WhatsApp Chatbot Setup: How to Do It in 5 Steps

Say "WhatsApp chatbot setup" and most people picture code, integrations and complicated screens. What we see in the field is different. Almost never is technology what slows setup down; it is that nobody has decided what the bot will do, the answers are scattered around, and the question "what happens when the bot doesn't know?" is left empty.

Below we split the job into five steps: choosing a scenario, preparing information, building the flow, testing, and handing over to an agent. If you go in order, you can put your first bot live within a few days.

What do you need to set up a WhatsApp chatbot?

You can put an auto-reply on the WhatsApp Business app on your phone, but a real chatbot needs the WhatsApp Business API. We build the bot, the team dashboard and the other channels on top of this infrastructure. If you want to know the difference between the app and the API, take a look at our article What is the WhatsApp Business API.

On the Chatopya side, we handle Meta business verification and the application on your behalf; your tax certificate and brand details are enough. Dashboard setup takes about 5 minutes. If your number has never been registered with Meta before, Meta's own review time is added on top.

Step 1: What job will the bot do?

The most common mistake is wanting the bot to do "everything". Projects that start this way either never finish or go live with a half-built bot. We recommend starting with a single pain point.

To find it, look at the messages that came in on WhatsApp over the last month. Which questions keep coming back? Which one takes the most of the team's time? The list usually looks something like this:

• Frequently asked questions: opening hours, address, payment options, product information.
• Taking orders: a form that collects the product, quantity and address from the customer step by step.
• Appointments: choosing a day and time at businesses such as hair salons, clinics and service centers.
• Product recommendations: asking about the customer's need and showing a suitable product from the catalog.
• Return requests: collecting the reason for the return and the order details and passing them to the team.
• Surveys: short feedback after a purchase or a service.

Pick one or two scenarios from the list. You can add the rest once the bot has settled in.

Step 2: How do you prepare the information the bot will know?

A bot answers only as well as the information you give it. In short: you need to write the answers once, correctly and up to date.

In Chatopya, the AI answers frequently asked questions around the clock from your own business information. You add your business information and the pages on your website to the knowledge base in the Bot Center. If a question comes in that has no match in the knowledge base, the bot does not make up an answer; it passes the conversation to an agent. That is why missing information does far less harm than wrong information.

You can use this checklist for preparation:

1. Write down the 20-30 questions that come in most often, in your customers' own words.
2. Prepare a short, clear answer for each. Note information that changes, such as campaigns or prices, separately.
3. Update information such as return terms, payment options and opening hours on your website; the bot will read from there.
4. Decide which topics the bot should never answer. These should go straight to an agent.

If you run an online store, there is one more step. Stock and price information from your connected store is refreshed at regular intervals, and the bot can list products from the catalog and send a PDF catalog and images.

Step 3: How do you build the flow?

There are two paths here, and most businesses use both together. You leave the free-form question-and-answer part to the AI. For jobs where you need to collect information in a set order, you build a flow.

What should you watch for in the flow designer?

The flow designer in Chatopya works by drag and drop; you don't write code. Still, good flows share a few traits. Keep the steps short. Ask just one thing at each step. Think ahead about what happens if the customer gives up in the middle of the flow or asks something else.

WhatsApp Flows is also available on the Business and Enterprise plans. It is handy if you want to collect multi-field information such as appointments or orders with form-like screens inside the chat.

How well does the bot understand the customer?

Customers don't write to a bot in neat sentences. Messages like "prc how mch" or "the product I asked about yesterday" are normal. Chatopya's AI follows the conversation history and tolerates typos, abbreviations and idioms. It can answer in other languages as well, with Turkish first among them.

Step 4: How do you test before going live?

Don't skip testing. In the Bot Center, you can try your flows before publishing; use that chance to have a few people from the team assess it. When the person who built the bot tests it alone, they usually ask the "right" questions and miss the problems.

During testing, be sure to try these:

• Ask the same question three different ways, with abbreviations and typos.
• Ask something that isn't in the knowledge base. Does the bot hand over to an agent instead of making something up?
• Change the subject in the middle of a flow, or abandon the flow halfway.
• Write like an angry customer. Sentiment analysis tags an angry or worried customer; see how that shows up for your team.
• Write outside working hours and check that the handover rule works correctly.

Step 5: When should the bot hand over to an agent?

What separates a good bot from a bad one is often the moment of handover. If a customer asks the same thing three times and gets the same answer, the bot is not doing its job.

In Chatopya, when the AI can't resolve a conversation, it hands it over to an agent together with the chat history. The agent doesn't make the customer start from scratch. You define working hours and the handover rule in the Bot Center. The same number can be used by several agents, and chats can be assigned and handed over. On the Conversations screen, the customer card also stays open on the side; what the customer previously wrote from Instagram or Web Chat shows up in the same timeline.

When writing the handover rule, think about these situations:

• When the bot can't find the answer in the knowledge base.
• When the customer clearly wants to talk to a person.
• When the customer seems angry or worried.
• When the topic is sensitive, such as payment or a complaint.

So set up your handover rule in a way that doesn't create long waits. If the window closes, you will need a template; we explain why templates get rejected in our template approval guide.

What should you track after launch?

Don't assume the job is done once the bot is live. In the first weeks, reading the conversations teaches you the most. Which questions end up with an agent? Which answer is missing from the knowledge base? As you find these, you update the knowledge base.

The analytics dashboard shows response time, resolution rate and satisfaction. The satisfaction (CSAT) survey sent after each conversation also tells you what kind of experience you gave from the customer's point of view. One more note: the AI bot runs on credits, and the AI Bot credit package is separate. When you're ready, you can find the details on our WhatsApp customer service and AI chatbot page.

Frequently asked questions