This paper presents the design and implementation of an AI assistant based on Retrieval-Augmented Generation (RAG) and Large Language Models (LLMs) to enhance organisational communication, focusing on university-student interaction. The system automates responses to frequently asked questions by leveraging semantic search and a dynamic knowledge base built from real user queries. Topic modelling using Latent Dirichlet Allocation (LDA) was applied to cluster questions into thematic areas, improving relevance and response quality. Deployed in a university setting, the assistant significantly reduced the administrative workload and improved user experience through faster and more accurate information delivery. The proposed architecture is modular, scalable, and adaptable to various organisational contexts, including public institutions and enterprises. This research demonstrates the potential of AI-driven assistants in supporting knowledge management, real-time communication, and digital transformation in complex environments.