How much does a LINE bot cost? A 2026 quote broken into six parts
LINE bot prices can differ by 100x: keyword replies are free, LINE's built-in AI costs NT$100 a month, and we quote custom AI customer service from NT$50,000. This article breaks a custom quote into six parts: where the money goes, what you can cut from version one, and what you'll still pay each month after launch.
The numbers first:Everything is called a "LINE bot," but prices range from free to well over NT$100,000.
- Keyword auto-reply: built into LINE Official Account,free
- LINE's built-in AI chatbot: an add-on to the advanced chat plan,NT$100 per month(currently in beta)
- Custom AI customer service (answers from your data, hands off to a human, shared with your website): we quote from NT$50,000, live in 3–5 weeks
- Bots that take orders, dispatch jobs, and collect payments:quoted by scope, after we review your existing systems
Things that differ in price by 100x aren't the same thing. This article first helps you figure out which one you need, then breaks down the quote for the custom option.
First: do you really need custom?
Most people don't. Of the three situations below, the first two don't require hiring anyone:
| Your situation | What's enough | Cost |
|---|---|---|
| Customers ask fewer than ten questions, with fixed answers | Keyword auto-reply + rich menu | Free |
| More questions, but only lookups (hours, prices, menu) | LINE's built-in AI chatbot | NT$100 per month |
| Needs to apply your rules, check live data, hand off to a human, and share one system with your website | Custom | From NT$50,000 |
LINE's built-in AI chatbot has a few limitations you should know about, all stated in LINE's official documentation:
- It's currently in beta,has a monthly reply cap, and LINE recommends it for accounts with fewer than 25,000 friends
- Medical, financial, religious, government, and legal businesses may not be eligible
- Its sources are the product information and FAQs you upload; the official documentation doesn't mention human handoff or integration with external systems
If you run a clinic, or your answers depend on checking inventory and available booking slots, the built-in option can't help. That's when custom makes sense.
A custom quote in six parts
Part 1: Foundation
LINE Messaging API integration, webhook message handling, signature verification, server deployment, and certificates.
Customers never see this part, but it determines whether the bot is stable.Without signature verification, anyone can send spoofed messages to your bot.Unusually cheap quotes often cut corners here.
Part 2: Knowledge base preparation (the most important part)
Organizing your product descriptions, price lists, FAQs, return policies, and past customer service conversations into a format the AI can work with.
This usually takes 1–2 weeks and is the most valuable part of the whole project.How well the AI answers depends about 80% on this step, not on which model you use.Your materials can be in any format: Word, Excel, web pages, even screenshots of LINE chats.
How to test it: ask it real customer questions from the past six months and see what share it gets right. If it fails here, nothing else matters.
Part 3: Answer only from your data, say "I don't know" otherwise
The technical term is RAG: first retrieve the relevant passages from the knowledge base, then have the AI answer based only on those passages.
The real work here is handling "I don't know." An honest AI customer service agent,when faced with a question the knowledge base doesn't cover, should say it doesn't know and hand off to a human, rather than make up a plausible-sounding answer.A made-up answer is a complaint waiting to happen.
Numbers that change, like prices, inventory, and booking slots, should never come from the AI itself. Code fetches the live data, and the AI just phrases it naturally.
Part 4: Human handoff
When a customer asks for a real person, or the AI decides it can't answer, the bot needs to step aside and pass the conversation to you.
The handoffmust carry over the conversation so far, or the customer has to explain everything again, which is worse than having no bot at all. Our approach routes it to your Telegram, where you reply to the customer directly, then hand it back to the bot when you're done.
Part 5: Rich menu and guided flows
Not every customer wants to type. Bottom menus, quick-reply buttons, multi-step flows: if they can tap, don't make them type.
This part is small, but it directly affects whether customers actually use the bot.
Part 6: Usage caps and cost guardrails
AI is billed by usage. Without caps, a message flood or a bug can blow up your bill overnight.
Set a daily cap on calls per person, and hand off to a human once it's reached.This belongs in the system, not in the contract.
Add-ons: pay only if you need them
| Item | When you need it |
|---|---|
| Multi-platform (Telegram, WhatsApp) | You have overseas customers. Same backend, priced by number of platforms |
| Subscription payments | The bot itself is a paid product |
| Booking flow | You want scheduling in the same bot. See LINE booking system |
| Existing system integration | Reading and writing CRM, POS, or membership data.We need to review the system before quoting |
What you can cut from version one
On a tight budget:
- Cut the add-ons first: multi-platform, payments, and integrations can all come in version two without a rebuild
- Keep the rich menu simple: three buttons is enough; adjust later based on what customers tap
- Don't cut: knowledge base preparation, "I don't know" handling, human handoff, usage caps
Cut those last four and things go wrong: wrong answers, made-up answers, stuck customers, runaway bills. The first two can be added later without redoing anything.
A real number: an honest AI says "I don't know" a lot
Xiao Ai, the AI assistant in the bottom-right corner of our website, is built from the six parts above. Here are its dashboard numbers for the last 90 days (early July to late September 2026):
- 33 visitors chatted with it, and it sent 81 replies
- Of those, 28 (about 35%) were "I'm not sure about this, let me get you a human"
- Another 3 times, a customer asked and it handed off to a human; the human then sent 10 more replies
- Average response time: 11.5 seconds
35% sounds high, but that's by design: questions on a website are all over the place, and if the knowledge base doesn't cover it, it doesn't answer.If an AI customer service tool claims it "can answer anything," worry about the answers it gets wrong.
(These numbers include our own test conversations, not just customers.)
What you'll still pay each month after launch
Beyond the one-time development fee, there are three costs:
- LINE Official Account plan: paid directly to LINE, tiered by broadcast volume. Replies to customers don't count toward the broadcast quota
- AI model usage: billed by message volume, with a hard cap built into the system. The account should be in your name, with no markup from the developer
- Maintenance (optional): hosting, monitoring, backups, certificate renewal, and routine changes. Our plan runs NT$1,000–3,000 per month depending on scale, with no contract. Cancel anytime, and the bot keeps running
What to watch isaccount ownership. Your LINE Official Account, AI API account, knowledge base, and source code should all be in your name. That way, if you ever switch maintainers, you won't find everything stuck with the previous vendor.
Five questions to ask once you have a quote
- Who prepares the knowledge base, and how many days will it take? (Tests whether Part 2 is actually done)
- How does it respond when a customer asks something the knowledge base doesn't cover? (Tests Part 3)
- When it hands off to a human, does the earlier conversation come with it? (Tests Part 4)
- If someone floods it with messages, what's the maximum bill? (Tests Part 6)
- Whose name are the LINE Official Account and AI account registered under? (Tests ownership)
If all five get specific answers, a higher quote is worth it. If the answers are vague, the low price usually comes at a cost.
FAQ
Can I build a LINE bot myself? Keyword replies and LINE's built-in AI chatbot require no coding; you set them up in the dashboard yourself. A custom build on the Messaging API requires someone who can code, plus a server.
Why does one vendor quote NT$20,000 and another NT$200,000? The difference is usually in Parts 2, 3, and 4: whether the knowledge base is properly prepared, whether "I don't know" is handled, and whether handoff carries context. Part 1 plus an AI connection can be done for NT$20,000, but that's just plugging ChatGPT straight into your customer service. For what happens then, see this article。
How long until launch? For us, 3–5 weeks: 1–2 weeks for the knowledge base, 2–3 weeks for development and a trial run. During the trial, you play the customer for a few days, and we fix the knowledge base on the spot whenever it gets something wrong.
Our LINE bot development and AI customer service starts at NT$50,000 and launches in 3–5 weeks. All six parts above are included, and your website shares the same knowledge base.
If you're still deciding whether you really need AI,for many people the answer is to start with LINE's built-in NT$100-a-month option. Saying so doesn't help our business, but being upfront saves everyone's time.