When does the AI send a message without asking me first?

This is on the Pro plan and up.

Every new kind of message starts with your approval

When a journey job has the AI write a message to a customer, or send them one of your forms, the AI prepares it and waits. You see it in the job's AI draft, and nothing reaches the customer until someone on your team presses Approve. Follow-up messages work the same way: each one waits on the customer's Follow-ups card until someone presses "Approve and send". That is how every kind of message starts, in every business.

How the AI earns it

Each time you approve or turn down one of these, Peakify counts it for that kind of message. For a follow-up, "Skip this one" turns it down.

• Approved as written, or with a small fix: it counts toward the run. A small fix is a few words, a comma or a capital letter.
• A changed number or price: changing any number (a time, a date, an amount), or adding or taking away a price, a percent sign or an email address, does not count, however short the change. The run starts again from zero.
• A bigger rewrite, or turned down: the run starts again from zero. Pressing Discard on a draft nobody has edited turns it down too, and so does pressing Redo on a message draft.

Once you have approved 20 in a row without real changes, covering at least 5 different customers, that kind of message has earned it.

Saying yes, and which jobs it affects

Only an admin is asked. Right after the 20th approval, the draft (or the customer's Follow-ups card, for follow-ups) shows "Let the AI send them on its own?". For journey messages and forms, the jobs it affects are listed:

1. Jobs set to AI does it that use this kind of message are listed under "These jobs will start sending on their own". They start sending without asking as soon as you say yes.
2. Jobs set to AI drafts, person checks are listed with a tick box, "Also let this one send on its own", which starts unticked. They keep waiting for you unless you tick them.
3. Press Let it send on its own to switch it on, or Keep checking first to carry on approving. If you keep checking, you are asked again only after another 20 in a row.

Nothing switches on by itself. You can also say yes later from the AI Assistant page, in the card "What the AI sends on its own".

Turning it on before it has earned it

An admin does not have to wait for the 20. On the AI Assistant page, the card "What the AI sends on its own" lists each kind of message with how far it has got, for example "6 of 20 approvals so far, from 2 of 5 customers".

1. Press Turn on anyway next to the kind of message.
2. Read the warning. It says how many approvals the AI has had, and that the safety checks below still apply.
3. Press Turn on anyway again to switch it on, or Keep checking first to leave it as it is.

The card then says "Turned on before it had earned it". Every safety check works exactly the same: one in ten still comes to you, and one message turned down or changed a lot puts it straight back to checking with you. Jobs set to AI does it that use this kind of message start sending on their own too. Only an admin sees this button.

The one in ten check

While a kind of message sends on its own, every tenth one still waits for you to check, the same way it did before. Approve it and the AI carries on.

How it goes back to checking with you

• You turn one down, or change one a lot: if that happens to a message the AI held back for you, it goes straight back to checking with you first. The card on the AI Assistant page says why.
• You switch it back: on the AI Assistant page, in "What the AI sends on its own", press Check with me first. It takes effect straight away.

What never sends on its own

Anything that moves money, signs a contract, or touches legal or health matters always waits for a person, every time, however many you have approved. Only the kinds of message shown on the AI Assistant card can ever send on their own.