They are not written by the assistant, and nothing about your conversation is sent anywhere to produce them. A short row of one-tap closings appears above the message box when the last message in the thread reads like a sign-off. It is worked out on your own device by a fixed rule.
When the row appears
All of this has to be true, or there is no row at all:
- The last message came from the other person, not from you.
- It is at most 120 characters and at most 20 words.
- It contains no question mark.
- It contains none of the words that mean a thread is still open: but, however, still, waiting, failed, broken, error, problem, refund, urgent, please, any form of cannot, and others like them.
- It reads as a thank you or a sign-off: thanks, got it, understood, that helped, all sorted, it worked, perfect, cheers, take care, and so on.
- You have not already reacted to that message.
Most messages fail at least one of these, so most of the time there is nothing there. That is deliberate. A canned "Glad it helped" underneath an unanswered question is worse than no suggestion at all.
What you get offered
In a direct message with another member: Anytime., Glad it helped. and Talk soon. In a thread with UNI Support the wording is longer, because it is signing a conversation off on the platform's behalf.
Pressing one sends it
There is no draft step and no chance to edit. The chip puts that exact sentence into the box and sends it, so treat it as a send button. If you want to say something of your own, ignore the row and type.
Reacting to the message says much the same thing without adding another message to the thread.