How to Create a WhatsApp AI Customer Service Bot

A whatsapp ai customer service bot for your business meets customers where they already are. WhatsApp has over 2 billion users, and for many markets it is the primary communication channel. I build these for service businesses that get flooded with the same questions on WhatsApp daily.

The setup uses Meta's WhatsApp Business API, a webhook server, and an AI model to generate responses grounded in your business data.

What You Need Before Starting

• A Meta Business account with WhatsApp Business API access
• A verified phone number for WhatsApp Business
• Python 3.8+ with Flask
• An Anthropic or OpenAI API key
• A public webhook URL (ngrok for testing)

Step 1: Set Up the WhatsApp Business API

Register your app in the Meta Developer Portal. Create a WhatsApp Business account and get your access token:

# Your WhatsApp credentials
WHATSAPP_TOKEN=your_access_token
WHATSAPP_PHONE_ID=your_phone_number_id
VERIFY_TOKEN=your_custom_verify_string

Step 2: Build the Webhook Receiver

Meta sends incoming messages to your webhook. Set up verification and message handling:

from flask import Flask, request, jsonify
import requests
import anthropic
import os

app = Flask(__name__)
client = anthropic.Anthropic()
WHATSAPP_TOKEN = os.getenv("WHATSAPP_TOKEN")
PHONE_ID = os.getenv("WHATSAPP_PHONE_ID")

@app.route("/webhook", methods=["GET"])
def verify():
    mode = request.args.get("hub.mode")
    token = request.args.get("hub.verify_token")
    challenge = request.args.get("hub.challenge")
    if mode == "subscribe" and token == os.getenv("VERIFY_TOKEN"):
        return challenge, 200
    return "Forbidden", 403

@app.route("/webhook", methods=["POST"])
def receive_message():
    data = request.json
    for entry in data.get("entry", []):
        for change in entry.get("changes", []):
            message = change["value"].get("messages", [None])[0]
            if message and message["type"] == "text":
                handle_message(message)
    return "OK", 200

Step 3: Generate AI Responses

When a message comes in, send it through your AI model with your business context:

BUSINESS_CONTEXT = """You are a customer service agent for [Your Business].
Hours: Mon-Fri 9am-6pm. Location: [Your Address].
Services: [List your services and pricing].
Answer in the same language the customer writes in.
Keep responses under 300 words. Be friendly and direct."""

def handle_message(message):
    sender = message["from"]
    text = message["text"]["body"]

    response = client.messages.create(
        model="claude-sonnet-4-20250514",
        max_tokens=400,
        system=BUSINESS_CONTEXT,
        messages=[{"role": "user", "content": text}]
    )

    reply = response.content[0].text
    send_whatsapp_message(sender, reply)

Step 4: Send Replies Back

Use the WhatsApp API to send the AI response:

def send_whatsapp_message(to, text):
    url = f"https://graph.facebook.com/v18.0/{PHONE_ID}/messages"
    headers = {
        "Authorization": f"Bearer {WHATSAPP_TOKEN}",
        "Content-Type": "application/json"
    }
    payload = {
        "messaging_product": "whatsapp",
        "to": to,
        "type": "text",
        "text": {"body": text}
    }
    requests.post(url, json=payload, headers=headers)

Step 5: Add Conversation Memory

Store recent messages per user so the bot has context across multiple messages:

import sqlite3

def save_message(phone, role, content):
    conn = sqlite3.connect("whatsapp_chats.db")
    conn.execute(
        "INSERT INTO messages (phone, role, content, timestamp) VALUES (?, ?, ?, datetime('now'))",
        (phone, role, content)
    )
    conn.commit()

def get_history(phone, limit=10):
    conn = sqlite3.connect("whatsapp_chats.db")
    rows = conn.execute(
        "SELECT role, content FROM messages WHERE phone = ? ORDER BY timestamp DESC LIMIT ?",
        (phone, limit)
    ).fetchall()
    return [{"role": r[0], "content": r[1]} for r in reversed(rows)]

What to Build Next

Add template messages for common flows like appointment booking and order status. WhatsApp Business lets you send pre-approved templates outside the 24-hour window, so you can follow up with customers proactively.

Related Reading

• AI in Customer Service - the full landscape of AI-powered support channels
• The Automation Decision Tree - decide which support channels to automate first
• Connecting Two Systems with APIs: A Practical Guide - how webhook integrations work under the hood