This paper researches efficient techniques for retrieving precise information from similar documents in AI chatbot systems, particularly in enterprises with large document collections. Artificial Intelligence (AI) chatbots, powered by Large Language Models (LLMs), are increasingly used for customer service, with Retrieval Augmented Generation (RAG) employed to mitigate hallucinations. However, document types like terms and conditions or policy documents frequently contain highly similar content, leading to ambiguity in retrieval results. Therefore, we propose efficient search techniques using extended metadata and separate vector databases. This technique improves accuracy in retrieving relevant information. We experimented with this technique using a prototype chatbot service, demonstrating its potential to enhance chatbot performance in business applications.