Abstract:

Agile is an approach to software development that emphasizes flexibility, collaboration, and customer feedback. It focuses on iterative development cycles, where requirements and solutions evolve through the collaborative effort of self-organizing and cross-functional teams. Scrum is a framework used within agile software development but has found application in various other fields as well. It provides a structure for teams to collaborate effectively on complex projects, allowing them to adapt to changes quickly and deliver high-quality products efficiently. New software engineers need to learn to work in teams with other engineers in complex systems. One method to do so is through capstone classes. These capstone classes tend to be led by one or two professors who must take care of at least 5 groups each. In this paper we propose the usage of Large Language Models (LLMs) as Scrum Masters for the groups to light up the load work of professors. LLMs are a type of artificial intelligence (AI) model designed to understand and generate human-like text. These models are built using deep learning techniques, particularly using architectures like transformers, and are trained on vast amounts of text data to learn patterns, relationships, and language structures. Due to the vast amount in knowledge contained within LLMs and their natural language understanding, the development team and product owners could utilize LLMs as their Scrum Master. Scrum Masters oversee facilitating the Scrum process, removes impediments, and ensures the team adheres to Scrum principles and practices. In this paper we show an approach that, through the use of template queries, allows LLMs to capture the work of a Scrum Master during the daily standup meetings. In our approach, every member of the development team will answer the three daily standup questions to the LLM. The LLM will condense the information, e.g., if a developer is facing an issue that is blocking their progress, the LLM will create a concise report of this. In this work we show how the LLM can point out a potential solution. It is envisioned that the professor gets reported on the progress and on anything preventing their progress which the LLM cannot assist on solving. Even though there are still some things that LLMs cannot comprehend nor provide such as empathy, or leadership skills, LLMs may have the necessary technical skills required to successfully carry out Scrum Master related tasks. The main issue LLMs face is that there are some physical impediments that LLMs cannot directly overcome. However, by using the proposed approach, could save time for professors to directly focus on fixing other issues the students are facing.

Keywords:

Large Language Model, Scrum, Education, Project Management, Software Engineering, Agile, Scrum Master.