AI chatbots on WhatsApp let e-commerce brands answer product questions, share order updates, recover abandoned carts and handle returns inside the app many customers already use daily. They work through the WhatsApp Business Platform, usually via a provider, and combine a language model with your store data. The best results come from narrow, well-defined use cases and a quick handoff to a human when the bot is unsure.
How do WhatsApp AI chatbots work?
Businesses connect to WhatsApp through the WhatsApp Business Platform, which Meta offers as the Cloud API. Most brands use a business solution provider such as Twilio, Infobip, 360dialog, Wati or a commerce-focused tool that plugs into Shopify or WooCommerce. The provider handles the connection; you add the logic.
An AI chatbot sits on top. When a customer writes, the message goes to a language model with instructions, store policies and, through integrations, live data such as order status or stock. The model writes a reply, and the provider sends it back through WhatsApp.
Two rules shape everything. Customers must opt in before you message them first, and business-initiated messages outside a 24-hour window after the customer’s last message must use pre-approved templates. Meta charges per conversation or message depending on category and country, so check current pricing.
Which use cases work best?
The strongest use cases are high-volume, repeatable and tied to data you already hold.
- Order tracking: the customer types an order number or phone number and gets the current status and courier link.
- Product questions: sizing, materials, compatibility and stock answered from product data and FAQs.
- Abandoned cart recovery: a template reminder for opted-in customers, with the bot answering any follow-up questions.
- Returns and exchanges: the bot checks eligibility against policy, collects the reason and issues a return label or form.
- Back-in-stock and restock alerts for customers who asked to be notified.
- Post-purchase care: usage tips, warranty registration and review requests.
Each of these has a clear end point, which makes it easier to measure and to keep the bot from wandering off topic.
What should the bot not do?
Some tasks should stay with people, or at least require confirmation. Refunds above a set amount, disputes, complaints about damaged goods and anything involving health or safety are better handed to a human agent. The cost of a wrong automated answer here is a lost customer or a legal problem.
The bot should also not invent policies. Instruct it to answer only from your published shipping, returns and warranty information, and to say it will connect the customer with the team when it cannot find an answer.
How do you design a good conversation?
Customers on WhatsApp expect short, friendly messages, not paragraphs.
Open with a clear menu and free text
Start with interactive buttons or a list for the top three or four needs, such as Track my order, Product question, Returns, Talk to a person. Let customers type freely too; the AI handles that.
Use rich message types
WhatsApp supports product catalogue messages, images, list messages and quick reply buttons. Showing a product card is better than describing it.
Make handoff obvious
Offer a human at any point, and pass the conversation history to the agent so the customer does not repeat themselves. Most providers support a shared inbox for this.
What rules must you follow?
Beyond Meta’s opt-in and template rules, WhatsApp’s Business Messaging Policy restricts certain product categories and prohibits some content. Read it before launch, especially if you sell supplements, alcohol or financial products.
Data protection law also applies. In the EU and Turkey, for instance, GDPR and KVKK cover the phone numbers and order data you process. Record consent, tell customers how their data is used and do not send personal data to an AI provider without a suitable agreement.
Finally, avoid spam. Template messages that customers block or report can lower your quality rating and messaging limits.
How do you measure success?
Track a small set of metrics from the first week.
- Containment rate: share of conversations resolved without a human.
- Handoff rate and reasons, which show where the bot needs better data.
- Average first response time.
- Revenue from recovered carts and WhatsApp-attributed orders.
- Customer satisfaction, using a one-question rating at the end of a chat.
- Block and report rates on template messages.
Review a sample of transcripts every week. Numbers show where problems are; transcripts show why.
What does a realistic rollout timeline look like?
Most brands move in three stages. First, set up the business account, verify the number and launch one use case such as order tracking, which uses data you already have. Second, after a few weeks of transcripts, add product questions with a knowledge base built from your catalogue and FAQs. Third, add proactive templates such as cart reminders and back-in-stock alerts, once opt-in volume is large enough to matter.
Rushing all stages together tends to produce a bot that does many things poorly. One use case at a time makes problems easy to spot and fix.
How Alien Road applies AI chatbots on WhatsApp
In a typical engagement we begin by analysing the client’s support inbox and order data to find the top questions and where customers drop off. We then choose a WhatsApp provider that fits the store platform, set up the business account and opt-in points, and get template messages approved. The AI layer is configured with the client’s policies and connected to order and product data, with a clear handoff route to the support team.
Before launch we test the bot against a list of real customer questions. After launch, weekly reporting covers containment, handoffs, recovered revenue and satisfaction, and we update the knowledge base based on transcripts. For example, a fashion retailer might start with order tracking and size questions, then add returns once the first two run smoothly.
Where to start this week
Export last month’s customer service emails and chats and count the five most frequent question types. If order tracking or product questions lead the list, they are your first WhatsApp use case. Ask two or three providers for a demo connected to your store platform, and check how each handles opt-in, templates and human handoff before comparing prices.
Keep reading: AI Marketing Agency · AI Marketing