Auto-reply rules explained: who the bot answers and when it steps back

This tab is about when the bot speaks, as opposed to what it says. Five settings, and the first two matter far more than most people expect.

When you reply yourself

If you or a colleague answers a customer from the phone, the bot notices and stops answering that conversation. This setting decides how long it stays out.

Waits two hours, then picks it up again is the default, and it fits a team that answers from the phone through the day. The clock restarts every time somebody replies, so a conversation you are actively working stays yours for as long as you are working it. Two hours after the last human message, the bot resumes.

Waits a day suits threads a person normally finishes, where you still want the bot back eventually if the customer returns the next week.

Stays out until I let it back in never answers that conversation again until you press the button on the thread. It is the safest option and it has a cost: every thread anyone touches becomes a thread the bot has permanently stopped handling. On a busy number that quietly turns the bot off, one conversation at a time.

The reason this exists at all is that the alternative is worse than either extreme. A bot replying two seconds after a human reply, contradicting them, is the failure customers remember.

Who the bot answers

New conversations only is the recommended setting and the default. Chats that were already on the phone when you connected stay yours. The bot answers people who message for the first time after connection.

Everyone hands the bot every chat on the number.

Start narrow. The reason is not that the bot is bad at old threads, it is that you cannot predict what is in them. Long-running conversations with suppliers, staff and regulars all live on a business number, and none of them benefit from an assistant joining halfway through.

Widen it once you have read twenty or so conversations and stopped being surprised.

Voice replies

Whether the bot answers with a voice note instead of text. This is a Pro feature and it is off by default even there.

Worth separating two things that sound alike: understanding voice notes that customers send is included from Starter upward, because transcription is part of reading the message. Sending voice notes back is the Pro part.

Off is the right default for most businesses. A synthetic voice note is more noticeable than synthetic text, and the customers who send voice notes are usually happy to read a reply.

Pause the bot

Stops the bot replying for a set period while the number stays connected. Messages keep arriving and you answer them yourself.

This is the correct response to almost any incident: a price went out wrong, a campaign misfired, something is being discussed that you want a person on. Pausing is instant and reversible.

Disconnecting is not the same thing and is usually the wrong instinct. It unlinks the device, which means a rescan to come back and, on some accounts, the history sync running again.

Sending limits

Two caps: per hour and per day, across everything the number sends.

This is the setting that protects the number, and it is deliberately separate from the pace setting inside a campaign. A campaign's pace controls that campaign. This controls the number: campaign messages, ordinary replies and order notifications added together.

Defaults are 30 an hour and 200 a day, which is the range a linked device sustains without drawing attention. Raise them only after a number has been sending at the default for several weeks without trouble, and raise them a little at a time.

The failure mode worth understanding: a campaign running at its own safe pace, plus a busy afternoon of replies, plus order notifications from a promotion, can add up to a volume none of the three would have produced alone. This cap is what notices.

What to do next

Once the bot's timing is right, the next tab worth a visit is Business, which holds the opening hours that decide when an away message answers instead of the bot.