AI chatbots on WhatsApp let a brand answer questions, qualify leads and confirm orders inside the app customers already use every day. They run on the WhatsApp Business Platform, Meta’s API for businesses, and combine a language model with business rules: approved message templates, a 24-hour reply window and user opt-in. Used well, they shorten the path from ad click to conversation. Used carelessly, they get numbers flagged and blocked.
How do AI chatbots on WhatsApp work?
A WhatsApp chatbot is not an app inside WhatsApp. It is a server that receives messages from the WhatsApp Business Platform through a webhook, decides what to reply and sends the answer back through the API. Businesses reach the platform either directly through Meta’s Cloud API or through a Business Solution Provider such as Twilio, 360dialog, Infobip or MessageBird.
The AI part sits in the middle. A language model reads the customer’s message, looks up product data or order status, and drafts a reply. Business rules decide whether that reply can be sent automatically or should go to a human. Most good setups use the model for understanding and wording, and keep prices, stock and policies in a database the model reads from rather than remembers.
What rules shape a WhatsApp campaign?
WhatsApp is stricter than most channels, and the rules shape every flow. The ones that matter most:
- Users must opt in before a business sends them marketing messages.
- When a user writes first, the business can reply freely for 24 hours.
- Outside that window, the business can only send pre-approved templates.
- Templates are categorised as marketing, utility or authentication, and pricing differs by category and country.
- Quality ratings drop if users block or report the number, which can limit sending.
These rules push campaigns toward user-started conversations. Click-to-WhatsApp ads on Facebook and Instagram are the common entry point, because the user taps a button and sends the first message, which opens the free reply window.
Where do chatbots help most in a campaign?
The strongest results usually come from the first minutes after a click. A person who taps a Click-to-WhatsApp ad is curious right now and gone in an hour. A bot that replies instantly, answers the obvious questions and asks one or two qualifying questions keeps that interest alive until a sales person can step in.
Other good uses include order updates and delivery questions in e-commerce, appointment booking for clinics and service businesses, and re-engaging past customers with a template that invites them to start a new chat. Each one has a narrow goal that a bot can finish or hand off.
What a bot should not do
A bot should not argue with an angry customer, negotiate prices or invent policies. It should also not pretend to be a person. Clear wording like “I’m the store’s assistant, a team member can join if you need” keeps trust and reduces complaints.
How Alien Road applies AI chatbots on WhatsApp
In a typical engagement the work starts with a conversation audit. We read a few hundred real chats from the client’s current WhatsApp Business app or support inbox and list the questions that come up again and again. That list, not a feature wish list, decides what the bot handles.
Setup comes next. We connect the client’s number to the WhatsApp Business Platform through a provider, write and submit templates for approval, and build the flow: greeting, intent detection, answers from a product or FAQ source, qualification questions, and a clear handover to a human inbox. The language model only drafts replies from the client’s own content, and anything about price changes, refunds or complaints routes to staff.
Campaign launch is tied to ads. Click-to-WhatsApp ads in Meta Ads Manager point to the bot, with a pre-filled first message that tells us which ad the user came from. Measurement tracks each step: ad click, conversation started, questions answered, lead qualified, handover, sale. Reports show where users drop out, so the next round of changes targets the weakest step rather than the bot as a whole.
Writing the knowledge source
The bot is only as good as the content it reads. Short, plain answers work better than pasted policy pages. For a clothing store that might mean one entry each for returns, delivery times by city, size guidance and payment options, each written in two or three sentences. Updating this source is a weekly task, not a launch task.
Which metrics show whether the bot is working?
Avoid reporting only message volume. A bot can send thousands of messages and still lose customers. More useful measures:
- Time to first reply, which should be seconds for automated answers.
- Share of conversations resolved without a human, checked against satisfaction.
- Share of qualified leads passed to sales and their close rate compared with other channels.
- Block and report rates on the number, which protect long-term sending ability.
- Cost per conversation, including Meta’s template charges and provider fees.
If resolution rises while satisfaction falls, the bot is closing conversations it should hand over. That pattern shows up quickly when both numbers sit on the same chart.
Common mistakes to avoid
The most frequent mistake is sending marketing templates to contacts who never opted in, often from an old customer list. It saves effort for a week and then the quality rating drops. Another is building a long menu of buttons that feels like a phone tree. Users on WhatsApp expect to type normally, which is exactly where a language model helps.
A quieter mistake is skipping the human handover. When nobody answers after the bot says a team member will join, the brand looks worse than if it had no bot at all.
Where to start this week
Export one month of WhatsApp or website chat logs and sort the questions into groups. If three or four groups cover most conversations, a focused bot can handle them. Write the ideal answer for each group yourself first. Those answers become the content the bot draws from, and they will show you quickly whether automation is worth it.
Keep reading: AI Marketing Agency · AI Marketing