Response guidelines and guardrails

Last updated: 26 September 2026 · 4 min read

Guidelines decide how the assistant answers; guardrails decide what it won't answer.

Steps

  1. 1

    Response guidelines

    In Clomni AI, open the “Response Guidelines” page. Click “Add this” on the example guidelines that fit, or type your own rule in the “Type in another response guideline...” field and click “Add and save”. For example: “Don't write more than 3 sentences”, “Address the customer formally”.

  2. 2

    Guardrails

    On the “Guardrails” page, add the topics the assistant must not touch, for example: “Don't talk about competitors”, “Don't give medical advice”. You add them the same way.

  3. 3

    Test them

    The “Test all” button in the list checks how the guidelines work. Then try a real question in the Playground too.

  4. 4

    Delete

    Select the guidelines you don't need and click “Delete”.