A WhatsApp AI support bot
that answers where your customers already are
Support over WhatsApp today usually means a staff member typing the same answer for the twentieth time. An AI bot answers from your actual documentation instantly, and hands the conversation to a person the moment it is genuinely unsure.
What it is and who needs it
A WhatsApp AI support bot answers customer questions from your actual documentation, policies and FAQ, in the app where a large share of your customers already expect to reach you, rather than redirecting them to email or a separate app. It suits any business fielding repetitive support questions on WhatsApp today, a shop, a service provider, a property manager, where a staff member currently retypes close variations of the same answer throughout the day.
It is for answering questions, not taking action. A bot that needs to actually change an order or issue a refund needs an AI agent runtime with scoped tools and approval rules on top of this, a more involved build than a pure support bot.
What is inside
The bot answers from a knowledge base you control and maintain, with a hard rule to hand off to a person rather than guess when a question falls outside what it knows. Every conversation is logged, which doubles as free market research: we have seen this exact logging reveal that one lead channel converted at a very different rate than another, once dialogues were classified and visible rather than scattered across individual staff phones.
Multi-language support is often essential on WhatsApp given how international a customer base can be, and we build the knowledge base structure to support that from day one rather than retrofitting translation later. Admins update answers themselves through a simple panel, the pattern we have run for a visa consulting centre’s AI consultant bot, where staff update pricing and policy content without ever involving a developer.
How we build it
We start from your real support conversation history, the actual questions and the actual phrasing customers use, not an assumed FAQ list that misses how people really ask. Every answer is grounded in a specific document so the bot cannot drift into confident-sounding guesses, and we test it with adversarial questions, someone deliberately trying to provoke a wrong answer, before it goes near a real customer.
Hand-off is tuned to trigger early rather than late: a bot that keeps trying to answer past the point of real confidence erodes trust faster than one that hands off a bit too readily. We deploy on the official Cloud API with monitoring and rate limits that keep AI spend predictable regardless of a sudden spike in traffic.
We also keep a human in the loop for the first few weeks after launch, reviewing a sample of real conversations, because the gap between how a bot performs in testing and how it performs against real customers, with real typos, slang and impatience, only shows up once it is live.
Timeline and price
| Tier | Price | Timeline | What changes |
|---|---|---|---|
| MVP | from $1,300 | 2 to 3 weeks | One core flow, one channel, enough to go live and test demand |
| Production | from $2,900 | 3 to 6 weeks | Full flow, admin tools, integrations, monitoring, ready for real volume |
| Full control, handover-ready | from $4,930 | 5 to 8 weeks | Same build, plus documentation, test coverage and a handover package so any developer can take over, not just us |
Running cost is typically $25 to $90 a month in AI usage depending on volume, separate from WhatsApp API and hosting fees.
What you own at the end
The knowledge base, the conversation logs, and the full source code behind a model gateway that is not locked to one AI provider. Switching models or providers later does not mean rebuilding the bot.
Related
Pairs with a WhatsApp ordering bot so support and sales share one number and one knowledge base, and with a Telegram support bot with AI for customers who prefer Telegram. See help center and knowledge base for the content layer. For a real AI support build with admin-editable content, see the visa consulting centre case study.
Answering the same WhatsApp questions every day? Get in touch with a sample of your chats and we will show you what a bot could handle.
FAQ
How much does a WhatsApp AI support bot cost?
A bot answering from a knowledge base of a few dozen topics, with hand-off to a human, starts at $1,300 and is live in two to three weeks. Multi-language support and a full admin panel run $2,800 to $4,500.
Is this built on the official WhatsApp API?
Yes, through a verified WhatsApp Business account and the Cloud API. We do not use unofficial automation that risks the number being banned.
What stops the bot from giving a wrong answer?
It only answers from documents you provide, with a rule to say it does not know and hand off rather than guess. We test this against questions designed to tempt a wrong answer before launch.
Who owns the conversation data and knowledge base?
You do. Conversations are logged in a database you control, the knowledge base lives in your own admin panel, and the bot's code is yours.
Can it support more than one language?
Yes, this is a common need on WhatsApp given how international the audience often is. We build the knowledge base and response logic to handle multiple languages from the start if you tell us which ones matter.