AI Sales Agent for WhatsApp (RAG with Product PDFs)

Unlock 24/7 Sales & Support with this AI Agent

This workflow transforms your static product PDFs into a dynamic, conversational AI Sales Agent on WhatsApp. It uses a Retrieval-Augmented Generation (RAG) approach, allowing the agent to provide factual, context-aware answers based solely on your provided documents. Customers can ask questions in natural language and get instant, accurate responses about product specs, features, and pricing, directly within their favorite messaging app.

Key Features & Benefits

• AI-driven Q&A: The agent uses OpenAI's GPT-4o to understand and respond to user queries.
• Factual & Secure: Leverages a vector store built from your PDF, ensuring the agent's answers are grounded in your data, preventing hallucinations.
• Stateful Conversations: Remembers the context of each conversation per user, providing a natural and helpful chat experience.
• 24/7 Availability: Automates responses to common questions anytime, freeing up your human agents for more complex interactions.
• Easy to Adapt: Simply swap the source PDF to create a knowledge base for any product line, user manual, or FAQ document.

• An n8n instance (Cloud or self-hosted).
• OpenAI API Key with access to a chat model (e.g., gpt-4o) and an embedding model (e.g., text-embedding-3-small).
• A Meta for Developers account with a configured WhatsApp Business App and phone number.
• A product brochure or knowledge base in PDF format, accessible via a public URL.

1. Import the workflow into your n8n instance.
2. Create the Knowledge Base (Run Once): a. In the 'get Product Brochure' node, replace the sample URL with a link to your own product PDF. b. Manually execute the workflow using the 'When clicking ‘Test workflow’' trigger. This will download, process, and save your document into the agent's memory (vector store).
3. Configure the AI Agent: a. Select each of the OpenAI nodes ('Embeddings OpenAI', 'OpenAI Chat Model', etc.) and add your OpenAI API credential. b. In the 'AI Sales Agent' node, review and customize the system prompt to define your agent's persona and instructions.
4. Connect WhatsApp: a. Configure the 'WhatsApp Trigger' node with your WhatsApp Cloud API credentials and webhook verification. b. Configure the 'Reply To User' node with the same credentials and your Phone Number ID.
5. Activate Workflow: Enable the workflow to go live. Your AI agent is now ready to answer customer questions on WhatsApp.