For years my answer to any small business asking about a chatbot was to wait. That answer has changed, not because the technology got better, but because the way it is billed did.
Meta moved the WhatsApp Business Platform from conversation-based pricing to per-message pricing on 1 July 2025. That sounds like an accounting detail. It reverses the logic of the build-or-not decision.
What actually changed
Under the old model you paid for a 24-hour conversation window, so it made sense to fit as many messages into it as possible. Now, according to the official pricing documentation, you are charged when a template message is delivered, and the rate depends on the template category and the recipient country code.
Every extra message now has a price. That makes a chatty bot more expensive rather than more helpful.
The second detail matters more: category. Marketing templates are always billable. Utility templates, meaning updates about something the customer did such as an order or a booking, are free inside an open customer service window. And ordinary replies inside an open conversation are free.
The two windows that decide half the maths
Anyone making this decision should know both windows precisely.
The first is the customer service window. It opens for 24 hours when the customer messages you, and inside it your ordinary replies and utility templates cost nothing.
The second is the free entry point window. It runs for 72 hours and opens when someone reaches you through a Click to WhatsApp ad or a page button and you respond. While it is open, any message type is free.
That second window is the one most small businesses have never heard of, and it is exactly what makes WhatsApp advertising work economically when replies are fast. Reply two days later and you have spent 72 free hours and started paying.
So when is a bot actually worth it
The rule I use is simple. A bot earns its cost when one question arrives with the same answer more than a hundred times a month.
Opening hours, price, branch location, order status, whether a size is in stock. None of these need intelligence. They need a correct answer immediately, and that is where a bot genuinely wins, which is the same economic logic behind why companies invest in chatbots, because it replies in seconds instead of two hours, and the customer waiting two hours has already gone somewhere else.
It fails at the opposite: when every conversation is different, or when the sale itself needs a discussion. A bot trying to sell a consulting service repels more people than it converts.
The calculation to run before spending anything
Work out two numbers. First, how many messages you answer a month and how long they take. If that number is under roughly a hundred to a hundred and fifty, replying by hand is cheaper than any bot, and it is not close.
Second, what share of those messages have a fixed answer. If 70 percent of them have five answers between them, a bot takes that share and leaves you the thirty percent that needs a person. If the ratio is the other way round, you are about to build something that creates work rather than removing it.
Notice that this calculation is about time, not about message cost. Message cost matters when you are the one initiating, which is marketing. For replying, most of what you send falls inside a free window.
The cheapest version worth starting with
You do not have to start with a bot. The cheapest version is quick replies and a greeting message inside the WhatsApp Business app itself. Free, set up in half an hour, and it captures the first forty percent of the value.
If volume grows after that, move to the official platform. The order matters, because it lets you discover which questions genuinely repeat before paying to automate them. Most people do it the other way round and build a bot for questions they imagined. If you do reach the build stage, there is a step-by-step example in building a chatbot on Chatfuel.
On the types of bot and which suits which business, I wrote about that in types of chatbot, and the core idea still holds after the pricing change.
The part nobody mentions: templates get reviewed
Before you can send any message you initiate, you need an approved template. Meta reviews it and assigns it a category, marketing or utility, and that categorisation is theirs rather than yours.
This is the most important operational point in the whole subject, because the category decides the cost. A message you think of as a customer update can be classified as marketing if it contains a selling line, and then it is billable every time you send it.
The practical rule: keep utility templates completely clean of any offer. An order status is an order status. A phrase like "and there is a discount right now" turns the template into a marketing one, and turns a message that was free into one you pay for.
Also note that an approved template can be rejected at a later review. Do not build an entire process on a single template with no fallback.
One thing worth saying plainly
A bad chatbot is worse than none. A customer going in circles with a bot that does not understand them leaves with a worse impression than one who waited an hour and got a person.
So if you build one, put a visible exit in the first message: a word that hands the conversation to a human. And do not hide that it is a bot. People do not mind that it is a bot. They mind being trapped in it.
A final point: this is not a channel you own. The pricing changed once, templates get reviewed and can be rejected, and the category is decided by Meta rather than by you. Do not make it the only route by which a customer can learn something important.
And if your real goal is to remove manual work rather than to sell more, the team at Rivl wrote something practical on automating manual reporting without buying a platform. The same logic applies here: start from the thing that repeats and can be measured, not from the tool.