When an AI agent decides a chat needs a person, LINE and Facebook’s platforms don’t decide who that person is. Meta’s own documentation says that without routing, any connected app can answer the same message; LINE’s permissions manage logins, not who owns a chat. A second staff member turns that gap into a duplicate reply, or an unanswered case.
What LINE and Meta actually hand you
LINE’s Official Account Manager help center documents exactly one way to split work across a team: Administrator and Operator permissions, added from Settings > Permissions > Add, where “only an existing administrator can add administrators” and “operators cannot add administrators” (LINE Official Account help center). That page governs who can log into the account and change its settings — it says nothing about which operator a given chat belongs to. Every operator who’s logged in sees the same open conversation.
Meta’s Messenger Platform describes the same gap from the other side, for a business running more than one connected app or inbox on a Page. Its own documentation on Conversation Routing is explicit about what happens without it: “all applications can respond to the same user message without restrictions,” and left alone, “coordination of responses from both the application and page inbox is necessary” (Meta for Developers: Conversation Routing) — meaning the platform won’t stop two people, or a bot and a person, from typing a reply to the same customer at once. Neither platform’s native tools were built to answer “whose case is this,” which is exactly the question that matters the second an AI agent decides to hand a case over in the first place.
Three ways to decide who catches it
1CS’s escalation strategy picks one rule and applies it to every case an AI agent hands off: unassign leaves the case open with nobody attached, for a supervisor or the first free agent to claim; specific agent always routes to one named person, for a channel or topic one person owns; round robin cycles the case to the next agent on the account, for a team that answers evenly. As a recommended default rather than a platform requirement: start with unassign if one person already triages everything by eye, move to a specific agent the moment a channel or product line clearly belongs to someone, and reach for round robin only once the team is large enough that an even split actually matters more than who’s free right now.
Set up in 1CS
Open AI > AI Agent, then the Escalation & handoff tab, and write the conditions that should hand a case to a person — a concrete event like “customer asks for a refund” or “customer asks for a human,” not a broad category (writing clear escalation rules). Below that, pick one escalation strategy: Unassign (the default — the case sits open for the team or a supervisor to claim), Specific agent (always the same named person), or Round robin (cycles through every agent on the account). Turning on “Send a message to the customer when escalating” lets the AI say a person is checking in before it goes quiet, which matters more the longer a routing strategy leaves a case waiting. The full click-by-click walkthrough is in Thai at เขียน Escalation Rules ให้ชัด; the underlying AI Agent capability — including this same choice of unassign, assign, or round robin — is covered in English on the AI Agents feature page.
Something to copy: a claim note for the team
Whichever strategy is live, paste this into the team channel or chat note the moment someone opens a handed-off case, so a second person doesn’t start typing the same reply:
Claimed: [case/customer name] by [your name] at [time]. Give me [5] min — if I haven’t replied by then, it’s back up for grabs.
It only works if releasing it is as automatic as claiming it — a case nobody is actually working should never sit marked “claimed” for an hour.
