AI agents · Customer service WhatsApp bot and chatbot for business: requirements, rules and a good start

Two routes: Business app or Business Platform

WhatsApp offers businesses two products. Which one fits depends on whether people answer the messages or whether they are automated.

For an AI chatbot that looks up knowledge and queries systems, the Business Platform is the route. How the chatbot is built content-wise is explained in our article Website chatbot. The set-up is the same for WhatsApp. How we use such chatbots in AI customer care is described on the service page.

Meta rules you need to know

• Consent (opt-in): A business may only message people who have given their number and agreed to be contacted. The opt-in must clearly say that someone wants to receive messages, name the business and comply with applicable law. You choose how to collect it, for example by website, SMS, phone or signed form.
• 24-hour window: When a customer writes or calls, a 24-hour window starts. Within it you may reply without a template, and every new customer message resets it. Outside the window only message templates approved by Meta are allowed.
• Templates: Templates belong to categories such as marketing, information on ongoing matters (for example an appointment reminder) and verification codes. They must be approved before you send them.
• A route to people: Automation is allowed within the 24-hour window, but under Meta’s policy together with a way out: handover to staff in the chat, phone, email or a form.
• Billing: Since 1 July 2025 Meta has charged per delivered message, no longer per conversation. Prices change, so the current price list belongs in your planning.

What a WhatsApp chatbot does well

• Answer common questions: opening hours, prices from your price list, process, requirements, with a reference to the source.
• Look up status: order, repair, appointment or quote, through an interface to your systems.
• Bundle appointments: offer free slots, confirm, remind by template.
• Receive documents: accept photos and documents for a request and assign them to the right case.
• Hand over: for complaints, special cases and uncertainty, the conversation goes to a person with the history and a summary.

Law: AI disclosure, data protection, marketing

• AI disclosure: Under Art. 50 of the AI Act, people must be able to recognise that they are writing with an AI, unless this is obvious. Put the notice in the first reply. More on the transparency obligation.
• Data protection: Phone number, messages and photos are personal data. You need a legal basis (Art. 6 GDPR), must inform people (Art. 13 GDPR) and must sign a data processing agreement with the provider (Art. 28 GDPR). Special categories such as health data (Art. 9 GDPR) should not be requested in the chat without a separate review.
• Marketing: Under the ePrivacy Directive, “electronic mail” means any text, voice, sound or image message that can be stored in the network or in the recipient’s device (Art. 2(h)). This definition is wide enough that messenger messages may fall under it. For direct marketing, Art. 13 requires prior consent in principle, with a narrow exception for similar own products offered to customers.

This overview is general information, not legal advice. Check your individual case with your data protection officer or lawyer.

How to start

1. Collect requests: Which questions and cases reach you today by phone and email? Begin with the five most frequent.
2. Set up the opt-in: Where and with what wording do customers agree? Website, form, SMS or a paper form in the shop.
3. Define content and handover: checked answers, limits, handover with history, responsibilities and times.
4. Create templates: appointment reminder, status message, follow-up question. Allow time for approval by Meta.
5. Test internally: with real questions, including unfriendly and unclear ones.
6. Start small and evaluate: resolution rate, handovers and unanswered questions. They become your next content.