Abstract
This paper proposes an approach that employs generative AI, specifically GPT models, to enhance decision-making in software architecture through a sequence of prompt patterns. Five prompt patterns are introduced, each targeting software architects’ specific challenges when navigating complex design decisions. Through a structured and context-aware decision flow, we demonstrate how these patterns can mitigate risks, manage uncertainties, and optimize functional and non-functional requirements. The proposed approach is evaluated in two real-world scenarios and one fictional case, illustrating its practical application in optimizing operations and ensuring scalability, security, and performance. While AI demonstrates transformative potential in aiding architectural practices, we also highlight its limitations, emphasizing the importance of human oversight in validating technical assumptions and avoiding over-reliance on automated tools. This paper contributes to the ongoing dialogue on how AI can be integrated into software architecture to foster more efficient, informed, and context-sensitive decision-making processes for architects.