Zoona AI Agent Commands

Commands let your AI Agent do more than just reply to customers. Instead of only sending messages, the agent can take real steps in the background, like updating a conversation, looking up a contact in your CRM, or routing a ticket based on the outcome.

Commands are set up in advance by your team's admin, so the agent works within a defined playbook. It can decide when to run a command and which path to take, but it cannot invent steps on its own.

How it works

When a conversation matches a trigger (such as a customer reporting an error or a specific intent being detected), the AI Agent evaluates whether to run a Command. A Command is a sequence of predefined steps that execute in order.

The agent decides:

• Whether to run the command at all
• Which path to follow if there is a condition (for example, "if contact exists, do X, otherwise do Y")

Once triggered, the steps run automatically and deterministically.

What can a Command do?

Conversation updates

• Change priority, assignee, or conversation state
• Add or update tags

Contact updates

• Update contact properties (for example, marking a customer as an existing subscriber)

External API calls

• Look up or send data to connected tools like HubSpot
• Supports GET and POST requests with API key or OAuth authentication

Logic and conditions

• Branch based on whether a value exists, matches, or contains something
• If / Else conditions to handle different customer situations

Responses

• Send a message to the customer based on the outcome of the above steps

Example: A customer reports an error

Here is what an AI Agent might do when it detects an error report:

1. Tags the conversation with the relevant label
2. Sets priority to urgent
3. Calls the HubSpot API to look up the customer by email
4. If the contact exists, marks them as an existing customer
5. If not, marks them accordingly
6. Sends the customer an appropriate response based on the result

All of this happens without a human having to manually intervene.

Viewing what the agent did

Every time a Command runs, a step-by-step execution log is created. You can see:

• Each step that ran
• The inputs and outputs at each step
• Whether each step succeeded or failed
• A timestamp for the full execution

This makes it straightforward to understand what the agent did and debug anything unexpected.

What the agent cannot do

To keep things safe and predictable, the agent cannot:

• Create its own workflows or invent steps not already defined
• Run steps outside of predefined Commands
• Automatically retry failed steps (this is handled by your admin configuration)

Who can set up Commands?

Commands are created and managed by anyone with Manage AI Agent permission. If you need a new command or workflow added, reach out to your team admin.