Configuring an Agent

Putting an AI Agent on a stage is a configuration change within the workflow builder, not a separate system to set up.

Step 1: set the stage's assignment type to AI Agent

Open the stage's configuration panel (see Building a Workflow) and change the assignee type from a human role to "AI Agent." This replaces the role/skill/roster filtering chain entirely — an AI Agent stage always routes to the agent, not through the human assignment engine.

Step 2: choose an agent type

Select which of the four available agent types (see Agent Types) fits the stage's purpose — Script Drafter for a drafting stage, Reviewer Agent for a review gate, and so on. The agent type determines both what input it expects and what output shape it produces.

Step 3: configure its context

Each agent type is automatically given the ticket's own fields (Subject, custom fields like language and market) as context. For agent types that benefit from more, you can also configure it to receive the outputs of specific prior stages — a Reviewer Agent, for instance, is typically configured to read the Script stage's submitted text as its primary input.

Step 4: how output fills the completion form

When the stage activates, the agent runs automatically, and its output is mapped directly onto the same completion form fields a human assignee would fill in — a Script Drafter's generated text lands in the script_text field exactly as if a writer had typed it there. The stage then completes and routes onward using the same on_submit/on_approve/on_reject rules as any other stage.

For a description of what each specific agent type expects and produces, continue to Agent Types.