Go from "what I want to evaluate" to a working judge in one sentence. You can now create an AI judge in AgentControl by describing your evaluation goal in plain language, for example "is the answer grounded in the provided context and free of PII." AgentControl selects the best-fit judge template and drafts a customized scoring rubric for you.

Writing a good judge from scratch is one of the hardest parts of getting started with evaluations: you need to know which templates exist, which criteria matter, and how to phrase a rubric. AI-assisted creation removes that barrier, so your team can start measuring AI quality in production sooner.

You stay in control of the result. Review and edit the template, the rubric, and the pass/fail threshold before saving, then attach the judge to your configs for online or offline evaluations. As with any judge, review a generated one before relying on it in production.

AI-assisted judge creation is available for all accounts eligible for AI-powered features, and the manual judge creator remains available to everyone. Learn more about judges and online evaluations.