I built an AI-powered chatbot that answers customer questions using ONLY real data from a business's own website — no hallucinations, no generic AI guesses. Most AI chatbots businesses deploy today have one major flaw: they make things up. A customer asks about a return policy, and the bot confidently gives a wrong answer. That's a trust problem, not just a technical one. So I built a Retrieval-Augmented Generation (RAG) chatbot that solves this directly. 📌 How it works: ✅ Scrapes and processes real website content (products, policies, FAQs) ✅ Converts that content into a searchable knowledge base using vector embeddings ✅ Retrieves only the most relevant information before answering — grounded, not guessed ✅ Transparently says "I don't have that information" instead of fabricating an answer when it's outside its knowledge 🔧 Built with: n8n (workflow automation) | Google Gemini (LLM + embeddings) | Vector Database (RAG retrieval) This is the kind of AI agent businesses actually need for customer support automation — accurate, on-brand, and safe to put in front of real customers. Watch the full demo below I'm currently building hands-on expertise in AI agent development and workflow automation — including chatbots, lead qualification systems, and WhatsApp/CRM integrations using n8n. Currently taking on freelance automation projects — feel free to DM me or check the link in my profile.

Muhammad Usama Manzoor

n8n Automation Developer | AI-Assisted Lead Follow-Up & Work