AI Judgment
Routes the flow on an AI agent's answer to a yes/no question about the data passing through—"Is this on-brand?", "Does this already exist?". Reach for it when the decision needs reading and judgment rather than a comparison. Routes to true when the agent answers yes, otherwise to false.
Fields
| Field | Description |
|---|---|
| Agent | The agent this judgment runs as—for example, a brand-aware evaluator whose guidelines it should apply. Its model and tools apply too. Leave it blank where the deployment permits, and the judgment runs on the deployment's own AI configuration. |
| Context | Optional framing: the criteria or background the AI should weigh, such as "On-brand means a positive tone, no politics, and lifestyle topics." |
| Question | The yes/no question that decides the route. Accepts variables, so it can ask about the trigger's data and earlier results. |
Overrides
| Field | Description |
|---|---|
| Model Override | Run this judgment on a different Provider Account and Model than its agent's, with an optional Reasoning Effort. Leave unset to inherit. |
| Ignore Agent Instructions | Run this judgment without the agent's standing instructions. The context and question still apply, and so does the classifier framing this step type always imposes. |
| MCP Servers | Tool sources this judgment adds to its agent's—for example, a CMS, so it can check whether matching content already exists. Leave empty to judge from the question alone. |
Context and Question are simply joined into the task the agent is given, so where you put which text is your own organization. The question is what has to be there: a judgment with no question fails at run time.
What later steps see
The incoming payload, passed through unchanged on whichever path was chosen.
The agent's reasoning for its answer is written to the step's execution log, so a branch taken weeks ago is still explainable.
Read-only by design
Deciding a route must not change anything, so this node's agent is only ever offered read-only tools—in a live run as much as in a dry run. Mutating work belongs in an AI Agent action placed after the branch.
Which model it runs on
The same three tiers as the AI Agent action: the step's own Model Override, then the agent's model, then your Default Model, with the same note at the top of the form reporting what the step will actually run on. See AI agents and models.