BSP pricing, session windows, rate limits, and fallback logic for enterprise WhatsApp AI assistants.
WhatsApp is where Indian customers already are, which makes it the most attractive channel to automate and the one with the most rules to get wrong. Most guides skip the rules and go straight to the chatbot. The rules are the hard part — they decide what you're allowed to build, what it costs, and whether your number survives.
Three different products share the name:
The API has no interface of its own. You either build one or use a provider's inbox. Access is through Meta directly or, more commonly, through a Business Solution Provider (BSP) who handles onboarding, number provisioning and billing.
These constrain the product, so understand them before you scope anything.
The 24-hour customer service window. Once a customer messages you, you can reply freely with any content for 24 hours. After that window closes, you cannot send a free-form message. You may only send a pre-approved template. This single rule shapes every WhatsApp automation: your agent can converse naturally inside the window and essentially cannot outside it.
Template approval. Any message initiating a conversation, or sent after the window, must use a template submitted to Meta and approved in advance. Templates support variables, but the structure is fixed. You cannot have an AI compose a novel outbound message — it can only fill approved slots.
Opt-in. You need documented consent before messaging someone. Where and how you collected it matters, and "we had their number" is not consent.
Quality rating and messaging limits. Meta scores your number based on how recipients react — blocks and reports drive it down. A poor rating restricts how many unique users you can message daily and can end in a ban. This is the real penalty for aggressive outbound, and it's enforced automatically.
Conversation-based pricing. You're billed per conversation window, not per message, with different rates by category (marketing, utility, authentication, service) and by country. India is a large and competitively-priced market, but marketing conversations cost meaningfully more than service ones. Model your volume by category — a design that turns service conversations into marketing ones can multiply your bill.
Inside the 24-hour window, and specifically here:
Where AI does not fit: composing outbound messages (templates are fixed), and deciding to message someone who hasn't opted in.
The realistic architecture is a hybrid. Templates and buttons carry the structured, compliance-bound parts. AI handles the free-text middle. Interactive list and button replies are underrated — they're faster for the customer than typing and they remove entire categories of misunderstanding.
Three lines:
The pattern that catches people out: costs scale with conversations, and a badly designed flow that forces customers to start new conversations repeatedly multiplies the bill. Design to resolve inside one window.
Quality rating is driven by blocks and reports, so:
The uncomfortable truth: the cheapest way to protect quality rating is to send less. Volume targets set without reference to it are how businesses lose the channel.
WhatsApp automation is constrained by the 24-hour window, template approval, opt-in and quality rating — and every good design works with those constraints rather than around them. AI belongs inside the window, handling free text and retrieval; templates carry everything else.
Get the compliance model right and the channel is exceptional. Get it wrong and you lose the number. See WhatsApp AI chatbot development for what we build, or tell us your use case and we'll say whether it fits inside the rules.
Join high-growth companies that chose intelligence over mediocrity. Book a free consultation and see what Velex can build for you.
