SAGAI's second edition aims to contribute to the establishment of a community of researchers focused on exploring novel ways of using generative AI to improve both development time and runtime aspects of software architectures. To achieve this, we have the following specific objectives:
- To provide a forum for scientists and practitioners within the software architecture community to reflect on the potential and limitations of generative AI in software architecture;
- To disseminate early results in the field of generative AI for software architecture;
- To explore how generative AI can support different audiences (e.g., software architects, developers, requirements engineers, and end users) across architecture activities (e.g., identification of architecture-significant requirements, architectural design, evaluation, implementation, and governance).
- To stimulate the investigation of generative AI at runtime in software architectures to address quality attributes such as adaptability, fault tolerance, interoperability, operability, and recoverability, among others.
TOPICS OF INTEREST
- GenAI@DT
- Approaches to use generative AI in architecture design and evaluation;
- Generative AI to support architecture knowledge management;
- Generative AI to support the elaboration of architecture scenarios descriptions;
- Generative AI to support the elaboration architecture evaluation protocols;
- Integration and interaction of generative AI methods and tools with existing state of the practice tools;
- Improvement of context knowledge used by co-pilots and LLMs in multiple contexts;
- Approaches to use generative AI in modernization of legacy systems;
- Experiences and ideas on teaching generative AI techniques to software architects;
- GenAI@RT
- Architectures to use generative AI at runtime to implement functional requirements;
- Approaches to use generative AI at runtime to address quality requirements;
- Tactics to deal with the non-deterministic character of generative AI to allow for effective and efficient solutions at runtime.
TYPES OF CONTRIBUTION
- Full papers: complete research approaches and industry experiences (max. 10 pages).
- Short papers: ongoing research initiatives with early results, argued positions, and emerging novel ideas (max. 6 pages).
Submissions proceed via EasyChair.
All submissions must adhere to the IEEE Template for conference proceedings.
PROCEEDINGS
All accepted papers will appear in the ICSA 2026 Companion proceedings.
Fraunhofer Institute for Experimental Software Engineering IESE