It picks up where you left off.** When you open the chat again, your conversation from the last 24 hours with that agent is still there.
- **Clear** at the top of the chat clears the screen to start fresh. The agent still keeps any notes it saved about how you like things done.
- **Each module's agent is separate.** What you told the Work Orders agent isn't known to the CRM agent.

## If it doesn't work

| You see | What to do |
|---|---|
| **AI isn't set up for your organization yet** | Your workspace has no AI key for this agent. A Workspace Admin sets one up in **AI providers**; everyone else, ask your Workspace Admin. See [AI providers](article:e8edfb1d-8462-4a7c-bfc9-b923bbe1343e). |
| **This month's AI allowance is used up** | A Workspace Admin can add more in **System settings**, under **Billing**. Otherwise it resets on the 1st. See [Billing](article:74594451-2ed3-4d0f-b24a-33799c898a43). |
| **Permission denied** in a step | Your role doesn't allow that action, so the agent can't do it for you either. Ask your Workspace Admin. |

## Good to know

- **The agent can only do what you can.** It uses your permissions, never more. See [What an AI agent can and can't do for you](article:d4e4354f-ec97-45b6-93bc-11f36b91fec3).
- **Which AI model** an agent uses is set by your Workspace Admin. Admins see a model chip by the chat box to change it for everyone.
- **Agents can fill in forms on your screen** while you watch, if your workspace has switched that on. See [Shotgun Assistant: fill in a form](article:e1c1bbeb-58ac-43b7-8b3d-6914203b8540).


Source: https://datavisions.ai/docs/everyday-work/agent-chat
