Capturing high-intent inbound inquiries on WhatsApp and manually transferring contact details into your CRM is slow, error-prone, and costs valuable revenue. When a prospective client reaches out on WhatsApp, their purchase intent peaks within the first five minutes. Delaying response times or asking repetitive, manual qualifying questions often leads to dropped conversations and lost opportunities.
By connecting n8n, the official Meta WhatsApp Cloud API, OpenAI (ChatGPT), and Zoho CRM, you can deploy an autonomous, 24/7 lead qualification agent. The bot answers customer questions directly from your markdown product knowledge base, extracts vital prospect data (Full Name, Email Address, and Phone Number), and instantly logs a qualified enquiry inside Zoho CRM Leads.
This complete, step-by-step tutorial walks you through setting up the external webhook endpoint in n8n, handling Meta's verification handshake, deploying a contextual ChatGPT assistant backed by a knowledge base file, and mapping lead data into Zoho CRM without writing complex backend applications.
The system operates on an event-driven architecture designed to respect Meta's strict 15-second webhook processing limit while maintaining conversational context across multiple messages:
.md) knowledge base file detailing your services, pricing, and qualification criteria.Before beginning the implementation, make sure you have the following assets ready:
gpt-4o or gpt-4o-mini.The Meta WhatsApp Cloud API requires an externally reachable HTTPS endpoint with a valid SSL certificate. Meta will reject raw IP addresses, self-signed certificates, and unencrypted HTTP connections.
Ensure your n8n environment variables define your public URL so that webhook URLs generated by n8n match what Meta accesses:
WEBHOOK_URL=https://n8n.yourdomain.com/
N8N_ENFORCE_SETTINGS_FILE_PERMISSIONS=true
If you are developing locally on Docker or a desktop machine, expose your port (default: 5678) using a secure tunnel such as Cloudflare Tunnel or ngrok:
ngrok http 5678
Copy the resulting HTTPS URL (e.g., https://your-tunnel-subdomain.ngrok-free.app) and use it as your base webhook URL.
Meta requires a two-step webhook setup:
hub.mode, hub.verify_token, and hub.challenge. Your endpoint must validate the token and return the hub.challenge value as plain text with HTTP status 200.whatsapp-webhookConnect the Webhook node to an If node that validates the verification token sent by Meta:
{{ $json.query?.['hub.mode'] }} === 'subscribe' && {{ $json.query?.['hub.verify_token'] }} === 'YOUR_CUSTOM_VERIFY_TOKEN'
When the condition is true, route the execution to a Respond to Webhook node:
={{ $json.query['hub.challenge'] }}When an incoming message arrives via POST, immediately branch to a Respond to Webhook node configured with:
EVENT_RECEIVEDThis guarantees that subsequent slow API calls (such as OpenAI generation or Zoho CRM lookups) will never cause Meta to trigger timeout retries or deactivate your webhook subscription.
After acknowledging the webhook, parse the nested structure of Meta's payload. Incoming WhatsApp messages arrive under the entry[0].changes[0].value.messages[0] path.
{
"object": "whatsapp_business_account",
"entry": [
{
"id": "YOUR_WABA_ID",
"changes": [
{
"value": {
"messaging_product": "whatsapp",
"metadata": {
"display_phone_number": "15550001234",
"phone_number_id": "YOUR_PHONE_NUMBER_ID"
},
"contacts": [
{
"profile": {
"name": "Jane Doe"
},
"wa_id": "12345678901"
}
],
"messages": [
{
"from": "12345678901",
"id": "wamid.HBgL...",
"timestamp": "1730000000",
"text": {
"body": "Hi, I would like to know pricing for enterprise cloud backup."
},
"type": "text"
}
]
},
"field": "messages"
}
]
}
]
}
Add a Code node in n8n to normalize the incoming payload into clean variables:
const body = $input.first().json.body || $input.first().json;
const messageData = body.entry?.[0]?.changes?.[0]?.value?.messages?.[0];
const contactData = body.entry?.[0]?.changes?.[0]?.value?.contacts?.[0];
if (!messageData) {
return [];
}
return [{
json: {
senderPhone: messageData.from,
senderProfileName: contactData?.profile?.name || '',
messageId: messageData.id,
messageType: messageData.type,
messageText: messageData.type === 'text' ? messageData.text.body : '',
timestamp: messageData.timestamp
}
}];
To ensure accurate responses without hallucinations, ground the ChatGPT model using a markdown (.md) file containing your company's product catalogue, pricing tiers, operational policies, and qualifying requirements.
Create this file and save it on your server or host path accessible by n8n:
# Apex Cloud Solutions - Product & Sales Guide
## Products and Services
1. **Cloud Backup Starter**:
- Price: $49/month
- Storage: 1TB encrypted storage, single-node backup, standard email support.
2. **Enterprise Resiliency Suite**:
- Price: $199/month
- Storage: 10TB multi-cloud replication, zero-recovery-time SLA, dedicated account manager.
## Sales & Qualification Policy
- To create a consultation or trial inquiry, collect the customer's Full Name, Business Email Address, and Phone Number.
- If the customer does not provide an email address, politely request it before proceeding.
- We support businesses in North America, EMEA, and APAC.
/data/knowledge/business_knowledge.mdknowledge_contentgpt-4o-mini.Configure the system prompt in your AI node to handle knowledge retrieval and lead extraction simultaneously:
You are the official WhatsApp sales and inquiry assistant for Apex Cloud Solutions.
Use the following markdown knowledge base to answer customer questions:
{{ $json.knowledge_content }}
Your responsibilities:
1. Answer visitor questions politely and accurately using ONLY the knowledge base.
2. If the user expresses interest in a demo, pricing quote, or consultation, collect their:
- Full Name
- Email Address
- Phone Number (default to their WhatsApp sender phone number if verified)
- Specific Enquiry Requirement
3. If any detail is missing, ask for it naturally. Do not ask for all items in a robotic list.
4. When ALL four fields (Full Name, Email Address, Phone Number, Enquiry) are known, generate your response in strict JSON format:
{
"lead_ready": true,
"first_name": "Jane",
"last_name": "Doe",
"email": "[email protected]",
"phone": "+12345678901",
"enquiry_summary": "Inquiry regarding Enterprise Resiliency Suite pricing.",
"chat_reply": "Thank you, Jane! I have registered your inquiry with our enterprise team. An account executive will contact you shortly."
}
5. If the lead is not complete, return:
{
"lead_ready": false,
"chat_reply": "Your natural language response asking the missing question."
}
Add an If node following the AI node to inspect whether all contact details have been collected:
{{ $json.lead_ready }} === true
chat_reply directly to the WhatsApp Message Dispatch node to continue the conversation.Zoho CRM organizes prospective buyers inside the Leads module. Creating a lead requires mapping standard attributes, with Last Name being a mandatory field in Zoho CRM schemas.
https://api-console.zoho.com).ZohoCRM.modules.leads.CREATE, ZohoCRM.modules.leads.READ.Add the Zoho CRM node to the True output of your filter node:
={{ $json.last_name || 'Inquiry' }}={{ $json.first_name }}={{ $json.email }}={{ $json.phone }}WhatsApp={{ $json.company || 'WhatsApp Lead' }}={{ $json.enquiry_summary }}If you prefer using direct REST API calls, use the HTTP Request node in n8n:
https://www.zohoapis.com/crm/v6/Leads
{
"data": [
{
"First_Name": "{{ $json.first_name }}",
"Last_Name": "{{ $json.last_name }}",
"Email": "{{ $json.email }}",
"Phone": "{{ $json.phone }}",
"Lead_Source": "WhatsApp",
"Company": "WhatsApp Inquiry",
"Description": "{{ $json.enquiry_summary }}"
}
],
"trigger": [
"approval",
"workflow",
"blueprint"
]
}
The "trigger": ["workflow"] array ensures any existing Zoho CRM auto-responder emails, lead-assignment rules, or sales-rep alerts execute immediately.
To complete the cycle, send an automated confirmation message back to the customer's WhatsApp account confirming their inquiry ticket.
Add an HTTP Request node in n8n:
https://graph.facebook.com/v21.0/YOUR_PHONE_NUMBER_ID/messagesAuthorization: Bearer YOUR_META_PERMANENT_ACCESS_TOKENContent-Type: application/json
{
"messaging_product": "whatsapp",
"recipient_type": "individual",
"to": "{{ $('Extract Message Data').item.json.senderPhone }}",
"type": "text",
"text": {
"preview_url": false,
"body": "{{ $json.chat_reply }}"
}
}
You can import this complete workflow JSON into your n8n workspace via Workflows > Import from File / Clipboard:
{
"name": "WhatsApp Bot with AI Knowledge Base to Zoho CRM Leads",
"nodes": [
{
"parameters": {
"httpMethod": "POST",
"path": "whatsapp-meta-webhook",
"responseMode": "responseNode",
"options": {}
},
"id": "node-webhook-meta",
"name": "WhatsApp Webhook",
"type": "n8n-nodes-base.webhook",
"typeVersion": 2,
"position": [200, 300],
"webhookId": "whatsapp-meta-webhook"
},
{
"parameters": {
"conditions": {
"options": {
"caseSensitive": true,
"typeValidation": "strict"
},
"conditions": [
{
"id": "cond-meta-verify",
"leftValue": "={{ $json.query?.['hub.mode'] }}",
"rightValue": "subscribe",
"operator": {
"type": "string",
"operation": "equals"
}
}
]
}
},
"id": "node-if-verification",
"name": "Is Verification?",
"type": "n8n-nodes-base.if",
"typeVersion": 2.2,
"position": [420, 200]
},
{
"parameters": {
"respondWith": "text",
"responseBody": "={{ $json.query['hub.challenge'] }}",
"options": {
"responseCode": 200
}
},
"id": "node-respond-challenge",
"name": "Return Meta Challenge",
"type": "n8n-nodes-base.respondToWebhook",
"typeVersion": 1.1,
"position": [640, 100]
},
{
"parameters": {
"respondWith": "text",
"responseBody": "EVENT_RECEIVED",
"options": {
"responseCode": 200
}
},
"id": "node-respond-ack",
"name": "Acknowledge Event",
"type": "n8n-nodes-base.respondToWebhook",
"typeVersion": 1.1,
"position": [640, 300]
},
{
"parameters": {
"jsCode": "const body = $input.first().json.body || $input.first().json;\nconst msg = body.entry?.[0]?.changes?.[0]?.value?.messages?.[0];\nconst contact = body.entry?.[0]?.changes?.[0]?.value?.contacts?.[0];\n\nif (!msg || msg.type !== 'text') {\n return [];\n}\n\nreturn [{\n json: {\n senderPhone: msg.from,\n senderName: contact?.profile?.name || '',\n messageText: msg.text.body,\n messageId: msg.id\n }\n}];"
},
"id": "node-extract-payload",
"name": "Extract Message Data",
"type": "n8n-nodes-base.code",
"typeVersion": 2,
"position": [860, 300]
},
{
"parameters": {
"filePath": "/data/knowledge/business_knowledge.md",
"options": {}
},
"id": "node-read-knowledge",
"name": "Read Markdown KB",
"type": "n8n-nodes-base.readWriteFile",
"typeVersion": 1,
"position": [1080, 300]
},
{
"parameters": {
"model": "gpt-4o-mini",
"messages": {
"values": [
{
"role": "system",
"content": "=You are the sales assistant. Reference knowledge base:\n{{ $json.data }}\n\nExtract Full Name, Email Address, and Phone Number for enquiries.\nIf all contact details are known, reply strictly in JSON:\n{\"lead_ready\": true, \"first_name\": \"...\", \"last_name\": \"...\", \"email\": \"...\", \"phone\": \"...\", \"enquiry_summary\": \"...\", \"chat_reply\": \"...\"}\nOtherwise reply:\n{\"lead_ready\": false, \"chat_reply\": \"...\"}"
},
{
"role": "user",
"content": "={{ $('Extract Message Data').item.json.messageText }}"
}
]
},
"jsonOutput": true
},
"id": "node-openai-chat",
"name": "OpenAI ChatGPT Agent",
"type": "n8n-nodes-base.openAi",
"typeVersion": 1.4,
"position": [1300, 300],
"credentials": {
"openAiApi": {
"id": "YOUR_OPENAI_CREDENTIAL_ID",
"name": "OpenAI account"
}
}
},
{
"parameters": {
"conditions": {
"options": {
"caseSensitive": true,
"typeValidation": "strict"
},
"conditions": [
{
"id": "cond-lead-ready",
"leftValue": "={{ $json.message?.content?.lead_ready || $json.lead_ready }}",
"rightValue": true,
"operator": {
"type": "boolean",
"operation": "equals"
}
}
]
}
},
"id": "node-if-lead-complete",
"name": "Is Lead Qualified?",
"type": "n8n-nodes-base.if",
"typeVersion": 2.2,
"position": [1520, 300]
},
{
"parameters": {
"resource": "lead",
"operation": "create",
"lastName": "={{ $json.message?.content?.last_name || 'Lead' }}",
"additionalFields": {
"firstName": "={{ $json.message?.content?.first_name }}",
"email": "={{ $json.message?.content?.email }}",
"phone": "={{ $json.message?.content?.phone || $('Extract Message Data').item.json.senderPhone }}",
"leadSource": "WhatsApp",
"company": "WhatsApp Prospect",
"description": "={{ $json.message?.content?.enquiry_summary }}"
}
},
"id": "node-zoho-create-lead",
"name": "Create Zoho CRM Lead",
"type": "n8n-nodes-base.zohoCrm",
"typeVersion": 1,
"position": [1760, 200],
"credentials": {
"zohoOAuth2Api": {
"id": "YOUR_ZOHO_CREDENTIAL_ID",
"name": "Zoho CRM account"
}
}
},
{
"parameters": {
"method": "POST",
"url": "https://graph.facebook.com/v21.0/YOUR_PHONE_NUMBER_ID/messages",
"sendHeaders": true,
"headerParameters": {
"parameters": [
{
"name": "Authorization",
"value": "Bearer YOUR_META_PERMANENT_ACCESS_TOKEN"
},
{
"name": "Content-Type",
"value": "application/json"
}
]
},
"sendBody": true,
"bodyParameters": {
"parameters": [
{
"name": "messaging_product",
"value": "whatsapp"
},
{
"name": "to",
"value": "={{ $('Extract Message Data').item.json.senderPhone }}"
},
{
"name": "type",
"value": "text"
},
{
"name": "text.body",
"value": "={{ $json.message?.content?.chat_reply || $json.chat_reply }}"
}
]
}
},
"id": "node-send-whatsapp",
"name": "Send WhatsApp Reply",
"type": "n8n-nodes-base.httpRequest",
"typeVersion": 4.2,
"position": [2000, 300]
}
],
"connections": {
"WhatsApp Webhook": {
"main": [
[
{
"node": "Is Verification?",
"type": "main",
"index": 0
}
]
]
},
"Is Verification?": {
"main": [
[
{
"node": "Return Meta Challenge",
"type": "main",
"index": 0
}
],
[
{
"node": "Acknowledge Event",
"type": "main",
"index": 0
}
]
]
},
"Acknowledge Event": {
"main": [
[
{
"node": "Extract Message Data",
"type": "main",
"index": 0
}
]
]
},
"Extract Message Data": {
"main": [
[
{
"node": "Read Markdown KB",
"type": "main",
"index": 0
}
]
]
},
"Read Markdown KB": {
"main": [
[
{
"node": "OpenAI ChatGPT Agent",
"type": "main",
"index": 0
}
]
]
},
"OpenAI ChatGPT Agent": {
"main": [
[
{
"node": "Is Lead Qualified?",
"type": "main",
"index": 0
}
]
]
},
"Is Lead Qualified?": {
"main": [
[
{
"node": "Create Zoho CRM Lead",
"type": "main",
"index": 0
}
],
[
{
"node": "Send WhatsApp Reply",
"type": "main",
"index": 0
}
]
]
},
"Create Zoho CRM Lead": {
"main": [
[
{
"node": "Send WhatsApp Reply",
"type": "main",
"index": 0
}
]
]
}
}
}
Do not rely on the 24-hour temporary token displayed on the Meta Developer Portal's Getting Started screen. Create a System User inside your Meta Business Manager:
whatsapp_business_messaging and whatsapp_business_management permissions.Meta Cloud API allows free-form session messages within 24 hours of the user's latest incoming message. If your team or bot responds after 24 hours, free-form text will fail with HTTP error 131047. In that scenario, you must send an approved Meta Template Message.
To prevent duplicate records when an existing customer sends multiple WhatsApp inquiries:
upsert API parameter or query existing leads by Email / Phone prior to creating a record.The WhatsApp Business App is designed for manual communication by single operators on mobile devices or desktop clients, limited to 4 linked devices. The WhatsApp Cloud API is Meta's enterprise developer interface that supports unlimited automated programmatic interactions, webhook delivery, AI bot integration, and multi-agent routing through automation platforms like n8n.
Meta validates webhook URLs in real time. Rejection typically happens because:
hub.challenge as raw plain text with HTTP status 200.Meta terminates webhook connections if your server does not reply within 15 seconds. In n8n, set the Webhook node's Response Mode to Using 'Respond to Webhook' Node and place a Respond to Webhook node immediately after the trigger to return HTTP 200 (EVENT_RECEIVED). This decouples webhook acknowledgment from slower downstream operations like OpenAI inference or Zoho CRM updates.
You can read local markdown or text files using n8n's Read / Write Files from Disk node and pass the binary or string output directly into the system message prompt of the OpenAI node. For large multi-page PDF documents, use n8n's LangChain Document Loader nodes combined with an in-memory vector store or Qdrant/Pinecone instance for Retrieval-Augmented Generation (RAG).
Yes. You can run n8n on localhost and expose port 5678 to the public internet using secure reverse tunneling tools like Cloudflare Tunnel or ngrok. Configure the resulting HTTPS URL as the WEBHOOK_URL in your n8n settings and supply it to the Meta Developer dashboard.
To create leads through the Zoho CRM API, your connected OAuth credential or Zoho API client must have the ZohoCRM.modules.leads.CREATE and ZohoCRM.modules.leads.READ scopes enabled. In addition, ensure the authenticating Zoho user profile has write permissions on mandatory fields such as Last Name.
By utilizing n8n's AI Agent node coupled with a Window Buffer Memory node (or an external Redis session store), the bot retains multi-turn conversational context. Each incoming message updates the conversation history, allowing the LLM to identify which fields (Full Name, Email, or Enquiry) remain incomplete and ask natural follow-up questions.
Meta delivers media messages with distinct types (image, audio, document). In your payload parsing node, inspect msg.type. For voice notes (audio/ogg), you can fetch the media file from Meta Graph API using its media ID and transcribe it using OpenAI's Whisper model in n8n before passing the text to ChatGPT.
Yes. n8n natively provides an Ollama Model node. You can run open-source models such as qwen2.5-coder, mistral, or llama3.1 on your local hardware or VPS. Simply replace the OpenAI Chat Model sub-node with the Ollama node and configure your Ollama endpoint URL.
Running this setup is cost-efficient:
gpt-4o-mini costs roughly $0.0003 to $0.001 per inquiry.Implementing enterprise-grade WhatsApp bots, real-time lead qualification workflows, and seamless CRM synchronizations can dramatically accelerate your sales pipeline and eliminate dropped opportunities. If you need expert assistance designing custom workflows, configuring private LLM agents, or integrating your tech stack, feel free to connect with me for specialized n8n services tailored to your business goals.
IT & Business Enablement Leader
IT & Business Enablement Leader and Full Stack Developer with 15+ years of experience delivering scalable, high-performance digital solutions across Pakistan and the UAE. Specialising in React, Next.js, AI integrations, and workflow automation.
Build a modern, decoupled CMS architecture with Next.js. Learn integration patterns, caching, ISR, previews, and webhooks for Sanity and Strapi at scale.
Stop signal loss. Set up Server‑Side GTM with a custom domain, GA4, and Consent Mode v2 to protect first‑party analytics and reduce ad‑blocker impact in 2026.