How to Build an AI WhatsApp Chatbot (Without Writing Code)

An AI WhatsApp chatbot that actually gives useful answers — not canned responses — requires three things: a good knowledge base, the right AI provider and proper fallback handling. Here's how to set it up in Waqly without writing a single line of code.

Step 1 — Connect your WhatsApp number

Before anything else, connect your WhatsApp Business phone number to Waqly. You'll need a Phone Number ID and an Access Token from your Meta Developer account. The setup guide covers this in under 10 minutes.

Step 2 — Choose your AI provider

Waqly supports four AI providers. You bring your own API key — there's no markup on AI usage.

DeepSeek

Best for: High volume, cost-sensitive use cases. Very fast and affordable.

OpenAI GPT-4oRecommended

Best for: Highest quality reasoning and nuanced replies. Ideal for complex queries.

Anthropic Claude

Best for: Long documents, safety-focused use cases. Excellent at following instructions.

Google Gemini

Best for: Indian languages and multilingual support. Strong at regional language queries.

In Waqly Settings → AI, paste your API key and select the model. Your key is encrypted at rest — Waqly never has access to the raw key after saving.

Step 3 — Build your knowledge base

The knowledge base is what makes your AI chatbot useful vs. generic. It's a set of plain-text documents the AI retrieves from when answering questions.

What to add to your knowledge base:

Frequently asked questions and answers
Product or service descriptions with pricing
Return, refund and cancellation policies
Business hours, locations and contact details
Onboarding or how-to instructions

How it works: when a customer sends a message, Waqly scores your knowledge base chunks using keyword-overlap — then injects the top-scoring chunks into the AI prompt. This keeps the AI grounded in your actual content rather than making things up.

Step 4 — Configure auto-reply rules

Waqly processes incoming messages in this order:

Active flow session — If the contact is mid-flow, the flow continues.

Keyword rules — Exact or contains-match rules trigger specific replies or flows.

AI intent rules — Your configured intent rules are matched using AI.

AI catch-all — If nothing else matches, the AI generates a reply from your knowledge base.

Step 5 — Test before going live

Send test messages to your own number and check: Is the AI using your knowledge base? Are keyword rules triggering correctly? Are flow sessions persisting? Once you're happy, enable AI auto-replies and your chatbot is live.

Handling voice messages

If you enable voice transcription in Waqly (Whisper, Deepgram, AssemblyAI or Gladia), customers can send voice notes and the AI will understand and reply to them the same as text. This is especially important in markets where voice messaging is common.

Build your AI chatbot today

Connect your WhatsApp number, upload your knowledge base and go live in under 10 minutes.

Start for free →