Create an AI Agent

Overview

Set up an AI agent that reacts to activity in one of your projects and creates tickets, adds comments or sends emails without anyone starting it.

Agent runs consume your AI usage limit, the same limit as the AI Assistant. Read more in AI Usage Limits and Upgrades.

Read more about Privacy & Security for AI Features.

See it in action

Where to find your agents

Click AI Agents in the sidebar.

What an agent needs before it can run

Four things. The Testing panel beside the agent lists them and marks off each one as you complete it.

Until all four are in place you can still save the agent, but you cannot test it and you cannot activate it.

Start from a template

Four ready-made agents come with your account, each one already written. Change any part of it before you activate it.

1. Click AI Agents in the sidebar.
2. Click Create new agent.
3. Under Our templates, click a template.
4. Check the Agent name, Description and Instructions, and change them if you want.
5. Under Projects, click Add and choose the project the agent runs in.
6. Check the trigger and the actions. Click Edit on either one to change it.
7. In the Testing panel, click Test run and read what the agent would have done.
8. Click Activate agent.

Describe your own agent

You do not have to start from a template. On the Explain step you can write what you want in your own words.

1. Click AI Agents in the sidebar, then Create new agent.
2. In Instructions, describe what the agent should do, for example: When a ticket is created without a due date, add a comment asking for one.
3. Click Next.
4. On the Configure step, click Add on Projects, Triggers and Actions in turn.
5. Click Review and test.
6. Give the agent a name. A description is optional.
7. In the Testing panel, click Test run.
8. Click Activate agent.

Click Rewrite with AI under the instructions to have your wording tidied up. Keep the result, undo it, or write your own version instead.

Click Save as draft at any point to come back later. A draft never runs.

Triggers

An agent runs when one of its triggers fires in its project. An agent can have more than one trigger.

Ticket triggers can be narrowed to chosen forms and layers. File triggers can be narrowed to chosen folders in Documents.

Agents do not run on a schedule, and you cannot start one from the AI Assistant.

Actions

An agent can always read the tickets and documents in its project. That is how it answers a question or checks for a duplicate, and you do not switch it on. Actions are what the agent is allowed to write, and you add them yourself.

An agent can use each action once.

On Create ticket, choose the form first. The form decides which fields a ticket has, so the other settings stay switched off until you select a form, and a field the form does not carry stays switched off.

On Send email, you can either choose recipients in the action or name a person in the instructions. Recipients chosen in the action win.

An agent can only email active members of its own project. External addresses are not possible.

Test before you activate

A test does everything a real run does except write. The agent reads the same data and reaches the same decisions, and the result shows you what it would have created, commented or sent. Nothing is created or changed in your project. A test works on a draft and on an agent that is already active or paused.

Test run appears in the Testing panel once the agent has instructions, a project, a trigger and an action. Before that the panel tells you what is still missing.

• The test runs against a real item from the project, chosen for you. If nothing in the project matches the trigger, the test comes back with no data instead of a result.
• You can test changes before you save them. The agent goes back to its saved settings afterwards.
• Tests appear in the run history alongside real runs, marked as tests.

A test consumes your AI usage limit, the same as a real run. If the account has reached its limit, tests resume when the limit renews.

Activate the agent

Click Activate agent. The agent shows as Active in the list and runs from then on whenever a trigger fires. Pause it at any time and it stops running until someone switches it back on.

Only the person who created an agent can switch it back on after it has been paused. An archived agent cannot be activated again.

Read what an agent did

Every run is recorded. Open an agent, then open a run to see what the agent read, what it decided and what it wrote. Activity log shows runs across all agents on the account.

A run can finish without the agent doing anything. That is a normal result, not a failure, and the run says so: "Your agent reviewed this event and decided no action was needed."

The person who created the agent and the account owner see the full detail of every run. Everyone else sees that the run happened.

When an agent pauses itself

An agent pauses itself in these cases. You get a notification, and the agent stays paused until its creator switches it back on.

• Your account reached its AI usage limit. The agent does not restart by itself when the limit resets.
• Runs kept failing because the content sent to the AI model was too large. Shorten the instructions, or point the agent at smaller forms or fewer documents.
• The agent reached its limit of actions in a single run.
• The agent project was archived or deleted.
• The owner of the agent was set to inactive, lost the AI Agents permission, or lost access to the project.
• Too many runs started in a short time.

Keep in mind

• Agents create tickets, add comments and send emails. They do not change or delete existing tickets, documents or plans.
• The agent decides whether to act. Two runs on similar tickets can end differently.
• Each agent has its own user in your account. Everything it writes appears under that user in the ticket history.

This content was generated with the assistance of artificial intelligence and may contain inaccuracies. Please review and verify the information before use.

Account settings for AI Agents

The Configuration page holds three settings for the whole account, not per user. It needs an in-house user with the user permission Account.

FAQ