How to build a WhatsApp bot that qualifies leads in 60 seconds

WhatsApp has overtaken SMS as the primary messaging channel in most European markets. If your leads prefer chat over phone calls, WhatsApp Business API + AI is throughput no call center can touch: one bot handles 500 conversations in parallel, 24/7, multi-language, replying in 3 seconds.

This tutorial shows how to build one with Make + OpenAI + WhatsApp Cloud API — a weekend project, under $50/month for 1,000 conversations.

Architecture

1. Lead messages you on WhatsApp (from a Click-to-Chat ad, your website, or a QR code).
2. WhatsApp Cloud API sends a webhook to Make.
3. Make logs the message in Supabase/Airtable (conversation history).
4. GPT-4o with a qualification system prompt generates the reply.
5. Make sends the reply through WhatsApp Cloud API.
6. Once qualification criteria are met (budget, timeline, decision maker), the bot writes the lead to CRM and sends a calendar link.

Step 1: WhatsApp Business Cloud API

Skip Twilio — Meta ships its own Cloud API, cheaper and direct. Setup:

• Create a Business app at developers.facebook.com.
• Add the "WhatsApp" product.
• You get a test number and 1,000 free conversations/month in sandbox.
• For production: verify your business number via WhatsApp Business Manager.
• Pricing: marketing conversation ~$0.025, utility ~$0.008 (varies by market, 2026 rates).

Step 2: Webhook to Make

In Make, build a scenario starting with "Webhooks → Custom webhook". Paste the URL in Meta Developer Console as the message callback. Subscribe to the messages field.

Gotcha: Meta requires webhook verification. Make handles it automatically, but the first subscription needs a static verify token — set it in "Custom webhook" → Advanced settings.

Step 3: Conversation history in Supabase

GPT has no memory between calls. For the bot to "remember" previous messages, you must pass them in. Simplest: a Supabase table with phone_number, role (user/assistant), content, created_at.

In Make, on every incoming message:

1. Insert the user message into Supabase.
2. Fetch the last 20 messages for that number, sorted chronologically.
3. Map them into the OpenAI API array format.

Step 4: The system prompt — the heart of the bot

Most tutorials fall over here. A weak prompt = a robotic bot that scares leads off. A skeleton that actually works:

You are a sales assistant for [Company]. Your job:
1. Greet naturally.
2. Understand their problem (open questions, not a survey).
3. Gather 4 pieces of info, spread through the conversation:
   - industry
   - scale of problem (numbers, time, money)
   - who decides
   - when they want it solved
4. Once you have all 4, reply with JSON:
   {"qualified": true, "summary": "...", "budget_signal": "...", "urgency": "..."}
5. One question per message. Max.
6. Don't pretend to be human. If asked "are you a bot" — say yes.
7. Never quote prices or timelines — hand off to the calendar.

Step 5: Router — conversation vs qualified

Add a Router after OpenAI. Branch one: if the response contains "qualified": true → write to CRM + send calendar link (Cal.com/Calendly). Branch two: normal reply → send to WhatsApp.

Step 6: Sending the reply

HTTP module in Make, POST to https://graph.facebook.com/v19.0/[PHONE_NUMBER_ID]/messages, with bearer token and body:

{
  "messaging_product": "whatsapp",
  "to": "{{phone}}",
  "type": "text",
  "text": { "body": "{{gpt_response}}" }
}

Monthly cost (1,000 conversations)

• WhatsApp Cloud API: first 1,000 free.
• OpenAI GPT-4o: avg 10 messages/conv × 500 tokens × $0.0025/1K = ~$12.
• Make: Core plan $9.
• Supabase: free tier covers tens of thousands of conversations.
• Total: ~$22/mo for 1,000 conversations.

Pitfalls

• Session window. 24h after the user's last message you can't send free-form — only approved templates. Plan your follow-ups inside the window.
• OpenAI rate limits. Traffic spikes hit 500 req/min. Make retries automatically, but add a 1s "Sleep" module before GPT.
• Price hallucinations. Always instruct "never quote prices". GPT will make them up; leads show up angry.
• Marketing consent. The user's first message is the consent to reply. Don't cold-message — it violates WhatsApp ToS.