n8n Multi-Channel AI Sales Agent: WhatsApp, Facebook, Instagram Auto-Lead & Booking

Route customer messages from WhatsApp, Facebook Messenger, Instagram DM, and n8n Chat through a GPT-4o agent to automatically qualify leads, answer product questions, log to CRM, and book consultations

Who

n8n users with self-hosted instances and Langchain node experience who need automated multi-channel sales follow-up

You get

After a customer initiates conversation on any channel, AI identifies intent, queries product knowledge, logs leads to Airtable, and calls a sub-workflow to book consultation times once interest is confirmed

You need

n8n · supabase · whatsapp

Steps

1. Import template and configure credentials

   Import workflow.json (58 nodes) into your n8n instance. Configure credentials in sequence: OpenAI API key, Airtable personal access token, Supabase project URL and service role key, PostgreSQL connection string, WhatsApp Business Cloud credentials, Meta Graph API (Facebook/Instagram) credentials. All nodes have empty parameters initially and must be filled per your environment.

2. Configure WhatsApp trigger and preprocessing branch

   Set up node WhatsApp (type: whatsAppTrigger) to receive messages. Add Get Lead node to query Airtable for existing leads by phone number. Configure Handle Message Types switch: text messages route to Edit Fields - chat1 to prepare AI input; audio messages flow through WhatsApp Business Cloud (get media URL) → HTTP Request (download) → OpenAI Whisper (transcribe) → Edit Fields - chat2 (prepare transcribed text with lead info); unsupported types trigger Reply To User1 with an unprocessable notification.

3. Configure Facebook Messenger trigger and preprocessing

   Set up Facebook Trigger (Webhook) to receive messages. Add Respond to Webhook - facebook get for webhook verification, Respond to Webhook - facebook post for acknowledgment. Configure If is not echo - facebook to filter out page-sent messages. Add Sales Agent Demo - typing_on to send typing indicator. Configure Edit Fields - facebook to prepare message text, sender ID, and Facebook context.

4. Configure Instagram DM trigger and preprocessing

   Set up Instagram Trigger (Webhook) to receive messages. Add Respond to Webhook - instagram get for verification, Respond to Webhook - instagram post for acknowledgment. Configure If is not echo - instagram to filter out business account messages. Configure Edit Fields - instagram to prepare message text, sender ID, and Instagram context.

5. Configure n8n built-in chat trigger

   Set up When chat message received (Chat Trigger) node. Add Edit Fields - chat to prepare user input and session information, routing to the aggregation node.

6. Configure input aggregation and AI Agent core

   All preprocessed channel data flows through No Operation, do nothing to the AI Agent. Configure AI Agent connected to OpenAI Chat Model node and Postgres Chat Memory node. Per source description, the main agent logic intends to use GPT-4o, but the OpenAI Chat Model node in workflow.json has empty parameters and requires manual model selection after import. You must write the system prompt yourself, setting the agent goal to qualify customers for Paint Protection Film (PPF), Ceramic Coating, or Window Tint.

7. Configure AI tools: knowledge base, CRM, and calendar

   Connect three tool nodes to AI Agent: (1) sales_technique_knowledge (toolVectorStore) → Demo Supabase vector store + Embeddings OpenAI + OpenAI Chat Model1 for technical and sales knowledge queries; (2) CRM Agent (toolWorkflow sub-workflow) → log name, email, service interest to Airtable and update opportunity status; (3) Calendar Agent (toolWorkflow sub-workflow) → book consultation after CRM logging. Note: crmAgent and calendarAgent are external sub-workflows; create them separately and configure their workflow IDs in the toolWorkflow nodes.

8. Configure response routing and channel reply

   AI Agent output text routes through Switch node by original channel: WhatsApp → Reply To User node (note: source node name is 'Reply To User' with double space); Facebook → Facebook Graph API - Sales Agent Demo; Instagram → Instagram Graph API - smb.sales.agent.demo (HTTP Request); n8n Chat → Output - chat node.

9. Configure Airtable form submission branch (optional)

   Set up Airtable Form Submitted Webhook. On receiving data: Airtable node fetches full record → Create Contact in Contacts table → Edit Fields - form prepares notification data → WhatsApp Business Cloud2 sends templated confirmation message to lead.

You are setting up an n8n multi-channel AI sales agent. I have imported the template workflow.json with 58 nodes, but all node parameters are empty. Please help me check and configure in this order:
1. Credentials: Are OpenAI, Airtable, Supabase, PostgreSQL, WhatsApp Business Cloud, and Meta Graph API all ready?
2. Triggers: Confirm WhatsApp/Facebook/Instagram webhook URLs are registered with their platforms and verification nodes respond correctly to challenges.
3. AI Agent: Help me draft or review the system prompt to ensure the agent properly qualifies for PPF/Ceramic Coating/Window Tint intent, calls CRM Agent after collecting name and email, and calls Calendar Agent after confirming a time.
4. Sub-workflows: crmAgent and calendarAgent are currently missing. Help me design minimum viable versions or confirm existing sub-workflow IDs are filled in the toolWorkflow nodes.
5. Testing: Send a test message from the n8n Chat interface to verify the full pipeline.