By the Onsites AI team · Last updated · 7-minute read
In Japan, Taiwan and Thailand, LINE is not one channel among several — it is where customers expect service to happen. The failure mode is familiar: an Official Account whose chats reach one staffer's phone, so coverage dies at 19:00 and every answer lives in one person's head. The fix is routing, not new habits: your existing account connects to a shared desk, and every chat joins the same queue as website chat and email. This guide wires that up in about ten minutes.
A free workspace (register here), the LINE Official Account you already run, and ten minutes. Customers keep using the account they already follow; nothing changes on their side. You get the change underneath: their messages now land in a queue your whole team works, with full history attached.
In your workspace open Settings → Channels → LINE and link your LINE Official Account. From the moment the link is live, customer messages arrive in the same shared inbox as website chat, email, WhatsApp, WeChat, Telegram and Messenger. Your team answers from one screen and one skill set; LINE coverage stops depending on whose phone got the notification.
Set working hours and a first-response SLA the same day so the queue escalates honestly at the breach instead of after it. Japanese customers are notably tolerant of a stated wait and notably unforgiving of a silent one — an honest window is the single highest-value setting here.
Two tag axes earn their keep immediately: language (ja, zh-tw, th, en) and intent (order-status, refund, shipping-delay). The language tag tells the next teammate which voice to answer in before they open the thread; the intent tag becomes your analytics. Both take one second to add while replying.
Internal facts — "this customer's shipment left the Osaka depot Tuesday" — go into private notes the customer never sees. Reusable answers live as shared drafts so the ninth reply is an edit, not a retyping. This is the same playbook as every other channel, which is the point: one skill set, seven channels.
The translate action costs 3 credits per use. The pattern that works: read the incoming message (translated if needed), draft your reply in your strongest language, translate it, then give it one human pass for register. Direct output of any translator is grammatical and somehow off; the five-second human pass is what separates "translated" from "cared for."
Conversations attach themselves to customer records automatically. A LINE thread about order status shows the same account panel — contacts, open opportunities, past conversations — as an email thread would. Regional repeat customers stop being a name in a phone and become a record with history. The free CRM guide covers building out that record set.
The desk itself is free on the cloud plan with 3 seats and 100 MB forever; seats beyond the third are $15/seat/month. AI actions burn prepaid credits only when used — translations 3, drafts 3, suggested replies 6 — packs from $10 with a soft spend limit on by default, so a month of heavy assistance still reads as a single modest line. See the pricing page for the full table. One governance rule covers the channel completely, which is rare: because every action's price is published and fixed, the monthly AI bill is arithmetic, never archaeology — and the balance that goes unspent simply waits. If data residency rules shape how you may serve Japanese customers, the same desk runs self-hosted inside your own network, fully offline-capable with your own model.
Two numbers are enough for the first quarter. First-response time: filter the queue by the tags you set, count from message-in to first human reply, compare week over week. Intent mix: your intent tags, counted monthly. When an intent cluster grows — say shipping-delay doubles in a month — that is a product or logistics problem caught upstream by a one-second tagging habit instead of a meeting. Everything else the suite can produce (visitor tracking, CRM reports) waits until you want it; a LINE desk that keeps those two numbers honest is already run better than most desks running five dashboards.
The pleasant property of wiring LINE into a shared desk first is that channel number two is a card, not a project. The website widget is a snippet paste; WhatsApp is a number link; WeChat is an account link — and every one of them lands in the same queue your team learned on. No retraining, no second tool, no per-channel etiquette to rediscover from scratch. Companies rarely need all seven channels on day one, and they never need to re-teach the team to add one. Start where your customers actually speak; let the queue prove itself; extend per market as the orders arrive.
Do I need a new LINE account?
No — you link the Official Account you already operate, and replies go out under the same identity.
Can multiple teammates answer LINE at once?
Yes. One shared queue: anyone picks up any thread, with private notes and shared drafts carrying the context.
Can we reply in Japanese if nobody speaks it?
Yes, via the translate action (3 credits per use) plus one human pass for tone — that pass is what keeps it natural.
Where do LINE chats count against storage?
Chats and emails count, but text is negligible; attachments are what add up. Beyond the free 100 MB it is $1 per GB per month.