Exploging Agentic AI for Software Engineering tasks and reasoning about intentions and their potential role for Software Engineering 3.0.
Motivation
The project aims to identify and evaluate effective patterns for AI agents in software engineering tasks, emphasizing not only technical performance but also seamless human–AI collaboration. Building on a systematic literature review, it will analyse which agentic approaches best support developers while preserving their final responsibility for code and design decisions. Particular attention will be paid to developer satisfaction and trust as critical success factors for adoption. Based on the insights gained, the project will design and implement agentic workflows using orchestration tools such as n8n for selected software engineering tasks, including automated test quality assessment and improvement, API documentation generation, and automated correction of code issues detected by SonarQube. These workflows will be integrated into the university’s GitLab instance and applied in student software projects, where multiple teams will use them throughout a semester-long development process. Data collected from technical metrics and user surveys will be analysed to assess both effectiveness and user experience, ultimately leading to empirically grounded best practices for designing and integrating AI-driven workflows in software engineering.
Aim
For intention-based Software Engineering the project seeks to develop a comprehensive understanding of the concept of “intentions” within software engineering and to explore existing classifications and their relationship to traditional requirements and user stories. It aims to clarify how intentions can serve as higher-level abstractions or complements to these established notions. Building on this theoretical foundation, the project will identify and critically analyze published examples of intention-based software engineering artifacts, assessing the advantages and limitations of the proposed approaches. In a final step, selected intentions—potentially focusing on software change intentions—will be used to illustrate how principles of Software Engineering 3.0 can be applied in practice, either conceptually or through a prototype implementation.
Duration 10/2025- 09/2026
Research Partner Siemens AG
Contact Reinhold Plösch