Turn your WhatsApp number into a helpful sales chat. Customers ask about your products and get clear answers from an AI agent that reads your brochure. It fits teams that share product PDFs and want fast, consistent replies.

Incoming WhatsApp messages start the flow. A message type check lets only text continue and sends a polite note for other formats. The agent uses OpenAI with short term memory per phone number to keep context across replies. A one time load downloads your brochure PDF, extracts the text, splits it into chunks, and builds an in memory search index with OpenAI embeddings. The agent searches this index to find the right facts and then sends a direct answer back on WhatsApp.

To launch, connect WhatsApp Business and add your OpenAI key. Run the manual step once to load your brochure, then activate the workflow. Expect faster answers, lower support load, and more consistent product info over chat. Great for specs, model comparisons, and pre sales guidance. Update the brochure URL any time and rebuild the index when the catalog changes.

• WhatsApp Trigger captures incoming customer messages in real time
• Message type filter routes non text messages to a friendly reply
• One time ingest builds a knowledge base from your PDF using HTTP Request and Extract from File
• Text is split into chunks and loaded into an in memory vector store for fast search
• OpenAI embeddings create vectors from brochure content for relevant matching
• AI agent uses OpenAI chat with session memory per phone number to keep context
• Vector store tool lets the agent search the catalog before it answers
• Replies are sent back to the same WhatsApp user with the agent output

• Reduce manual replies from hours to minutes by answering common catalog questions automatically
• Automate up to 70 percent of repeated product questions through chat
• Improve answer accuracy by using facts from your brochure as the source of truth
• Handle many concurrent WhatsApp chats without adding staff
• Connect WhatsApp and OpenAI in one flow with no extra systems

1. Import the template into n8n: Create a new workflow in n8n > Click the three dots menu > Select 'Import from File' > Choose the downloaded JSON file.
2. You'll need accounts with WhatsApp and OpenAI. See the Tools Required section above for links to create accounts with these services.
3. Open the WhatsApp Trigger node and click Create new credential. Follow the on screen steps to connect your WhatsApp Business account. Make sure the webhook from Meta points to the webhook URL shown in the node and that events for messages are subscribed.
4. Open each WhatsApp Send node and select the same WhatsApp credential. Set your phoneNumberId to the number connected in your WhatsApp Business account.
5. Open the OpenAI Chat Model and Embeddings nodes. Click Create new credential, paste your OpenAI API key from the OpenAI account page, and save. Keep the models gpt 4o 2024 08 06 and text embedding 3 small or choose others you prefer.
6. In the get Product Brochure node, replace the URL with your own product brochure PDF. This is the source that feeds the knowledge base.
7. Click Execute on the manual trigger path to run the ingest once. Confirm that Extract from File produces text and that Create Product Catalogue shows vectors created without errors.
8. Send a text message to your WhatsApp number. Check Executions in n8n to see the AI Sales Agent step and the final reply back to the same user.
9. Send a photo or voice note to verify the non text message reply. Edit that message in the Reply To User1 node if you want different wording.
10. Activate the workflow. If self hosting, ensure your n8n URL is public with HTTPS so WhatsApp can reach your webhook.
11. If answers seem empty or off, confirm the vector store contains documents and rerun the manual ingest after any brochure update. Also check that the Vector Store Tool is connected to the Product Catalogue.
12. If no messages arrive, recheck the WhatsApp webhook subscription in Meta, confirm the phone number id, and verify that the n8n webhook URL matches the trigger node URL.