Customers
WhatsApp chatbot for business: what it handles and how it is built
A WhatsApp chatbot with AI answers your customers with your company's own information, checks your tools and passes the conversations that need a person to your team. Obinova designs it with you, connects it with n8n and documents everything.
What an AI WhatsApp chatbot does for a small business
Many people picture a WhatsApp chatbot as a button menu that leaves the customer guessing which option to press. An AI WhatsApp chatbot, sometimes called a WhatsApp AI agent, reads what the customer actually wrote, in their own words, looks for the answer in your company's information and replies in your team's tone, at any hour.
Think of the WhatsApp inbox at an estate agency in Bristol. Most messages ask whether a flat is still available, whether a viewing on Thursday is possible or which documents a tenant needs, and someone spends hours replying to them one by one. With AI, a customer can type "is the three-bed near the harbour still going?" and get a useful reply without waiting for anyone.
AI here means a language model, a program trained on large amounts of text that understands and writes sentences much as a person would.
When a question falls outside what it knows, the assistant says so and passes the conversation to a person, with the full history attached. Your team stops pasting the same replies and spends that time selling, advising and handling the difficult cases.
Working out whether you need one: your messages in three piles
Before talking about technology, look at your own messages. This exercise takes about an hour and tells you whether an assistant is worth it and which use case to start with.
- 01
Collect two weeks of conversations
Pick a normal fortnight, with no sale or holiday period. Go through the chats on the business phone, or ask whoever handles WhatsApp to log each enquiry in a spreadsheet with the date, the topic and who resolved it.
- 02
Sort every message into three piles
Pile A holds questions answered by information you already publish, such as opening hours, terms or required documents. Pile B needs a lookup in a tool, such as your calendar or an order status. Pile C needs human judgement, such as a complaint or a negotiation.
- 03
Read the proportions
If piles A and B make up most of the messages, an assistant will remove a lot of repetitive work. If pile C dominates, a chatbot adds little, and better reply templates or team organisation are the smarter investment.
- 04
Choose the first use case
Start with the most frequent topic in pile A, which is the easiest to automate. Pile B comes in a second phase, once the connection to your tools has been tested.
- 05
Keep the sheet as your baseline
The same sorting becomes your measurement. A few weeks after launch, compare how many A and B messages the assistant resolves on its own and how far the time to first reply has dropped.
Which enquiries a WhatsApp chatbot can handle
These are the most common uses. Sorting your messages into piles shows which one weighs most in your business.
Frequent questions
Opening hours, published prices, delivery times, return conditions or the documents a client needs. The assistant answers from information you have approved, and that information is updated when it changes.
Bookings and appointments
It checks availability in your calendar, suggests times and confirms the booking. It works for a property viewing, a first meeting with an accountant or a slot at a garage.
Order status
With an order number or the customer's phone number, it checks your online shop or order system and says where the parcel is and when it should arrive.
Lead qualification
It asks what your sales team would ask: what the person needs, by when and on what budget. The answers go into your CRM, the tool where you keep customers and opportunities, and the right salesperson is notified.
After-hours replies
At night or at the weekend, customers get a useful reply straight away. Anything that can't be resolved is logged so your team can pick it up first thing.
When the conversation should go to a person
We agree the handover rules with you before launch. Whoever takes over gets the conversation history and doesn't need to ask the same questions again.
Complaints and frustrated customers
A complaint needs empathy and the authority to make a decision. The assistant picks up on the tone, collects the basic details and passes the conversation to your team.
Decisions involving money or commitments
Discounts, returns outside your usual policy, custom quotes or contract terms. The assistant prepares the case and a person decides.
Sensitive topics
Health details, personal circumstances or specific legal and tax questions. At an accounting firm or a clinic, these go straight to a professional.
When the customer asks or the AI is unsure
If someone wants to talk to a person, the assistant makes it easy and doesn't push back. It also hands over whenever it can't find the answer in the approved information.
WhatsApp Business API: what it is, how to access it and what it costs
To automate WhatsApp with AI and connect it to your tools, the phone app is not enough. These are the Meta rules we go through with you from day one, based on its documentation in November 2026.
The app versus the API
The WhatsApp Business app is run by a person from a phone or computer. The WhatsApp Business API is part of the WhatsApp Business Platform and lets software receive and send the messages. That is how an AI assistant can check your calendar or CRM and log every conversation.
Direct access or through a provider
You can connect directly to Meta's Cloud API, with a business account at Meta and a phone number set aside for this use, or go through one of what Meta calls Business Solution Providers, which usually charge for their service on top. We help you choose and handle the setup.
Templates and the 24-hour window
When a customer writes to you, a 24-hour customer service window opens in which you can reply freely. Outside it, or when your company starts the conversation to remind someone of an appointment or confirm a dispatch, the message must use a template, which Meta generally approves first.
How Meta charges
Meta charges per delivered message, by category (marketing, utility, authentication or service) and the recipient's country, and does not charge for messages customers send you. Each number gets 1,000 free service messages a month. Since 1 October 2026, the ones after that are charged, as are utility templates sent inside the window. Rates per country are in Meta's rate card.
Consent and GDPR
Before you message a customer on WhatsApp, they must have agreed to hear from you there, and we look at how you collect that consent today. We define which personal data is stored, where and for how long. Customers know they are talking to an AI, and every supplier involved is documented.
How Obinova builds a WhatsApp chatbot with n8n
We build with n8n, an automation tool that connects applications through visual workflows. Everything is documented so your team can maintain and improve it.
- 01
Map the conversations
We start from your three-pile sort, reviewed with your team: what customers ask, who answers today and which cases need a person.
- 02
Prepare your company's information
We gather and organise your catalogue, published prices, terms, FAQs and procedures. The assistant answers from that material and is instructed not to answer what isn't in it.
- 03
Connect your tools
We link WhatsApp to your calendar, CRM, online shop or spreadsheets. The assistant works with real data and logs every contact in the right place. We also set up how your team is alerted on a handover.
- 04
Test before going live
Your team tries the assistant with real questions, including awkward and badly spelt ones. We adjust answers and rules before launch.
- 05
Monitor and improve
After launch we review conversations, unanswered questions and handovers. We track the share of enquiries resolved without help and the time to first reply, then adjust based on what we see.
Which businesses it pays off for, and what the cost depends on
It suits companies that receive a steady flow of WhatsApp messages and whose staff spend hours giving the same answers: online shops, estate agencies, accounting firms, clinics, training providers and sales teams handling lots of leads.
The cost depends on the number of use cases, the tools to connect, the state of your information and the languages the assistant needs. On top of that come Meta's messaging charges, which vary by message type and country, and the AI model's usage costs, which depend on volume. We give you a quote after the first meeting.
The usual stumbles are launching every use case at once, letting the assistant answer without an approved source, naming nobody to keep the information up to date and sending promotional templates to customers who only expected service messages.
For the first meeting, bring your two-week sorting sheet, the list of tools your team uses and the name of the person who handles WhatsApp today.
Frequently asked questions
How much does an AI WhatsApp chatbot cost?
It depends on the number of use cases, the tools to connect and your message volume. There are three parts: design and setup, Meta's messaging charges, which vary by type and country, and the AI model's usage. We give you a fixed quote after the first meeting, once we have seen your conversations.
How long does it take to set up?
It depends on the scope. A first use case, such as FAQs with a human handover, is ready sooner than an assistant connected to your calendar, CRM and shop. Setting up the Meta account and getting templates approved also takes some time. We give you a concrete timeline after the first meeting.
Can I run a WhatsApp chatbot on the WhatsApp Business app?
Only up to a point. The WhatsApp Business app is for replying from a phone or computer, with quick replies and automatic greeting and away messages. The WhatsApp Business Platform lets you connect WhatsApp to other software and to an AI assistant, so you need it for a chatbot that checks your calendar or CRM. An existing number can sometimes be moved across, with conditions worth checking first.
Is a WhatsApp chatbot worth it for a small business?
It is worth it when a large share of your messages repeat and can be answered with information that already exists. If you receive few messages, or most of them need human judgement, the benefit is small. Sorting two weeks of conversations into the three piles gives you the answer from your own data.
What happens if the AI gives a wrong answer?
We reduce that risk in three ways. The assistant answers only from information your company has approved, it is instructed to hand over to a person when unsure, and we review conversations regularly. When we spot a mistake, we fix the information or the rule that caused it so it doesn't happen again.
Is an AI WhatsApp chatbot GDPR compliant?
It can be, if it is designed with care. We define which personal data is collected and why, where it is stored, how long it is kept and which suppliers are involved. Customers know they are talking to an AI assistant and can ask for a person. We review these points with your data protection lead if you have one.