The manager
Every automation the product runs, in one list, pointing at the line of code where it happens. They cannot be deleted, because the behaviour is the product — but each one can be switched or configured.
Every built-in behaviour is listed, with a switch where a switch is allowed and the reason where it is not. Beside the list, a workshop where you build your own from tiles that clip together.
A line drawing of the Automations page. On the left, the list of everything the product does on its own: a count of forty-six, then ruled rows, each saying On or Off at its right. One row is open, and says the switch is the same setting already shown in Settings. Two rows cannot be switched at all: where the switch would be they say Always on, and one is open with the reason written underneath. On the right, the workshop: tiles clip together into a chain — an email arrives, AIRIS is asked a question, an If takes the yes arm, the two arms come back together — and the agent presses Try it once. The run walks the chain and stops at the last tile, book appointment, which turns dashed and waits for a person.
Everything that runs by itself, in one list. A switch where you may have one, the reason where you may not. What you build stops where she stops, and waits for you.
Every automation the product runs, in one list, pointing at the line of code where it happens. They cannot be deleted, because the behaviour is the product — but each one can be switched or configured.
A switch here moves the same setting the Settings page uses. Never two settings for one thing.
A STOP text, a written unsubscribe, a bounce becoming a suppression — no toggle is offered for these. The page shows why instead.
A dotted board and tiles that snap together instead of wires. What you can see is exactly the program; nothing to drag out of alignment.
Nine ways to start; “any of these” fans several triggers into one chain; “back together” rejoins branches; conditions, loops, delays and a full stop.
Two AI tiles carry their own prompt and answer in a shape — yes/no, one of, a number, a list, text — so the tiles beneath can branch on what she said.
Every assistant tool is a tile, marked with whether it asks a person first. An automation runs exactly as trusted as her own turn — a tool that asks still asks.
What this page may claim. Built automations run with the assistant's own permissions and nothing more. There is no outbound-call tile, because there is no outbound calling.
Put it on a real call today — no card, no commitment.