Download and read the papers before the MiniPLoP

Generative AI is increasingly used in software engineering, yet its integration into Agile workflows remains fragmented and largely limited to isolated features. This paper investigates how generative AI can be systematically structured to support coordination activities with- out undermining Agile principles. We propose a three-layer pattern hier- archy that separates motivation, operational representation, and appli- cation level usage of generative AI. The hierarchy is instantiated using semantic embeddings derived from software engineering artifacts and evaluated through a work-in-progress task-allocation case-study. Using the TAWOS dataset, tasks and developer histories were embedded with CodeBERT and compared using cosine similarity. The historically or manually-assigned developer was frequently ranked among the top rec- ommendations, achieving a Mean Reciprocal Rank of 0.69. Our results demonstratethatsemanticrepresentationsmayprovideameaningfulsig- nal for aligning tasks with project-managers and developers’ experience in Agile environments. Our preliminary results support the contribution and validity of the proposed pattern hierarchy.