This guide covers a WhatsApp chatbot on the official WhatsApp Business Platform: how customers start the conversation, which rules apply once they have, how to set it up with Waki’s inbox, flows and integrations, and what to measure once it is running.
Read alongside this guide: WhatsApp Chatbot and Waki plans and pricing.
Key takeaways
- With a WhatsApp chatbot the customer writes first, so replies inside the 24-hour window need no template; ask for opt-in before you send anything after it.
- Give every menu an option that reaches a person, and make the handover visible to the customer.
- Answer by hand first, then let a flow take over the questions that repeat.
- Measure against your own earlier results, not against figures quoted elsewhere.
A WhatsApp chatbot explained
A WhatsApp chatbot is a flow of rules: when a message matches a trigger, send this reply, offer these options, and when the customer picks one, continue down that branch until a person or an answer is reached.
How it works on the WhatsApp Business Platform
Messaging on the WhatsApp Business Platform needs the customer’s opt-in, and any message that starts a conversation must use a template approved by Meta. Once a customer writes to you, a 24-hour customer-service window opens in which agents and automations can reply freely; outside it, only templates go through. Meta prices conversations rather than individual messages.
Setting it up with Waki
Waki’s flow builder has keyword and event triggers, delays, conditions, webhooks and an assign-to-agent step. AI answers from a knowledge base are optional and use your own OpenAI or Anthropic key.
- Create a Waki account and connect your WhatsApp Business Account through Meta’s embedded signup, or ask for a Waki managed number.
- Complete Meta Business Verification so your messaging limits can grow; the verification guide on this site walks through it.
- Create the templates you need and submit them for approval. Keep the first ones short and factual.
- Import contacts who have opted in, and record where each consent came from as a contact attribute.
- Send a test to your own number, then to a small segment, and read the report before you scale up.
Good practice
A few habits keep a WhatsApp chatbot useful to customers and safe for your number.
- Give every menu an option that reaches a person, and make the handover visible to the customer.
- Ask for consent in plain words and say what you will send and how often.
- Send at sensible local hours and keep the frequency you promised.
- Give every message a reason to exist: a date, an amount, a document, a decision.
- Make it easy to reply or to stop; an opt-out that works protects your quality rating.
Mistakes to avoid
The same errors come up again and again.
- Trapping customers in a loop with no exit. One frustrated reply undoes the time the bot saved.
- Messaging people who never opted in, including bought or scraped lists.
- Using a utility template for an offer; Meta recategorises it and your campaign stops.
- Ignoring replies. A broadcast that nobody answers teaches customers to ignore the next one.
What to measure
Each campaign report shows how many messages were sent, delivered, read and replied to; tracked links add clicks, and the contact list records opt-outs. Compare each campaign with your own previous ones rather than with figures quoted elsewhere: read receipts are a signal, not proof of interest, and the number that matters is how many people did what you asked.
Plans and Meta charges
Plans from ₹699/month on Waki, plus 18% GST. You pay for the first month or year at checkout and can turn on auto-pay later if you want. The conversation charges set by Meta are separate from the plan: on a number connected through your own Meta account Meta bills them directly, while on a Waki-provided number a prepaid wallet covers them. Pay yearly and you get two months free, and you can cancel from the app at any time.
Next steps
Conclusion
With a WhatsApp chatbot the customer starts the conversation, so the work is in answering well. Reply by hand first, note which questions repeat, then let a flow answer those and hand the rest to a person. Give every menu an option that reaches a person, and make the handover visible to the customer.