What is LINE chatbot development?
LINE chatbot development is building the software that answers messages sent to your LINE Official Account automatically. When a customer writes to your account, LINE sends a webhook event to a server you control; that server decides what to reply and sends it back through the Messaging API.
LY Corporation’s Messaging API overview describes this loop: the platform delivers events to your webhook URL, and your bot responds with text, stickers, images, templates or Flex Messages, the card-style layouts used for receipts, menus and product lists. Users can also tap a rich menu, the panel of buttons at the bottom of the chat.
The “bot” is therefore not a product you switch on but a small application. It can be as simple as a set of rules (“if the user taps Opening hours, send this card”) or as capable as an AI assistant that reads your manuals and looks up a customer’s order. The difference between a helpful bot and an irritating one lies almost entirely in how that application is designed.
LINE’s built-in auto-replies in LINE Official Account Manager handle keyword responses without code. Custom LINE chatbot development starts where those stop: when answers depend on who the customer is, what they bought, or what your documents say.
What can a LINE chatbot do for a Japanese business?
A LINE chatbot can answer repeat questions, look up live information, collect details before a human steps in, and route people to the right place. The best bots do two or three of those well rather than trying to replace your whole support team.
Answer repeat questions
Hours, prices, parking, returns, delivery areas and product details, answered instantly from approved content.
Look things up
Order status, reservation times, point balances or stock levels, read live from your systems after the customer is identified.
Take actions
Book, change or cancel an appointment, request a callback, or register for an event without a phone call.
Qualify and route
Ask two or three questions and send the chat to the right staff member with a summary, so nobody asks the customer twice.
Nudge at the right time
Segmented follow-ups after a purchase or a missed booking, sent only to the customers they concern.
Clinics, salons, online shops, real estate agents, schools and tour operators all use combinations of these. If you run a site in Japan already, our clinic website and real estate website guides show where a LINE bot fits alongside it.
Rule-based, AI or hybrid: which kind of LINE bot do you need?
Use rules for anything with a fixed answer or a transaction, AI for open questions answered from your documents, and a hybrid for most real businesses. Pure AI bots are the wrong choice for bookings, payments and anything with legal weight.
A rule-based bot follows buttons and keywords. It is predictable, cheap to run and easy to test, and it is what you want for “book an appointment” or “check my order”, where the steps must happen in a fixed order and nothing should be improvised.
An AI bot uses a large language model to understand free-text questions such as “can I bring my dog to the terrace?” and to compose an answer. Left alone, a language model will guess; a well-built one only answers from content you approved and says when it does not know.
The hybrid is what most LINE chatbot development projects end up with: a rich menu and rule-based flows for the common tasks, AI for the long tail of questions, and handoff to staff for anything sensitive. Decision rule: if a wrong answer would cost you money or trust, that path is a rule or a human, not the model.
How does an AI LINE chatbot answer from your own FAQ?
It uses retrieval-augmented generation (RAG): the bot first searches your approved documents for passages that match the question, then asks the language model to answer using only those passages. The model supplies the wording; your documents supply the facts.
In AI-based LINE chatbot development we start by collecting your FAQ pages, product sheets, policies and staff manuals, removing contradictions and out-of-date versions, and splitting them into small sections with titles. Each section is converted into a searchable index stored in your cloud account. When a question arrives, the bot retrieves the closest sections, checks that they are relevant enough, and only then drafts a reply.
Three safeguards matter. A relevance threshold stops the bot answering when nothing in your content fits. Instructions tell the model not to invent prices, dates or policies. And every answer is logged with the passages it used, so your team can see why the bot said what it said and fix the source rather than the symptom.
The language model itself comes from a provider you choose, on an API account in your name, so usage is billed to you and governed by your agreement with them. We select settings that do not allow the provider to train on your data where the provider offers that option. For a deeper look at the method, see RAG chatbot development.
Can a LINE chatbot answer in both Japanese and English?
Yes. The bot detects the language of each message and replies in the same one, drawing on Japanese content for Japanese questions and English content, or a translated answer, for English ones.
Tone matters in Japanese. A clinic bot and a streetwear shop bot should not sound alike, and the right level of politeness is a brand decision. We work in English, so we ask you to supply or approve sample Japanese replies; the bot’s instructions then reference those examples to keep the register consistent. Your staff review a batch of real answers before launch.
For English, the usual audiences are foreign residents, inbound tourists and overseas buyers. Some businesses keep separate English content; others let the model translate from the Japanese source and flag those answers as translated. Either works, but prices, legal terms and medical information should always come from approved text in the right language rather than a live translation.
Menus can be bilingual as well, and bilingual LINE chatbot development usually covers them in the same build. We often give English speakers their own rich menu, switched automatically once the bot knows their preference, so buttons match the language of the conversation.
A rich menu should hold the four to six things customers want most, in plain labels, with the most common task in the largest tile. It is often used more than typed messages, so it deserves as much design effort as the bot.
The Messaging API lets you set rich menus per user, so the menu can change with the customer’s situation: new friends see “First visit guide” and “Book”; members see “Member card” and “My bookings”; customers with an open order see “Track my order”. The bot switches menus when something changes, such as a completed sign-up.
Tabbed menus give you more buttons without cramming, and each tap can be tracked, which tells you what people actually want. We also connect menu buttons to a LINE MINI App when one exists, so “Member card” opens the card instantly. For that side of the build, see LINE MINI App development.
Segmented broadcasts: sending the right message to the right followers
Instead of broadcasting every message to every friend, the Messaging API’s narrowcast lets you target audiences and demographics, which keeps block rates down and makes each message count against your quota only for the people who receive it.
LY Corporation’s documentation lists the targeting options: audiences built from user IDs, message clicks or impressions, chat tags and rich menu interactions, plus demographic filters such as gender, age, OS, region and how long someone has been a friend, combined with AND, OR and NOT logic. It also notes that sent messages are counted by the number of people who receive them, not by how many message bubbles you include.
Custom LINE chatbot development adds your own data to that. When the bot knows a customer’s LINE user ID is linked to their account, you can build audiences such as “bought within 60 days but not reviewed” or “has points expiring this month”. We build the audience logic and a simple screen for your marketing team to schedule the sends.
LINE Official Account plans vs build cost: what will the messages cost you?
Your LINE plan and your build are separate costs. The plan decides how many messages you can send each month for free; the build is a one-time project. A well-designed bot relies on replies, which LINE does not count, and uses paid sends only for targeted campaigns.
According to LY Corporation’s plan page, LINE Official Account has three plans: Communication with 200 free messages a month, Light with 5,000, and Standard with 30,000. Only the Standard plan can send additional messages beyond the free allowance. The same page lists reply messages, including replies sent through the Messaging API’s reply endpoint, among messages that are not counted.
That shapes LINE chatbot development from the first flow map. When a customer asks something, the bot answers with a reply and costs nothing against your allowance. Push messages, multicasts, narrowcasts and broadcasts do count, per recipient. So a bot that answers questions and a campaign strategy that targets small segments can run on a modest plan, while a bot that pushes a message to every follower after every event will burn through it.
We check your expected volumes against the plans during scoping and design the flows accordingly. Choosing and paying for the plan stays with you.
How much does LINE chatbot development cost?
With BtechWaleTech, an AI LINE chatbot with a rich menu and staff handoff starts at US$600, typically over two to four weeks. Bots embedded in larger booking, ecommerce or CRM builds start at US$900. The itemised proposal lists each flow and integration so you can remove what you do not need.
Across the market, quotes for LINE chatbot development vary a lot, mostly because “chatbot” covers everything from a keyword menu to a full AI assistant reading live systems. When comparing, ask each developer the same questions: which flows are included, which systems it reads, who owns the AI account and the hosting, and what happens to unanswered questions after launch.
- Number of conversation flows (booking, order status, FAQ, lead capture)
- Systems the bot reads or writes: Shopify, booking engine, CRM, spreadsheet
- Size and state of your knowledge base, and whether it needs cleaning
- One language or both Japanese and English
- Rich menu variants and segment rules
- Admin screens for staff, such as a handoff inbox or audience builder
Running costs come from your providers: the LINE plan, AI model usage, and a small cloud setup. After two months of free maintenance, optional care starts at US$120/mo per month.
Connecting a LINE chatbot to Shopify, booking systems and a CRM
The bot becomes useful when it can see your data, which is why integration is the heart of most LINE chatbot development. We connect it through each system’s API after linking the customer’s LINE user ID to their record, so the bot can say “your order shipped yesterday” rather than “please check your email”.
Account linking comes first. The customer taps a button, signs in to your site or confirms a phone number or email once, and the link is stored. From then on the bot knows who it is talking to. Without linking, the bot can still ask for an order number and a verification detail, which works but adds friction.
For Shopify, the bot can read order and fulfilment status and send shipping updates to linked customers. For booking engines and calendars, it can show open slots and create or change reservations. For CRMs, it can log each conversation to the customer record and create a follow-up task when a lead asks for a quote. For systems without an API, a scheduled spreadsheet or CSV sync is a practical fallback.
One of us builds the integrations; another of us handles the AI layer and the AWS setup. Everything runs in your cloud account with keys stored in its secret manager.
When should a LINE chatbot hand over to a human?
Hand over whenever the customer asks for a person, when the bot’s confidence is low, and always for complaints, refunds, medical or legal questions. A good handoff tells the customer what happens next and gives your staff the whole conversation.
In our LINE chatbot development work, handoff is a clear state. The bot tells the customer “a staff member will reply during business hours” (or right away, if someone is on shift), stops answering automatically, alerts staff, and hands over a short summary of what was asked. When staff close the case, the bot resumes.
Staff can reply from LINE Official Account Manager’s chat screen or from a small inbox we build that shows the bot’s history and the customer’s linked record. Which one suits you depends on volume and whether several people share the work.
For slow lookups, the Messaging API can show a loading animation in one-to-one chats for between five and sixty seconds, according to LY Corporation’s documentation, so customers know a reply is on its way while the bot checks an order or drafts an AI answer.
LINE chatbot development timeline in four short sprints
An AI LINE chatbot usually takes two to four weeks; one inside a larger system takes six to twelve. Most of the calendar time goes into content and testing rather than code.
- Sprint 1 — content and flows: collect FAQ and documents, map the rich menu, agree handoff rules, set up the Messaging API channel under your provider
- Sprint 2 — build: webhook server, rule flows, retrieval index, first integration, test account connected to your LINE app
- Sprint 3 — test with staff: your team asks real questions, flags wrong or awkward answers, and approves Japanese tone
- Sprint 4 — launch and tune: switch on for all friends, watch the unanswered-question log daily for the first two weeks, adjust content
Integrations with older systems, or a knowledge base that needs heavy cleaning, are the usual reasons a project runs longer. We flag both during scoping.
Chat logs, personal data and APPI: what a LINE bot must handle carefully
Chat logs contain personal data, sometimes sensitive data, so privacy is part of LINE chatbot development; so decide before launch what you store, for how long, who can read it, and what you send to an AI provider. Your privacy policy should match what the bot actually does.
We store conversation logs in your cloud account in the Tokyo region, encrypted, with staff access by role. Retention is a setting you choose. Before a message goes to the language model, the bot can strip obvious identifiers such as phone numbers, and it never sends payment details. Account-linked lookups happen on your server, not inside the model.
Japan’s Act on the Protection of Personal Information applies to how you collect and use this data, including passing it to an AI provider or to developers overseas. How it applies to your setup is a question for your own lawyer; our part is to build the controls your policy requires and keep our own access to production logs minimal and logged.
How do you know whether your LINE chatbot is working?
Measure how many conversations the bot resolves without staff, how many it hands over, which questions it could not answer, and whether blocks rise after broadcasts. Those four numbers tell you what to fix next.
Every LINE chatbot development project we deliver includes a simple dashboard. Resolved conversations show the bot is earning its keep. The handoff rate, split by reason, shows whether the bot is too cautious or not cautious enough. The unanswered-question log is the most valuable of all: each entry is a missing or unclear piece of content, and adding it improves the bot the following day.
Rich menu taps show which tasks customers care about, and message clicks show which campaigns work. Blocks after a broadcast are an early warning that a segment was too wide. Your Official Account’s own statistics complement the dashboard.
For the first month we review the log with you weekly. After that, many clients fold the review into monthly care, and some keep it in-house with a short guide we write for their team.
Commissioning LINE chatbot development from a team in India
We are three and a half hours behind Japan, so our day overlaps your afternoon from about 12:30 JST. Staff testing and content review happen in your morning; questions, fixes and demos happen in the shared hours.
The project runs in English, with Japanese content supplied or approved by you. A shared spreadsheet of test questions, with the expected answer and the bot’s actual answer side by side, is how most of the collaboration happens; your staff mark each row, and we fix the source.
You create the Messaging API channel under your own LINE provider and add us as admins; the AI provider account and the cloud account are also yours. Proposals are in USD, payable in USD or JPY by Wise or bank wire, with nothing billed before you approve in writing. Invoices come from India.
In the first two weeks you can expect a kickoff call, a written flow map with the rich menu layout, and a test version of the bot on your own LINE app answering the first twenty or thirty FAQ questions. From there it is test, fix and tune.
Common LINE chatbot mistakes, and how to avoid them
Most irritating bots come from rushed LINE chatbot development and share a few faults: they guess, they trap people, or they spam. Each has a straightforward fix.
- Answering from the model’s general knowledge instead of your documents: set a relevance threshold and log sources
- No visible way to reach a person: keep “Talk to staff” on the rich menu and honour it every time
- Pushing every update to every friend: use replies and narrowcasts, and watch block rates
- Asking for an order number that the bot could have looked up after account linking
- Webhook server and AI account in the developer’s name, not yours
- No owner for content updates, so answers drift out of date after a price change
- Launching to all friends on day one without a staff test round
Worked example: a hypothetical Yokohama online shop adds an AI LINE bot
This is an invented scenario to show how a LINE chatbot development project might be scoped, not a client story. Imagine an online kitchenware shop in Yokohama running on Shopify, with a LINE Official Account that sends a weekly broadcast. Staff spend hours answering “where is my order?” and product-care questions, and a growing share of buyers write in English.
The first release would have a four-button rich menu (Track order, Product care, Returns, Talk to staff), account linking by email, live order status from Shopify, and an AI answer layer over the shop’s care guides and return policy in Japanese and English. Refund requests and anything about damaged goods would go straight to staff with a summary. That scope fits the AI automation plan, from US$600, over about three to four weeks.
The weekly broadcast would become two narrowcasts: one for recent buyers about care tips, one for lapsed customers with new arrivals, both counted only for their recipients. A later phase might add a restock alert bot or a LINE MINI App member card, quoted separately.