Agent Workspace

The Agent Workspace is where you meet your Org's AI workforce face to face. Instead of digging through separate screens for each department, you see every agent gathered in one visual space, and you handle anything that needs your attention right there, in front of the agent that raised it. Use this page whenever you want to check in on your agents, see what they are about to do on your Org's behalf, and decide whether to let it happen.

What you can do here

• See your Org's agents rendered as characters in a shared visual space, so you have a sense of your whole team at once.

• Open any agent to bring up their panel and see what they are working on.

• Review the outbound actions an agent wants to take before they happen: sending an email, posting a message, scheduling an event, filing a document, or signing an order.

• Approve an action so the agent carries it out, or reject it, optionally with a reason.

• Set how much you trust each type of action, from needing your sign-off every time to letting the agent act on its own for that action type.

• See a clear notice when an action can't run yet because the integration it needs isn't connected.

Before you start

You need an Org with agents already staffed and assigned to roles. If you haven't set up your AI workforce yet, do that first; Building Your Agent Team walks through hiring and assigning agents so there is a team here to review.

Working with your agents

1. Open the Agent Workspace for your Org. Your agents appear together in the hall, each represented by their own avatar.

2. Click on the agent you want to check on. Their panel opens and brings them into focus.

3. Look under "Actions awaiting you" in that panel. Any action the agent has proposed, and hasn't yet been allowed to carry out, is listed here with a plain description of what it will do.

4. Read the summary of the action and decide: Approve it to let the agent go ahead, or Reject it (you can add a reason) to stop it.

5. If you want to change how closely you supervise a particular kind of action going forward, use the autonomy control next to that action type. Raise it to let the agent handle that type of action on its own more often, or lower it to keep reviewing every one yourself.

6. If an action shows a note that it needs a connection that isn't set up yet, that action will stay pending until the integration it depends on is connected.

Example

Your support agent drafts a reply to a customer and proposes to send it. You open the Agent Workspace, click on the support agent's avatar, and find the draft listed under "Actions awaiting you." You read the summary, approve it, and the email goes out. The next time the same agent drafts a routine reply, you could raise the autonomy setting for "send email" so it no longer waits for your approval on that kind of action.

What to expect

Approving an action lets the agent complete it immediately. Rejecting an action stops it from happening. Any autonomy level you set for an action type stays in place, so future actions of that same type are handled according to the trust level you chose, until you change it again.

Common mistakes

• Leaving actions unreviewed for too long. Actions wait for you, so check the workspace regularly rather than letting them pile up.

• Raising an action type's autonomy to full trust before you've seen a few examples of what that agent proposes for it.

• Expecting an action to run when its required integration hasn't been connected yet; look for the connection notice before assuming an approval will take effect.

Related features

If you haven't assigned agents to your Org yet, see Building Your Agent Team to hire and equip the team you'll be reviewing here. Once your agents are active, Generating AI Reports shows you how to pull together a summary of what your team has been doing beyond the actions you approve one by one.