Everything about how the AI answers your clients is on one page: open AI in the menu, under Who answers. What each library knows is managed next to it, in AI › Libraries.
- Overview — Right now says on how many channels the AI replies, suggests or stays off, with the Pause everywhere switch. This month shows the AI replies used from your plan. To do lists new things to check and questions it couldn’t answer. Your AI names the model that answers now. A new workspace sees three set-up steps here instead.
- Channels — Where it replies: for each channel, the library it reads and Off · Suggests · Replies. How it sounds: the voice, and a different voice for one channel.
- Knowledge — What it knows: your libraries, how much each knows and how many channels use it, and New library.
- Models & keys — the AI providers, the model in use, your own API keys, and If the model cannot answer.
- Runs on its own — Sort new messages, Read the tone and Summarize conversations, then Keeps learning from new conversations and Reads your websites again, for every library at once.
- Limits — Safety (always on), the Monthly limit and five more limits.
- Test — Try it: ask as a client on one of your channels. Nothing is sent to anyone.
Good to know
- Each tab has its own address (for example /app/ai?tab=models), so you can bookmark it, and Back moves between tabs.
- The owner or an admin makes the changes here.
- The page also works in a phone’s browser, and the phone app has its own AI assistant screen.