How to create a chatbot on WhatsApp without programming (2026 guide)

Serving businesses that sell via WhatsApp, we saw a pattern: Anyone who implements a chatbot in the first month reduces the average response time by 70% and frees the team to focus on hot leads. This guide shows you how to build yours — no code, no BSP, and no headaches.

What is a WhatsApp chatbot (and what it is not)

A WhatsApp chatbot is a set of automated flows that respond to the customer based on keywords, time or context. It's not a "robot that talks like a human" — and trying to pretend like it is often turns off customers.

The chatbot that works does three things well: (1) answers frequently asked questions instantly, (2) qualifies the lead before the human enters and (3) delivers the right customer, at the right time, to the right agent.

And the demand is there: Meta reports that most people now prefer to speak to companies by message rather than by phone or email^[1], and messages on WhatsApp have an open rate above 90%^[2]. The chatbot is what allows you to meet this demand without hiring proportionally.

Official API vs extension for WhatsApp Web

Those who are starting out usually face the same question: using the WhatsApp Business API (with BSP) or an extension such as Zapext that runs on WhatsApp Web?

• Official API: recommended for volumes above 10 thousand conversations/month, with cost per message and template approval process.
• Extension for WhatsApp Web: ideal for up to a few thousand conversations/month, one-click installation, zero cost per message, maintains the personal/business number that your customers already know.

For 9 out of 10 Brazilian SMEs, the extension delivers faster and cheaper results. For the rest of this guide, we'll focus on this approach.

The 6 steps to create your chatbot

1. Map the 10 most frequently repeated questions

Before opening any tool, open your WhatsApp and read the last 200 conversations. Write down any question that appears more than twice. This is your first flow — don't try to automate what is infrequent.

2. Define triggers (keywords)

Each flow needs a trigger: a word or expression that triggers the response. Good triggers are short, common, and unambiguous. Examples:

• preço, valor, quanto custa → plan flow
• horário, funcionamento, aberto → schedule flow
• endereço, localização, onde fica → location flow

3. Draw the flow on paper before assembling

Each automatic response must have a clear decision: end the conversation, branch with buttons or transfer to an attendant. Avoid trees with more than 3 levels — after 3 clicks, the customer will give up.

4. Mount on the tool (in the case of Zapext, it takes 5 minutes)

With Zapext, you create flows by dragging blocks: Trigger → Text → Audio → Button → Tag. No lines of code. See the chatbot pillar page for the technical details of the feature.

5. Train the team on the "human enters when..." protocol

The chatbot needs to know when to shut up. Define in advance which signals the human attendant takes over:

• Customer types "I want to speak to someone" / "attendant" / "human"
• Customer asks the same question twice (frustration)
• Customer mentions value above X (hot lead)
• Message outside business hours (reply on the next business day)

6. Measure and refine weekly

Every Friday, open the reports and see: which flows were triggered, where the customer abandoned, what % of conversations were fully automated. Adjust, publish, repeat.

5 mistakes that kill chatbots (and how to avoid them)

1. Classic-robot-message: "Hello! I'm the virtual assistant for company X." It sounds cold. Use the tone of your business.
2. Menu with 8 options: reduce to maximum 3.
3. Hide the human: always offer "speak to customer service" on the first level.
4. Forget after-sales: the best use of the chatbot is to confirm delivery, schedule a return and request a review.
5. Do not update: review flows every 60 days. New product, new price, new flow.

Metrics you need to track

Forgetting metrics is the main reason teams abandon chatbots. The five essentials:

• Automation rate — % of conversations resolved without human (target: 30-50%)
• Average first response time — goal: < 60 seconds
• Throughput to Human — when it exceeds 70%, simplify flows
• CSAT post-bot — ask “from 1 to 5, how would you rate?” to the end
• Lead → sale conversion — compare leads that passed through the bot vs. 100% manual service

Conclusion: start small, scale

The most common mistake is trying to map out the entire service on day 1. Start with just one stream — timetable or price. When it's been running well for a week, add the next one. In 60 days you have 8-10 streams covering 40% of conversations.

Want to see it in practice? Create your free Zapext account and build your first stream in 5 minutes.

Sources

1. [1] How people use messaging to connect with businesses — Meta Newsroom (2024-01). accessed 2026-05-19.
2. [2] Email open rate vs messaging open rate benchmarks — Sinch (MessengerPeople) (2024). accessed 2026-05-19.