WhatsApp chatbot for a small business in India

If your phone fills up with the same WhatsApp questions every day, a bot can answer some of them. It cannot answer all of them, and a bad bot costs you customers faster than a slow reply does. There are three real choices: use the AI built into WhatsApp by Meta, build a custom bot on the official API, or wait.

This guide explains each choice in plain words, lists what to prepare, and names the questions a bot should never answer alone. WhatsApp features, availability and fees change often. Treat this post as a map and check Meta's own pages before you decide.

What a WhatsApp bot can and cannot do

A bot is good at repeated, factual questions where the answer is the same for everyone. Opening hours, address, which services you offer, how to book, which documents to bring, and where an order stands are typical examples.

A bot is weak when the answer depends on judgement, a person's situation, or money you may have to return. It also struggles when a customer is upset. It can sound confident and still be wrong, so the design must assume it will sometimes fail and give the customer a way to reach a person.

Two other limits come from WhatsApp itself. On the official API, you can send free-form messages only inside a customer service window that opens when the customer writes to you. Outside it, you can send only pre-approved message templates, and you need the customer's opt-in first. Read Meta's WhatsApp Cloud API documentation (opens in a new tab) for the current rules before you plan any messages.

Option 1: Meta's Business AI inside the Business app

Meta has an AI agent for businesses that now goes by the name Meta Business Agent (earlier coverage called it Business AI). Meta's announcement (opens in a new tab) says it can answer business-specific questions, recommend products from a catalogue, book appointments and qualify leads. It works on WhatsApp and Messenger, and Meta says it is expanding to Instagram. It replies in the customer's language, and the business decides when a team member steps in.

On the WhatsApp Business page for the app (opens in a new tab), setup has three steps. You connect the agent under Tools, teach it from past chats and, if you wish, your website or catalogue, then test and switch it on. You can mark topics it should avoid, set the tone, and choose who it replies to: everyone, only new customers, or customers who come from ads. You can take over any chat yourself.

Meta says this is "available to eligible businesses in select markets", and that some features may not yet reach every eligible business. For India, Meta's May 2026 post introducing Business AI on WhatsApp for small businesses in India (opens in a new tab) says it works in all native Indian languages and rolls out to eligible businesses, which must meet criteria set out in Meta's Help Center. So do not assume your account has it: open the Tools menu in your WhatsApp Business app and see whether the option appears. This post gives no prices because Meta's pages do not settle them. Check the current page before you decide.

This option suits a business that already runs on the WhatsApp Business app, has a simple catalogue or FAQ, and wants something working this week. The trade-off is control. You work within the settings Meta gives you.

Option 2: a custom bot on the official API

The WhatsApp Business Platform is the developer route. Meta's platform overview (opens in a new tab) describes a set of APIs for messaging customers at scale. Your software sends messages through the Cloud API and receives customer messages through webhooks.

You need a business portfolio, a WhatsApp account and a messaging account tied to it. You also need a business app with the right access tokens. Most small businesses reach this through a developer or a messaging provider.

Meta's pricing page (opens in a new tab) describes the structure. Meta charges per delivered message, based on the message category and the customer's country. Template messages fall into marketing, utility and authentication categories. Messages inside the open service window are a separate category. Rates and free allowances change, so read the page on the day you plan your budget.

A custom bot suits you when you need to connect to your own systems, such as a booking diary, a stock list or a CRM. It also suits you when you need strict rules about what the bot may say. It costs more to build and needs someone to maintain it. If you want help scoping this, our AI automation work covers it.

Option 3: do not build one yet

Skip the bot if any of these is true.

• Your questions are mostly different every time, and few repeat.
• Your replies depend on a person checking something first, such as stock, a diary or a diagnosis.
• You have no written answers yet. A bot cannot be better than your own FAQ.
• Nobody on the team can read the chats each week and fix the bot's mistakes.

In these cases, better first steps are saved replies in the WhatsApp Business app, a clear business profile with hours and address, and one person who owns the inbox. Revisit a bot once you can list your 20 most common questions with correct answers.

What to prepare

Whichever route you choose, prepare the same material.

1. Your top questions. Export or scroll through a month of chats and list what people ask most.
2. One correct answer for each. Write them in plain words, with dates and conditions. Remove old offers.
3. Opening hours and holiday dates. Keep them in one place so the bot and your profile match.
4. Handover rules. Decide who receives a handed-over chat, during which hours, and what the customer is told when nobody is free.
5. A test list. Collect 20 real past questions to try before the bot meets customers.

Questions a bot should never answer alone

Set the bot to hand these to a person, or to say it cannot help and give a way to reach you.

• Refunds, cancellations, compensation and disputes.
• Any medical question, dosage, symptom or report result.
• Legal questions, contracts and notices.
• Financial advice, loans, tax and insurance claims.
• Complaints, threats, or a customer who says they are unhappy.
• Anything that changes an order, a booking or a payment after it was made.
• Questions about another customer, or requests to share personal data.
• Anything the bot's files do not clearly cover.

Meta's developer documentation (opens in a new tab) lists business types the agent does not support, including finance, government and health, so check this before you plan around it.

Hand-off to a person

A hand-off has three parts. First, a trigger: a keyword such as "agent", a topic from the list above, or two failed answers in a row. Second, a message that tells the customer a person will reply and when, for example "A team member will reply during office hours, 10 am to 7 pm." Third, a clear owner who sees the chat with its history, so the customer does not repeat everything.

When you take over in the Business app, you simply reply in the same chat. On a custom bot, your developer must pass the chat to a shared inbox or an app your team uses. Test it by sending a refund question yourself, and see how long the chat takes to reach a human.

Privacy and consent for customer messages

Chats hold names, phone numbers and sometimes health or payment details. Tell customers when they are talking to an automated assistant. On the API, get opt-in before you send template messages, and make clear who you are and why you are messaging. Meta's documentation states this requirement.

Ask where chat data is stored, who can read it, and whether it is used to train anything. Meta's own pages did not answer this fully when this post was written. If you handle personal data of people in India, check your duties under the Digital Personal Data Protection Act with your lawyer.

Where to start

Start with the 20-question test list and a decision on who takes over. With those two things, either route becomes easier to judge. If you want a bot on the official API that links to your systems and hands over cleanly, see our AI automation service.