There is a meaningful difference between using a tool to edit a sentence and allowing a system to carry on a conversation independently. The person receiving the message may care about that difference.
A useful disclosure should explain the role AI is playing in language people can understand. ‘I used an assistant to help draft this, and I reviewed it’ describes a different process from ‘An automated assistant is answering routine questions while I’m away.’ Use either statement only when it is true.
The amount and placement of explanation should fit the situation. A confusing technical label is not helpful if the recipient still cannot tell who made the decision or whether the owner is currently present.
Personal and consequential conversations may require more direct human participation. A system should not rely on familiar wording to imply attention or agreement that the owner has not provided.
For product design, ask what the recipient would reasonably believe from the interaction. Can they tell when they are speaking to the person, receiving a reviewed draft or interacting with automation?
Transparency gives people information that may affect how they choose to participate. It also helps keep responsibility visible, so a natural-sounding response does not quietly blur who is speaking and what has actually been authorized.