Buyer's Guide

LINE OA Chatbots: Built-in, LINE's AI Chatbot, or Your Own

For shops and companies whose admins answer the same chat questions all day. The three options, what each really costs, and what to design before letting AI talk to customers.

Short answer

A LINE OA chatbot comes in three levels: free keyword auto-reply, LINE's AI chatbot (Beta) in the OA Chat package at THB 555 a month plus VAT, or a custom AI chatbot on the Messaging API that answers from live stock, prices and orders. Bot replies don't use the message quota, so cost sits in the package, model and build.

Key points

  • Keyword auto-reply is free and built in; it suits questions with one fixed answer, like opening hours or bank details.
  • LINE's AI chatbot (Beta) comes with OA Chat at THB 555 a month plus VAT and answers from Q&A sets you prepare.
  • Only a bot connected to your own systems can tell one customer where their order is or whether an item is in stock.
  • Prices and stock must always come from live data, and customers must be able to reach a person at every step.
  • Start from one to three months of chat history, count the frequent questions, and launch one question group at a time.

What kinds of LINE OA chatbot are there?

There are three. The first is keyword auto-reply in LINE Official Account Manager: set a keyword and a reply, and a customer who types the keyword gets that reply. It suits questions with one fixed answer, such as opening hours, parking or bank details.

The second is LINE's AI chatbot (Beta). According to the OA Chat package page on LINE for Business (checked 27 September 2026), the package is THB 555 a month plus 7% VAT. The bot picks answers from Q&A sets you build, and can generate those sets from PDFs or images. LINE notes that some account types cannot use it.

The third is a custom AI chatbot built on LINE's Messaging API and connected to your own data: stock, bookings or CRM. It can answer whether a model is in stock, where an order is, or which Saturday slots are free. The first two cannot, because they never see your live data.

Three levels of LINE OA chatbot
Keyword auto-reply
Free
LINE's AI chatbot in OA Chat (before VAT)
THB 555/mo
Custom AI chatbot on the Messaging API
Live data

Chatbot vs AI chatbot: what is the difference?

A rule-based chatbot follows the conditions you set: if the customer's wording doesn't match, it can't answer, but it never goes off-script. An AI chatbot uses a language model to understand questions however they are phrased and replies more naturally, but unless it is held to approved information it can make answers up.

Most shops do best combining the two. Anything that must be right every time, such as prices, promotions and return terms, comes straight from data the shop controls; the AI's job is to understand the question and phrase the reply.

Rule-based bot vs an AI chatbot grounded in your data

Rule-based (keywords)

  • Answers only when the customer types the keyword
  • The same question in other words gets no answer
  • Cannot see any one customer's data

AI chatbot with controlled sources

  • Understands varied wording and casual language
  • Prices and terms come from data the shop sets
  • Answers order or booking questions once connected

How much does a LINE OA chatbot cost?

The cost is in the package, the AI model and the build, not in message fees. LINE's developer documentation says replies to messages a customer sends do not count toward the monthly message quota. LINE package prices below exclude VAT.

Three chatbot options on LINE OA (OA Chat price from lineforbusiness.com, 27 September 2026)
Built-in auto-replyLINE AI chatbotCustom AI chatbot
Main costFreeOA Chat package, THB 555/monthBuild, hosting and model usage by message volume
Answers fromKeywords you setQ&A sets you prepare (can be generated from PDFs or images)Your live systems plus approved documents
Customer-specific answers (e.g. where is my order)NoNoYes, if connected to orders or CRM
Handover to a personCustomer keeps chatting with the adminPer LINE's featuresYour own handover rules, with a summary sent to the admin
Who maintains itShop adminShop admin updates the Q&A setsDevelopers run the system, the shop owns the content

How do you stop an AI chatbot making things up?

Limit what the bot may answer from instead of letting the model answer from general knowledge. A good design retrieves the relevant facts from your documents or database first, lets the model phrase an answer from those facts only, and, if nothing is found, says it isn't sure and hands over to an admin.

  • A list of topics the bot may answer, and topics that always go to a person: complaints, refunds, price negotiation
  • Which sources the bot may read, and who keeps them up to date
  • The handover message, and when the admin will reply
  • Transcripts kept and reviewed for accuracy, with unanswered questions fed back into the data
  • A model and effort level sized to answering questions, so you don't overpay for the model
How a bot grounded in your data answers
  1. Search your dataRetrieve the relevant facts from your documents or database.
  2. Phrase the answerThe model answers from those facts only.
  3. Admit gapsIf nothing is found, say so instead of guessing.
  4. Hand overPass the chat to an admin, then add the question to the data.

What are the risks of a customer chatbot?

The most expensive risk is the bot making promises for you: a wrong price, confirming stock that has sold out, or agreeing to something your policy doesn't allow. The other is customers getting stuck with a bot that won't let them reach a person.

The fix: prices and stock always come from the live system, never from the model's memory; there is a way to reach an admin at every step; and the bot is tested against real questions from past chats before customers see it.

What does PDPA mean for an AI chatbot?

You need to know how long your AI model provider keeps chat messages and whether it trains on them, because chats often contain names, phone numbers, addresses or health details. Strip what the model doesn't need before sending it, and say in your privacy notice that chats may be answered with automated help.

How to set up a chatbot on LINE OA

Start from the data you already have: your chat history. Count which questions come up most, then pick the kind of bot those questions need. Fixed-answer questions go to LINE's built-in features first; questions that depend on one customer's data become a connected system later.

Five steps from real chats to a live bot
  1. Pull chat historyOne to three months: group the questions and count them.
  2. Fixed answers firstUse auto-reply or LINE's AI chatbot.
  3. Connect data laterCustomer-specific questions get a system connected to your data.
  4. Test on real questionsMeasure the share answered correctly and how often it hands over.
  5. Launch in groupsSwitch on one question group at a time, not everything at once.

Before a chatbot talks to customers

  • Do you know the most frequent questions, from real chat history?
  • Do prices, promotions and stock come from live data, not the bot's memory?
  • Is there a list of topics that always go to a person?
  • Can customers reach a person at every step?
  • Is someone responsible for updating what the bot answers from?
  • Do you know which providers see customer chat data?
  • Has the bot been tested on real questions before launch?

Frequently asked questions

Is a LINE OA chatbot free?

Keyword auto-reply is free and built into every LINE OA. LINE's AI chatbot (Beta) comes with the OA Chat package at THB 555 a month plus VAT, per LINE for Business on 27 September 2026. Bot replies to messages customers send do not use the monthly message quota, so a busy bot does not run up message fees.

Can I connect ChatGPT or another AI model to LINE?

Yes, through the LINE Messaging API and a small service that calls the model. Don't let it answer from general knowledge alone: have it search your own data first, answer from that, and hand over to an admin when unsure. Model cost scales with message volume, so estimate it from your real monthly chats.

How well do AI chatbots handle Thai?

Current models understand casual Thai and misspellings far better than keyword bots. Shop-specific words, such as product model names, course names or nicknames customers use, must be in the data the bot reads, or it will guess. The reliable test is to run it against real questions from your own chat history.

What are the downsides of a chatbot?

The main one is a wrong answer the customer believes, especially on prices and stock; the other is frustration when customers cannot reach a person. Both are fixable: pull key facts from live systems, put a route to an admin at every step, and review transcripts regularly.

Can a LINE chatbot check a customer's order status?

Yes, but only a custom AI chatbot built on the Messaging API and connected to your order system or CRM. Keyword auto-reply and LINE's AI chatbot answer only from replies you set in advance, so they never see an individual customer's data.

Start with the problem

Which chatbot fits your chats?

We read your chat history, sort which questions LINE's built-in features can handle and which need your data, and propose the cheapest setup that works. Our engineers work with AI assistants and own the result.

Your business,in professional hands.