Engenharia de requisitos auxiliada por chatbots inteligentes: uma avaliação da qualidade dos artefatos

Abstract

Requirements Engineering is an essential process of collecting, documenting, and maintaining software requirements aligned with business objectives. To perform this role, professionals require specific skills such as effective communication, analytical reasoning, empathy, conflict resolution skills, moderation, self-confidence and persuasion. Unfortunately, there is a significant shortage of specialized professionals with an impact on production qualities, especially the quality of deliveries. Statistics reveal a specific number of failures in software products due to problems in the requirements engineering stage. However, recent research points to the promise of using generative artificial intelligence (AI) as a solution to improve requirements engineering. Artificial intelligence tools have the potential to automate tasks such as requirements gathering and analysis, as well as assist in identifying and correcting ambiguous or incomplete requirements. In this study, our objective was to evaluate how the use of generative AI chatbots can assist requirements professionals, taking into account the exploration of the ideal approach for preparing the prompt and the quality of the resulting artifacts. As a result, the chatbot tool tested proved to be useful for professionals to explore previously unconsidered requirements, highlighting their interest for brainstorming in requirements elicitation. Furthermore, the experiments carried out generated relevant practical recommendations for formatting prompts for requirements specification. However, the quality of the artifacts produced did not show a satisfactory level of completeness, which demonstrates the need for more research to be carried out in order to guarantee better quality. Still, the majority of professionals participating in the survey (63%) agree that the tool can be integrated into their daily practice. Therefore, it is expected that the contribution of this work will serve to improve practices in the area of Requirements Engineering when employing new AI technologies.