AI Agents¶
To help your agents focusing on more important tasks, you can create AI agents which can take over routine tasks. These AI agents are managed in Zammad’s admin settings under AI > AI Agents. Make sure to configure an AI provider before activating the feature. Otherwise, a warning message will tell you to do so.
While an AI agent works on a ticket, an avatar is displayed in the live user section in the bottom bar of the ticket detail view, see example below. In such a short period of time, the agents should not apply changes to the ticket because they could get overwritten.
Avatar of an AI agent which shows up in the live user section.¶
Important Information¶
To configure AI agents, the
admin.ai_agentpermission is requiredAI agents don’t run on their own. They have to be invoked by a trigger, a scheduler job or a macro with the action AI > AI Agent.
By creating a macro with an AI agent action, your agents can run it on demand.
If an AI agent applies changes to a ticket, you can find an entry in the ticket history with the name of the AI agent.
Manage AI Agents¶
Create an AI agent by clicking the New AI Agent button in the top right corner. This opens a wizard where you can choose the type of the agent and additional settings based on the selected type, see AI Agent Types for details. The action column with the ︙ button allows you to clone or delete an AI agent. You can only delete AI agents which are not used in triggers, scheduler jobs and macros. Check the Used in column and hover over an entry to see where an AI agent is used.
AI Agent Types¶
Define your AI agents and use them afterwards in triggers, scheduler jobs or macros. An example for a trigger configuration could look like this:
Activated by: Action
Action execution: Selective
Conditions for affected objects: Action is
createdExecute changes on objects: AI Agent > Select your AI agent
Hint
Depending on the agent type, you can limit the available options for the AI agents. If you do so, make sure to select at least two options the agent can choose from.
Ticket Categorizer¶
This AI agent is capable of assigning categories to tickets. Because there is no built in category field in Zammad, you first have to create a custom ticket attribute in which you want to track the ticket’s category. The field type must be one of the select types.
The ticket categorizer AI agent uses the content of all articles and the title to compare it with your category names (and optionally descriptions). To add descriptions for your categories, limit the categories the AI agent can choose from. Then add the categories which should be available for the AI agent and optionally provide a description for them. Additionally, you can decide in the wizard if multiple values are allowed (only if category field is a multi-select field type).
Ticket Group Dispatcher¶
This AI agent is capable of assigning tickets to different groups. It compares the content of the ticket in question with your group names (and optionally descriptions). To add descriptions for your groups, limit the groups the AI agent can choose from. Then add the groups which should be available for the AI agent and optionally provide a description for them.
Ticket Prioritizer¶
This AI agent is capable of setting a ticket priority based on the title of the ticket and the content of the last article. You can choose to use all priorities or limit them to specific ones. In case you limit the priorities, you can also provide a description for each priority the AI agent can choose from.
Ticket Title Rewriter¶
This AI agent is capable of rephrasing the ticket title based on the content of the last article. The title is not considered by the agent and gets completely rewritten. This can be useful if your customers often send requests with unclear subjects or if you have a specific channel which only creates tickets with generic titles (e.g. from a web form).