That sounds more like a Parent-Child relationship than Owner-Pet to me. I guess I take inspiration from the Culture series where AIs indeed have basically all the agency and "run" things but humans are not relegated to the status of pets.

>But who's going to be on the other side, arguing that listening to the AI advice is a net bad?

If an AI reached the point where it is objectively better than the human, across all intellectual metrics, to which it is giving advice would it be a net bad?