Rethinking AI Design Practices: A Reference Architecture for Generative AI Systems

Generative AI (GenAI) holds the potential to transform various facets of society. While research and practice increasingly explore its diverse use cases and assess its impact, many companies continue to face significant challenges in designing and developing advanced GenAI systems, hindering their ability to derive value from these technologies. To address these challenges and offer guidance on emerging GenAI development practices, we propose a reference architecture that highlights key design decisions. Using a design science research approach, we developed the reference architecture informed by consortium workshops, literature reviews, and expert interviews. The reference architecture is grounded in eight key design requirements and serves as a strategic guide to streamline system design and assist organisational decision-making and communication. Additionally, the architecture contributes to the academic discourse by offering a structured framework for future research on the design peculiarities of GenAI systems, helping bridge the gap between practical implementations and theoretical insights.