How does an AI chatbot work?
An AI chatbot works in four steps: it reads the message, works out the intent, looks up the answer in your own business knowledge and writes a reply.
The knowledge is what you give it: products and prices, opening hours, FAQs, policies and past conversations. A large language model turns that into a natural reply in the customer's language. Good platforms keep the model inside a flow you designed, so it asks for the details you need (name, phone, preferred date) in order, and it can tag the lead as hot, warm or cold. When a question is outside what it knows, or a customer asks for a person, it hands the conversation to your team in the same inbox.
What is the difference between an AI chatbot and a rule-based chatbot?
A rule-based chatbot follows a fixed menu of buttons and keywords; an AI chatbot understands messages it was never scripted for.
If a customer types "ada slot Sabtu ni tak?" instead of tapping "Book appointment", a rule-based bot usually replies with its menu again. An AI chatbot understands the question, checks the knowledge it has, and answers. Rule-based bots are predictable but break easily; unconstrained AI can improvise. The practical middle ground is an AI chatbot that follows a designed flow: the AI understands and phrases the reply, and the flow keeps it on the steps your business needs.
What is the difference between an AI chatbot and an AI agent?
An AI chatbot answers and guides a conversation; an AI agent also takes actions to reach a goal, such as qualifying a lead, booking a slot or following up.
In practice the line is blurry and most business tools combine both: the chatbot handles the conversation and agent-style steps handle the booking, the CRM update and the handover. See our guide to AI agents for the full comparison.
Why do Malaysian businesses use an AI chatbot on WhatsApp?
Because that is where their customers write, and often when nobody is there to answer.
In a one-week sample of customer conversations on Mampu AI (1 to 7 September 2026), 67.1% started on WhatsApp, 17.1% on Facebook Messenger, 9.8% on TikTok, 5.3% by email and 0.7% on Instagram. And 47.9% of customer messages that week arrived outside Monday to Friday, 9 am to 6 pm, with 22.3% on the weekend. A chatbot on WhatsApp answers those customers at the moment they ask instead of the next morning.
Does an AI chatbot work in Bahasa Malaysia and Chinese?
Yes, a modern AI chatbot can reply in Bahasa Malaysia, Chinese and English, and switch when the customer switches.
That matters more than most buyers expect. In the same one-week sample, of the conversations where the customer's language was detected, 41.7% were in English, 32.5% in Malay and 25.8% in Chinese. A bot that only handles English well misses more than half of a typical Malaysian SME's customers. Ask any vendor to show a real conversation in Malay and in Chinese before you sign up, including mixed messages such as Manglish.
How much faster does an AI chatbot reply than a person?
In seconds instead of minutes or hours.
In the same week, inboxes with a published Mampu AI flow agent replied to 92.5% of customer-initiated conversations, with a median first reply of 8 seconds; inboxes without one replied to 52.7%, with a median of 109 minutes. When a conversation started after hours, customers waited a median of 13 minutes for a person, and 44.5% waited more than an hour. These are correlations across businesses that chose their own setup, not a controlled test, but the direction is clear.
How much does an AI chatbot cost in Malaysia?
From RM0 on a free plan to roughly RM40–RM600 a month for a small team, plus Meta's WhatsApp message fees.
As checked on each vendor's pricing page on 21 September 2026: Mampu AI Lite is RM119 a month (RM99 billed annually) and Pro RM499 (RM399 billed annually); Dah Reply starts at RM39 for one user; ForwardChat at RM299 (RM249 annually); SleekFlow Pro at RM569 (RM439 annually) for 3 users; WATI at US$39 for 3 users without AI; respond.io at US$99 (US$79 annually) for 5 users. Check the billing unit as well as the price: per-seat plans grow with your team, per-contact plans grow with your customers. Our roundup of nine platforms compares them side by side.
Is there a free AI chatbot?
Yes. Mampu AI's Free Builder plan is RM0 with no time limit: 50 monthly active contacts, 2 users and one channel of your choice (WhatsApp, Facebook, Instagram or TikTok), with the AI flow builder included.
ManyChat also has a free plan, for 25 active contacts on Instagram and Messenger only. Most other platforms offer a 7-day trial instead. On any platform, Meta's own WhatsApp fees still apply to the messages Meta charges for.
What does WhatsApp itself charge for chatbot messages?
Meta charges per delivered message for templates, and from 1 October 2026 also for replies after the first 1,000 a month per phone number.
After the free 1,000, each reply to a Malaysian customer costs RM0.0564 (US$0.014), according to Meta's rate card effective 1 October 2026. Messages in the 72 hours after a customer clicks a Click-to-WhatsApp ad stay free, and marketing broadcasts are always charged. A chatbot's replies are billed the same way as a person's. Our WhatsApp pricing guide explains what is billed and how to estimate your bill.
How do I build an AI chatbot without coding?
Describe what the chatbot should do, let the AI build the flow, add your business knowledge, connect a channel and test it.
On Mampu AI that looks like this: write one prompt such as "answer questions about our clinic, book scaling appointments and collect name and phone", or pick an industry template; FlowAgent builds the conversation flow; you add your FAQ and product list; connect WhatsApp, Instagram, Messenger or TikTok; then test it in the playground before you go live. Most businesses with a WhatsApp Business number can go live the same day.
Which channels can an AI chatbot answer on?
The main ones in Malaysia are WhatsApp, Facebook Messenger, Instagram, TikTok and website chat, and many tools add email, Telegram and LINE.
A shared inbox matters as much as the channel list: your team should see every conversation, from every channel, in one place, with the chatbot's replies and the lead details alongside. Check whether each extra channel is included in the plan or charged separately.
When should an AI chatbot hand over to a person?
Whenever the customer asks for one, the question is outside what the chatbot knows, or the deal is worth a person's attention.
Set the handover rules before you go live: complaints, refunds, custom quotes and high-value leads usually go straight to your team, with the conversation history attached so the customer doesn't repeat themselves. The chatbot should also stop replying once a person has taken over.
