WhatsApp automation for business: what's possible, and what it costs

There is the consumer WhatsApp Business app, a free app for a single business phone, and there is the WhatsApp Business Platform, an API built for automation and scale. The app is fine for a small operation replying by hand. Anything programmatic, such as bots, AI replies, or sending from your own systems, needs the Platform, usually through a provider like Twilio. Registering a number on the Platform means it can no longer be used on the consumer app, so choose deliberately.

On the Platform you can automate the full round trip. Outbound: order confirmations, shipping updates, appointment reminders, and re-engagement messages. Inbound: greeting and away messages, quick replies to common questions, and an AI assistant that answers from your catalog and policies using RAG. You can route conversations to a person when the bot reaches its limit, so automation handles the volume and humans handle the exceptions.

WhatsApp constrains business messaging on purpose. Outside a 24-hour window after a customer messages you, you can only send pre-approved message templates, not free-form text, to prevent spam. Businesses need a verified Meta Business account and a display name. And you opt users in to receive messages. These rules shape the design, so it is better to plan for them than to discover them mid-build.

Costs come in layers: the number and any provider fees, plus Meta's own charges. Since July 2025 Meta charges per template message delivered, not per conversation, at a rate set by the template's category (marketing, utility, or authentication) and the recipient's country. Replies inside the 24-hour customer service window cost nothing, and so do utility templates, such as an order update, sent while that window is open. Marketing templates are charged every time, and utility and authentication rates fall as monthly volume grows.

That shapes the design. Answering customers while their window is open is free in Meta fees, so an assistant that resolves questions promptly costs little to run, while outbound campaigns are where the bill grows. Budget for the provider, the marketing and out-of-window templates you plan to send, and the build itself.

Begin with the highest-value, rule-based messages, such as order and shipping updates, which are predictable and need no AI. Add greeting, away, and quick replies next. Introduce an AI assistant only once the basics are live and you know the questions customers actually ask, so the bot is grounded in real demand and your real catalog. It is the order we build conversational AI in, and as with any automation, the replies are only as good as the data behind them.