Your customers message you on LINE, and then someone on your team reads each message, types the details into a spreadsheet, and forwards them to whoever does the work. That copying is slow, it happens late at night, and it is where orders get lost. LINE OA automation removes the copying while leaving the conversation with your customer intact.
This guide explains what can safely be automated on a LINE Official Account, what should always stay with a person, and what it takes to set up. We run this pattern ourselves across the Rova brands, so the advice comes from using it, not from a vendor brochure.
Why LINE is the place to automate in Thailand
In Thailand, LINE is the default way people contact a shop, a clinic, a cleaner or a florist. A customer would rather send a photo of the stained sofa and ask "how much?" than fill in a website form. That is good news for conversion and bad news for admin, because every chat is unstructured text that a human has to interpret.
Automation gives that text a structure. A message that says "two-bedroom condo near Thong Lo, Saturday morning, deep clean" can become a record with an address, a date and a service type. Once it is a record, it can appear on a board, trigger a confirmation and be counted. Until then it is just a chat that scrolls away.
LINE OA versus a personal LINE account
This distinction decides whether automation is possible at all.
- A personal account is built for people. It has no supported way for software to read and answer messages, and running a business through one means a single phone holds your whole customer history.
- A LINE Official Account (OA) is built for businesses. It supports several admins, a shared chat inbox, and, importantly, the LINE Messaging API, which lets software receive messages from customers and reply to them.
If you are still on a personal account, moving to an OA is step zero. Plan it: tell regular customers the new account, keep the old one alive with a pinned message for a few weeks, and add the OA to your website, receipts and Google Business profile. Check LINE's current plan page for message limits, because plans cap how many broadcast messages you can send each month. Replies inside an ongoing conversation are treated differently from broadcasts, so confirm the current terms before you design anything around volume.
What is worth automating
Four jobs repay the effort. Each one is repetitive, has a clear right answer, and is boring for staff.
1. Order and booking capture
A customer sends the details of what they want. The system pulls out the fields (service, address, date, contact) and asks for anything missing. When the record is complete, it lands on your job or dispatch board as a structured job.
This is the highest-value automation for most service businesses, because it removes the re-typing step where mistakes creep in. A misspelled soi name or a wrong condo tower number sends a driver or technician to the wrong place.
2. Answers from a price list
"How much for a 40 square metre condo?" "Do you deliver to Bang Na?" These questions have answers in your price list. An assistant that reads only from that list, and not from the open internet, gives the same answer at 11pm as at 11am. We describe how this works in our guide to an AI assistant that speaks Thai, and the service itself is on our AI assistants page.
3. Confirmations and reminders
When a job is confirmed, the customer gets a message with the date, time and address. Two days before, a reminder goes out. After the job, a short thank-you and a request for a review. None of this needs a person to press send.
4. Follow-ups
A quote sent three days ago with no reply deserves one polite nudge. A customer who booked every month and has gone quiet might appreciate a reminder. These are easy to forget when you are busy, and easy for a system to do on schedule.
What should always stay human
Automation is most trusted when it knows its limits. Keep these with a person:
| Situation | Why a person handles it |
|---|---|
| Discounts and price exceptions | They are a business decision, not a lookup |
| Complaints and damage claims | Tone matters, and so does accountability |
| Refunds and cancellations of paid jobs | Money is involved and the rules have exceptions |
| Anything ambiguous or unusual | A wrong guess costs more than a short wait |
| Distressed or urgent customers | They need a human voice quickly |
| Anything involving health, safety or legal matters | The assistant should not offer advice |
The rule we use is simple: if the answer is not written in the price list or the FAQ, the assistant does not make one up. It says a team member will reply and passes the chat on.
The hand-over moment
The hand-over is where most chatbots fail. The customer has already explained their problem twice, and then a person arrives and asks them to explain it again.
A good hand-over has four parts:
- A clear trigger. Keywords like "complaint" or "refund", a request for a discount, a low-confidence answer, or the customer typing "talk to a person".
- The full conversation travels with it. The staff member sees everything the customer wrote, plus the fields already captured.
- A notification to the right person. Not a general inbox that nobody watches at 9pm, but a named person or rota.
- An honest message to the customer. Something like "I have passed this to the team, and someone will reply during opening hours." Include the hours.
Our own sales assistant works this way: it answers price and availability questions across the website, LINE and WhatsApp from a single price book, and hands the chat to a person when a discount or an edge case comes up. A shared inbox lets staff see both assistant-handled and human-handled conversations in one place.
A worked example: a cleaning order
Here is what a single order looks like end to end.
- A customer writes on LINE: "Deep clean for my condo, 55 sqm, Ari, this Saturday afternoon."
- The assistant checks the price list and replies with the price for that size and service.
- It asks for the missing details: building and unit number, a preferred start time, and a contact number.
- The customer replies. The assistant fills the fields and shows a summary to confirm.
- The job appears on the dispatch board as "pending confirmation", with the full chat attached.
- A coordinator confirms the slot. The system sends the customer a confirmation and schedules a reminder.
- After the job closes, a thank-you and review request go out.
At step 2 the customer asks, "Can you do 10 percent off if I book three?" The assistant does not answer. It passes the chat to a person, who decides. That single handover is the difference between a helpful tool and a liability. If you work across more than one channel, the same engine can serve a website widget and WhatsApp as well.
What it takes to set up
Most of the effort is not code. A typical project runs in this order:
- Map the real process. Read a few weeks of actual chats. Find the ten questions that make up most of them.
- Write the price list down. Services, sizes, areas, extras, exclusions. If it is not written, it cannot be automated.
- Decide the hand-over rules. What always goes to a person, who gets notified, and what the customer is told.
- Connect the OA. Link your Official Account to the system through the Messaging API, and decide where jobs will land (a board, a sheet or your existing tool).
- Test on real past conversations. In Thai and English, including slang, voice-note replies and photos.
- Launch quietly. Run it with staff watching for a couple of weeks, then adjust.
- Hand over. You receive the accounts, documentation and a walkthrough, so you are not dependent on us.
Customer data in chats is personal data, so decide who can see it and how long it is kept; our guide to PDPA basics for a Thai website is a sensible starting point, and your own lawyer should confirm your specific obligations.
What it costs
Published prices start at ฿39,000 for the automation starter, which covers one workflow end to end, typically in 2 to 3 weeks. The AI assistant package starts from ฿89,000 and covers a website widget, LINE OA and WhatsApp from one price book, typically in 3 to 5 weeks. Both are fixed in writing after a free scoping call. Third-party costs, such as LINE plan fees and AI model usage, are listed at cost in your proposal. You can see everything on the pricing page.
What automation will not fix
It will not fix an out-of-date price list, a team that disagrees about what a discount policy is, or a process nobody has agreed. It speeds up what already works. If your quoting is inconsistent, you will get inconsistent quotes faster. Fix the rule first, then automate it.
If you would like to talk through which of your LINE chats could be automated, get in touch and we will look at your real conversations on a free scoping call.