Ask three vendors what a WhatsApp chatbot costs and you will get three numbers that are not comparable, because they are quietly quoting different things. Some quote the build only. Some bundle a platform subscription. Almost nobody separates out the part Meta charges you regardless of who builds it.
There are three costs. Understanding them separately is most of the battle.
Cost 1: the WhatsApp Business API itself
This is Meta's charge and it is unavoidable for any automated bot at scale. You cannot run a proper chatbot on the free WhatsApp Business app. That app is designed for a human replying by hand, and automating around it puts your number at risk of being banned.
The official API is accessed through a Business Solution Provider. Well-known ones include WATI, AiSensy, Interakt, 360dialog, Twilio and Gupshup. You pay the provider a platform fee, and conversations are billed on top. For a small Malaysian business with moderate volume, this layer typically lands in the low hundreds of ringgit per month.
Ask any vendor: is the API account in my company's name? It should be. If the bot is built on the vendor's account, moving away from them later means starting over.
Cost 2: the AI model, if the bot uses AI
A rule-based bot that walks people through a fixed menu has no model cost. A bot that understands a question phrased however the customer felt like phrasing it uses a language model, and that is billed per use.
The good news is that this has become dramatically cheaper. Reporting through 2026 puts a complete small-business bot in the region of USD 40–100 a month all-in, roughly half what the equivalent cost a year earlier. For most shops the model is not the expensive part any more.
Cost 3: the build
This is the one-off, and it is the one that varies most, because it depends entirely on how much your bot needs to know and do:
| Scope | What it does |
|---|---|
| Simple | Answers a fixed set of FAQs, captures a name and number, hands over to a human. |
| Connected | The above, plus live answers from your real data: stock, prices, opening hours, order status. |
| Transactional | Takes bookings or orders, writes them into your system, sends confirmations, handles payment links. |
Each step up roughly multiplies the integration work, because each one means touching another system that has its own login, its own API and its own edge cases.
Where quotes hide things
- Per-seat inbox fees. Some platforms charge per agent using the shared inbox. Fine at two people, unpleasant at fifteen.
- Template message approvals. Any message you send first (rather than replying) must use a pre-approved template. Ask who handles submissions and rewrites when one is rejected.
- "Unlimited" conversations. Meta bills conversations. If a quote says unlimited, someone is absorbing the cost, and that usually shows up as a higher platform fee.
- Who owns the conversation history. It is customer data. It should be exportable.
Is it worth it?
The real test is not whether the bot is impressive. It is whether you are currently losing enquiries because nobody replies fast enough.
Count the messages you get outside working hours for one week. Count how many of those people messaged a competitor too. If the first number is meaningful and the second is a guess you would rather not make, the bot pays for itself on response time alone, before anyone mentions AI.
And if the answer is that you get four enquiries a week and answer them all within the hour, you do not need a chatbot. We will tell you that rather than sell you one.
Pricing described here reflects publicly reported market rates in 2026 and varies by provider, message volume and country. Get a written quote broken into these three layers before committing.
