Telling an AI reply from yours, and taking it back
What the AI looks like from inside a thread. How its messages are marked, the choice between it suggesting and it sending, and how you stop it on one conversation.
Watch
What it does
What is on the screen
AI messages are marked as AI
They sit in the thread like any other message but carry their own label, so nobody on your team mistakes one for something a person wrote.
It can suggest instead of send
In suggesting mode it drafts the reply and waits for someone to press send. In sending mode it answers on its own. You pick which.
You can stop it on one thread
Turning the AI off for a single contact leaves it running for everyone else. There is no all-or-nothing switch to reach for.
How it works
Working alongside it
- Open a thread the AI has been handling. Its replies are labelled.
- Read up from the bottom to see what it has already promised.
- Take over by turning it off for that contact, or by pausing it for a set time.
- Reply yourself. Turn it back on when the conversation is routine again.
Worth knowing
Worth knowing
- Customers usually work out that it is a bot. That is fine if it hands over quickly and dishonest if it pretends otherwise.
- Suggesting mode is safer and slower. If nobody presses send, an overnight message waits until morning anyway.
- It waits a moment before answering, so it can read a customer's three quick messages as one. That pause is deliberate, not a fault.
Related
Related
Conversation AI
What it does across all your text channels, and where it stops.
Answers outside working hours
The overnight case, which is where it earns its keep.
Unified inbox
The list these threads open from.
Read one AI thread
Log in, open a conversation it handled, and read it as a customer would before you widen what it may say.