A different channel. The same answers.

Questions that arrive are answered by the AI from the same knowledge as LINE and site chat.

Once 24 hours have passed since their last message, only templates approved in Meta go out.

Messages go only to people recorded as willing to receive them — someone who writes to you first is recorded that way. When just “STOP” arrives, nothing more is sent.

Make a wa.me link per entry point, and each contact records which one they came through.

Buttons, confirmations, images, and product or event cards — inside the conversation.

Whatever the channel, they land in the same list.

Every talk that arrives is recorded, sales pitches included. A problem becomes a support case, anything else a lead, and the exchange the AI answered stays with it. Nothing is dropped in silence; what you do with it is a later step.

When asked, the AI answers on the spot from the knowledge you decided to publish. What the knowledge does not cover it does not guess; it hands over to a person. The knowledge it reads is the same one as the website chat, and customer cases or internal learnings never mix in.

Messages go only to people whose consent to receive is recorded, and only when you check the recipients and the message and approve. The exception is a scope you delegated to the AI yourself, with a limit, a budget and an expiry. After 24 hours, only approved templates.

No talk is used to train an AI model. The AI gets better when the knowledge grows, and knowledge grows only when a person adopts something in Knowledge. The learning drafted when a case is closed becomes knowledge once a person approves it, too.