"How much does a WhatsApp chatbot cost?" The short answer is "it depends", which helps nobody. The honest answer is more awkward: the figure on the provider's pricing page is one of five cost lines, and often not the largest one.
This guide deliberately contains no numbers. Figures change by provider, by country and by quarter; the structure does not. Once you understand the lines, you can read any quote — ours included — and know exactly which questions are still unanswered.
The five cost lines
1. The platform licence. What you pay the provider for the software. It is the only number that appears on websites, which is why everyone compares only this. That, as we will see, is a mistake.
2. Meta's API conversations. If the bot lives on WhatsApp — rather than only on your website — Meta charges separately per conversation, at its current rates, which vary by country and by who starts the exchange. It is a different bill from the provider's: some pass it through as-is, others fold it into their fee, and it is worth knowing which yours does before you sign anything. We cover how that billing actually works in WhatsApp Business API: what it is and what it costs.
3. Setup and training. Someone has to register the number, connect the diary or the CRM, load your services and teach the system how your business talks. Some providers include this; some sell it as a separate project. Neither is wrong — what matters is that it appears in the quote rather than on the third invoice.
4. The content. The invisible line. A bot is exactly as good as the answers it knows how to give. Who writes those answers the first time? And who updates them when you change a service, a price band or your opening hours? If the answer is "you do", those are your hours — and the owner's hours are the most expensive in the business, even though they never appear on a spreadsheet.
5. Maintenance and review. A bot is not a sign you hang on the wall and forget. Businesses change, customers' questions change, and what answered well in March may answer poorly by September. We read real conversations every week; it is the only method we have found that tells you whether an agent genuinely answers well. Whoever your provider is, ask who does that work in your case — and whether it is in the price, or simply nobody's job.
The costs that never appear on a quote
The cheap bot you maintain by hand. Flow-builder tools look like a bargain until you learn how they are maintained: every new question a customer asks is a new branch someone has to draw, test and reconcile with the rest of the tree. The licence is cheap; the afternoons someone on your team spends gardening it are not.
The "free" project you pay for yourself. "We'll set it up at no cost" usually means the real work — writing the answers, testing them against real enquiries, fixing what breaks — lands on you. It is not free; the price is simply denominated in your hours instead of pounds.
The cost of not replying. The biggest one, and the only one missing from every quote — because nobody invoices you for it. You just lose it. Every enquiry that lands out of hours and goes cold overnight had a value; we wrote about how that gap behaves in The weekend problem.
Three billing models, three very different incentives
Per conversation. You pay for every chat the bot holds. It sounds fair, until you notice the incentive: your bill grows with your success, and the channel that should always be open becomes a tap you have to watch. Some businesses end up throttling their own bot to keep the invoice down.
Per contact. You pay by the size of your database, whether people message you or not. It comes from the email marketing world and it punishes having history: dormant contacts count too.
Flat monthly fee. You pay the same whether the bot talks to ten people or a thousand. The incentive switches sides: the provider now wants the system to resolve things quickly and well, not to generate volume. It is the model we chose for Zatio; the actual numbers live where they should, on the pricing page, not in an article that would age badly.
The questions that make quotes comparable
With the five lines in front of you, comparison comes down to a handful of questions:
- Is Meta's charge included in the fee or billed separately? Who pays it?
- Who writes the initial answers, and who maintains them when something changes in my business?
- What does the bot do when it does not know something: admit it, make it up, or hand over to a person?
- Is anyone reading real conversations, or just watching a dashboard?
- How many hours a month will this need from me or my team?
- If I cancel, what do I keep? Is the number still mine? What about the conversation history?
The last thought is the one we would give a friend: the most expensive quote is not the one with the biggest figure on it. It is the one that quietly hands the work back to you. A bot you have to watch, correct and maintain by hand is not a tool — it is one more employee, and one you cannot even hold to account.