Abstract
Software architecture decisions are critical in a software project to ensure that systems meet functional and non-functional requirements. Microservice architectures have become popular in the industry, having a high amount of material available that was used in the training of large language models (LLMs). This paper explores the use of generative AI tools, such as ChatGPT, guided by a prompt pattern sequence to support architectural decision-making in microservices architectures. The proposed approach aims to provide structured guidance to software architects, helping them navigate in complex design challenges. To evaluate the prompt sequence, we conducted studies that revisited important architectural decisions made by large companies in the context of microservices architectures. Two industry case studies are presented: one involving the management of a large set of components in a financial institution, and the other focused on the front-end approach for a large-scale e-commerce platform in a pharmaceutical chain. The results demonstrate how five distinct prompt patterns deliver actionable insights tailored to each project’s unique technical and business constraints, enabling more informed decision-making. Retrospective feedback from architects highlights the effectiveness of the proposed prompt pattern sequence, which proposed solutions aligned to what was actually implemented. The findings suggest that generative AI, guided by well-structured prompt patterns, can support the decision-making process in microservice architectures.