**Can I edit a flow that arrived by directive (you asked your assistant for it over chat)?** Yes — it opens in the visual editor exactly the same as one created by hand or from a template. Only its origin changes, which stays logged with the literal directive that generated it. **Does a deactivated flow stop running entirely?** It stops firing on its own, but you can still launch it by hand with "Run now" — useful for testing it before leaving it running. **What happens if nobody completes a step with no automatic action by hand?** The task (and with it, the flow) stays waiting in "Active" status until someone completes or cancels it — there's no time limit unless the task itself has a due date. **Can I assign a task or a flow to a specific channel?** Not directly: they're always assigned to an **agent** (that's the identity they run under). An "by event" trigger CAN be limited to a specific source channel, if you need it. **What happens if something fails?** If a step has an automatic action and that action fails (for example, a momentary outage connecting to an integration), the system retries it on its own, with nothing for you to do: up to 2 more times, waiting a little longer each time between attempts (about 5 minutes the first time, double the next, never exceeding a 1-hour wait). If after 3 attempts in total it still can't complete, the task moves to **"Failed"** — the exact reason is logged in its **"Step-by-step history"**, and a notice goes out via the bell (always) and also by email if you have that option turned on in Settings → Notifications, to whoever created the task or, if that person isn't available anymore, to the whole team. If the step belonged to a **flow**, its run record is also marked as failed — open it to see exactly which step it stopped at. **Are "Overdue" and "Failed" the same thing?** No. **"Overdue"** is a task you gave a deadline ("Due on") and that nobody got to in time — no action failed, the deadline simply passed without it being completed. **"Failed"** is different: an automatic action was actually attempted, several times, and didn't succeed (previous section). Both are notified the same way — always the bell, email if you have it on — and both stay visible in the task's "Step-by-step history". **What if the system was down for a while and a scheduled task's or flow's time passed?** Nothing is lost. As soon as the system is back up, it automatically checks what was left pending — it checks every few minutes — and fires it anyway, even if the scheduled time has already passed. Nobody needs to relaunch it by hand. **Where do I configure an assistant handing a conversation off to another one?** That's **agent-to-agent handoff**, a separate topic — Manual 56.