AI autonomy
Autonomy is how far the AI may go on its own. You set it once here, and you can change it per action type at any time on the Rules page (Workspace → Rules).
What are the settings?
| Setting | What happens |
|---|---|
| Off | The AI does not do that kind of action. |
| Ask me first | The AI proposes an action and you approve it. |
| Do it and tell me | The AI acts, and reports what it did in Activity. |
Which should I choose?
Start wherever you are comfortable - this is not a one-time decision. You can set it per action type, so the AI can act on its own for things you don't want to review while still asking first for anything you do.
How do I see what it did?
Everything the AI does lands in Activity with the reason it did it, whichever setting you choose. On Do it and tell me that log is how you keep track; on Ask me first it is the record of what you approved.
Can I stop everything at once?
Yes. Pause everything on the Rules page stops everything the agents do on their own until you carry on. Your storefront assistant keeps answering shoppers, Maestro still answers you in the chat, and you can still approve what is waiting.
Can I set limits the AI never crosses?
Yes, under Guardrails on the same page: a discount ceiling, products never to recommend, words never to use, the longest product title, the language of product copy, the hours email is sent in and how many paid checks run a month. The agents work inside them, whatever the settings above say.
What's next
That's setup finished. Home is where the results show up.