Chatbot for WhatsApp without programming: the fastest way
Complete guide to creating a WhatsApp chatbot without code: types, flows, integrations and metrics. Updated 2026.
One chatbot for WhatsApp is an automatic assistant that responds, qualifies and directs conversations within WhatsApp — by keyword, guided flow or AI. Well designed, it resolves 40 to 60% of messages without human intervention and returns your team's time to what matters: selling. This guide shows the types, when to use each one and how to assemble yours without programming.
Demand exists on both sides: Meta reports that most people prefer to talk to companies by message rather than by phone or email[1], and in Brazil 89% of companies already sell via WhatsApp[2]. The chatbot is what allows you to meet this demand without hiring proportionally.
Types of chatbot on WhatsApp
- By keyword (rules): responds when the customer types "price", "time", "address". The most common in Brazil — simple to assemble, covers most of the volume and never "hallucinates" a wrong answer.
- Per button (guided flow): offers numbered options; the customer chooses. It has the highest completion rate because it eliminates ambiguity — ideal for screening and qualification.
- By IA (LLM): understands natural language and responds flexibly. Newer and more expensive; requires supervision to avoid inventing price or deadline information. Use as a top coat, not as a foundation.
When to use each
| Scenario | Recommended type | Why |
|---|---|---|
| FAQ about opening hours, price, address | Keyword | Exact answer, zero risk |
| Lead qualification with 3-5 questions | Button/flow | Higher completion rate |
| Complex technical support | AI + human | Flexibility with supervision |
| After-sales and NPS | Scheduled flow | Triggers by itself on the right date |
Architecture of a chatbot that converts
- Quick Screening (3-5 seconds): Identify what the customer wants with a menu question. This is where the speed gain comes in — leads answered in the first hour have a 7x greater chance of qualifying[3], and the bot responds within seconds, 24/7.
- Direct answer: delivers what can be delivered automatically — price, catalogue, scheduling link, duplicate.
- Smart Handoff: passes it to the human with context (what the customer asked for, which option they chose) when the topic requires it. The attendant does not restart the conversation from scratch — and the lead already enters with a step and label in the CRM.
The most common mistake is skipping step 3: chatbot with no way out for humans becomes a maze, and customer stuck in a loop reports the number — the type of behavior that WhatsApp policy punishes[4].
How to create your chatbot in 5 steps (without code)
1. Map real conversations. Take the last 50 conversations and list what customers ask for. Don’t build flow on guesswork — build on history.
2. Design the sorting menu. A maximum of 4 options (Prices / How it works / Support / Speak to an agent). More than that, the customer leaves.
3. Write the answers for each branch. Short, with clear next action: link, next question or transfer.
4. Configure triggers. First message from the contact triggers screening; keywords trigger direct responses; out of hours triggers the night flow.
5. Test with 5 real customers and measure. Flow completion rate and how many human requested. Adjust what stuck. Repeat every two weeks.
Two flows ready to adapt
Flow: scheduling appointments
Gatilho: cliente digita "agendar", "horário" ou "marcar" → Texto: "Claro! Qual período prefere? 1️⃣ Manhã 2️⃣ Tarde 3️⃣ Sábado" → Resposta → envia link da agenda do período + tag "agendamento-pendente" → 24h antes do horário → lembrete automático com opção de remarcar
Flow: qualified budget
Gatilho: cliente digita "orçamento" ou "preço" → Texto: "Pra montar seu orçamento: o que você precisa?" (3 opções de serviço) → Resposta → "E pra quando?" 1️⃣ Urgente 2️⃣ Este mês 3️⃣ Só pesquisando → Urgente → transfere pro atendente com prioridade + tag "lead-quente" → Pesquisando → envia material + agenda follow-up automático em 7 dias
Metrics that matter
- Auto Solve Rate — % of conversations that the bot closes without a human. Good: 40-60%.
- Flow Completion Rate — % of those who start the screening and reach the end. Below 70%, the flow is long or confusing.
- Time until handoff — when the customer asks for a human, how long does it take for someone to take over. The bot cannot be an excuse for delay.
- Post-bot conversion — bot-qualified leads that turned into sales. It is the metric that justifies (or not) the investment.
Errors that kill chatbot
- Menu with 7 options. Above 4, the abandonment rate skyrockets. Group and create submenus if necessary.
- No way out for human. The capital error — customer stuck in a loop reports the number.
- Bot that pretends to be a person. When the customer finds out (and finds out), trust breaks down. Signal and be quick.
- Flow never revised. What customers ask for changes; the bot needs to keep up. Review abandoned conversations every 2 weeks.
- Reply and do not register. Bot that does not tag or move the lead in the funnel throws away half the value — the qualification needs to arrive in the CRM.
Chatbot with or without official API
By Official API, the chatbot requires BSP, template approval and cost per conversation — enterprise operation path. Due to the extension on WhatsApp Web, the bot runs on the number you already use, without approval and at no cost per message. For SMEs and teams of up to 10 agents, the extension delivers the same practical result in an afternoon of configuration.
What Zapext offers
- No-code visual editor — drag and connect flows
- Triggers by keyword, time or tag
- Text, audio, image, document and button support
- Automatic transfer to attendant, with context
- Native integration with the Kanban funnel and quick answers
- Works within WhatsApp Web — no API, no BSP, on your current number
It has 598 users on the Chrome Web Store, with a 7-day unconditional guarantee. Installation takes minutes: see how to install the extension.
Popular Related Terms
- WhatsApp automation: the complete playbook
- Automatic message on WhatsApp
- Customer service via WhatsApp: complete guide
- WhatsApp Multiservice: multiple agents
- WebWhats: what is it
Sources
- [1] How people use messaging to connect with businesses — Meta Newsroom (2024-01). accessed 2026-05-19.
- [2] Email open rate vs messaging open rate benchmarks — Sinch (MessengerPeople) (2024). accessed 2026-05-19.
- [3] The Short Life of Online Sales Leads — Harvard Business Review (2011-03). accessed 2026-05-19.
- [4] WhatsApp Business Messaging Policy — WhatsApp (2026-09-23). accessed 2026-10-25.
- [5] Como o WhatsApp pode ajudar a impulsionar suas vendas — Sebrae (2023). accessed 2026-05-19.
Frequently asked questions
Chatbot by keyword or by AI?
For Brazilian SMEs in 2026, start with keyword. Covers 80% of cases with 20% of the effort, never invents the wrong price and is ready in an afternoon. AI comes in later, as a complementary layer for open-ended questions.
How much does a chatbot for WhatsApp cost?
Via the official API: BSP setup + cost per conversation, usually hundreds of reais per month. By extension over WhatsApp Web, the chatbot is included in the tool's plan — at Zapext, starting at R$24.75/month on the annual plan.
Do I need to know how to program?
No. The Zapext editor is visual: you drag message blocks, connect the options and publish. Anyone on the team sets up and edits flows.
Can I only have a chatbot outside business hours?
Yes. Configure the flow with an 'out of hours' trigger and it only responds during this interval — collects name and interest and marks the lead for the team to take on at the opening.
Does the chatbot work on cell phones?
Zapext works via WhatsApp Web (computer), which needs to be open for the bot to operate. The customer chats on his cell phone normally — for him, it's a regular WhatsApp conversation.
Can chatbot block my number?
Reactive chatbot (answers whoever contacts you) carries low risk — the risk lies in sending active messages without cadence. Always keep a way out for humans: customers stuck in a loop report, and the report weighs against the number.
How do I make the customer not realize it's a robot?
Don't hide — signal. 'I'm [company]'s virtual assistant' followed by an objective menu converts better than a bot pretending to be human. Customer accepts fast robot; does not accept nonsense.
