This research explores the way to automate and improve software architecture design by utilizing Artificial intelligence (AI) based approaches, i.e., Large Language Models (LLMs) and Retrieval Augmented Generation (RAG). This study examines how AI can help formalize the architecture generation process by automating requirement interpretation, solution options creation, and stakeholder-in-the-loop iterative improvement given complex requirements & limitations resulting from resource-intense traditional design process. With their self-attention mechanisms and reinforcement learning characteristics, LLMs can provide powerful support in managing complex dependencies across requirements. RAG facilitates the integration of an external knowledge base to ensure contextual relevance and reduce design time by up to 40% for a more cost-effective solution than ever before. The preliminary findings suggest that the AI can be effective in facilitating an agile design of high-quality software architecture at low costs and deal with ethical considerations such as data privacy, transparency, and equality. Continued efforts will be made to enhance the accuracy of AI retrieval, explainability of retrieved results and validation procedures to further develop adaptable solutions that leverage AI’s capabilities on complex architectural projects.