Build a WhatsApp AI Sales Chatbot with n8n That Answers From Your Product Catalog

Turn your product brochure into a WhatsApp AI sales agent that answers customer questions 24/7 using n8n, OpenAI and a vector store.

4 min read
WhatsApp AI sales chatbot answering customers from a product catalog with n8n

WhatsApp is where many customers prefer to ask questions before they buy. The problem is speed: a buyer asking "which speaker works for a 200-person venue?" at midnight won't wait until your sales team logs in. This use case shows how to build a WhatsApp AI sales chatbot in n8n that answers from your own product catalog, so every reply is accurate, on-brand and instant.

Why Automate WhatsApp Sales Conversations

Manual WhatsApp support runs into the same issues at every growing business:

  • Slow replies outside business hours lose ready-to-buy leads.

  • Repetitive questions about specs, sizes and prices eat up sales time.

  • Inconsistent answers when different team members reply from memory.

  • Generic chatbots that guess instead of quoting your real product data.

Want a WhatsApp agent connected to your CRM, inventory or order system? Get a custom n8n workflow from Vyrade.

Download & Inspect the JSON Workflow

If you prefer starting from a ready-made workflow, subscribers can download the preconfigured n8n workflow JSON below and import it directly into their workspace.

The Automated WhatsApp Chatbot Workflow

The template has 28 nodes in two parts. Part one builds the knowledge base; part two runs the chatbot.

Step 1: Import your product brochure

An HTTP Request node downloads your product brochure PDF and an Extract from File node pulls out its text. The demo uses a public loudspeaker brochure; swap in your own catalog, price list or FAQ document.

Step 2: Build the vector store

A Recursive Character Text Splitter breaks the text into chunks, OpenAI embeddings convert each chunk into vectors, and an In-Memory Vector Store saves them. This is retrieval-augmented generation (RAG): the agent searches your documents before it answers.

Step 3: Receive WhatsApp messages

The WhatsApp Trigger listens for incoming messages through the WhatsApp Business Cloud API. A Switch node filters out audio, images and other non-text messages so the agent only handles what it can answer well.

Step 4: Let the AI agent answer

The customer's message goes to an AI Agent using an OpenAI chat model. Window Buffer Memory keeps the recent conversation per user, so follow-up questions make sense. A Vector Store Tool lets the agent query your catalog and quote real specs instead of guessing.

Step 5: Reply on WhatsApp

The agent's answer is sent back to the customer through the WhatsApp node, usually within seconds.

What You Need Before You Start

  • An n8n instance. See the n8n technology page for hosting options.

  • A Meta WhatsApp Business account with a phone number and Cloud API access.

  • An OpenAI API key (see our OpenAI technology overview).

  • A product brochure, catalog or FAQ as a PDF.

The n8n WhatsApp Trigger documentation explains the webhook setup step by step. Follow it carefully, as Meta's app verification is the most common blocker.

How to Use It

  1. Click Test Workflow once to populate the vector store from your brochure.

  2. Activate the workflow to switch on the WhatsApp agent.

  3. Message your business number and check the reply.

  4. Tune the agent's system prompt: tone, upsell rules, when to hand off to a human.

How to Customize the Workflow

  • Go production-ready: replace the in-memory store with a persistent one like Qdrant, Pinecone or Supabase so your catalog survives restarts.

  • Handle voice notes: add a transcription step so customers can speak their questions.

  • Capture leads: log name, number and intent to Google Sheets, HubSpot or your CRM.

  • Human handoff: route pricing negotiations or complaints to a sales rep on Slack.

  • Multiple catalogs: load several PDFs, one per product line.

Who Benefits From This Workflow

E-commerce stores answer pre-sale questions instantly and recover late-night leads. B2B distributors let buyers query technical specs without waiting for a rep. Agencies can deploy one chatbot per client with a different catalog each. Service businesses can swap the brochure for service menus, pricing and booking policies.

The Limitations of Manual WhatsApp Support

A human team can only reply during working hours, and every new product means retraining staff. Answers vary by person, and nobody tracks which questions come up most. An AI agent grounded in your catalog stays consistent, works around the clock and scales with message volume without new hires.

Architecture Overview

n8n orchestrates two flows. The ingestion flow (HTTP Request → Extract → Split → Embed → Vector Store) runs whenever your catalog changes. The conversation flow (WhatsApp Trigger → Switch → AI Agent with memory and vector tool → WhatsApp reply) runs on every message. Need help with a larger rollout? Explore Vyrade's AI agent development services.

Conclusion

A WhatsApp AI sales chatbot built in n8n gives customers instant, accurate answers from your real product data. The template is a fast first step: import it, load your catalog and go live, then layer in CRM sync, voice notes and human handoff as you grow. For more customer-facing automation ideas, browse all Vyrade use cases.

Frequently Asked Questions

Yes. The n8n WhatsApp nodes use Meta's WhatsApp Business Cloud API, which needs a Meta developer app and a verified business number.