The master thesis investigates the application of dialogue systems in conjunction with large language models (LLMs) to optimize requirements engineering practices. The primary objective is to explore the automation of requirements gathering and translation into technical specifications, reducing manual effort and time consumption. A prototype dialogue system, enriched with in-house company knowledge and augmented by access to a large language model is developed to facilitate this automation process. Moreover, the research delves into the theoretical foundations, methodologies, experiments, and discoveries toward a future where large language models assume a pivotal role in automating and augmenting requirements engineering, potentially eliminating the need for extensive expertise. This is mainly achieved by conducting a semi-systematic literature review and the application of design science research (DSR). Moreover, this research is deeply rooted in the context of addressing the persistent challenges within requirements engineering. The manual, time-consuming requirements elicitation processes, coupled with the complexity and evolving nature of requirements, pose significant hurdles in the project lifecycle. Existing literature underscores the importance of automating these processes, highlighting the potential benefits of AI-driven solutions like dialogue systems integrated with LLMs. However, current tools and techniques lack the emotional, conversational and relational capabilities required for a comprehensive "requirements engineering companion". This thesis steps in to fill this gap by exploring modern technologies to develop a prototype customized for AMAG, serving as an intelligent requirements engineering companion.

Following the design science research methodology, first the awareness of the problem is established by conducting a focus group workshop with AMAG. The problem awareness included understanding the challenges and gathering the requirements for the artifact. Based on the literature review and AMAG’s requirements, suggestions were made for a well-suited design of the artifact. In accordance with the suggestions, the prototype was developed following an iterative development cycle.

The evaluation of the artifact demonstrated its high usability, utility, quality and efficacy. All participants from AMAG reported positive experiences using the prototype and expressed high satisfaction with the generated technical requirements in the form of user story. These findings were supported by System Usability Scale (SUS) survey results and interviews, where participants indicated their intention to use the RE-Companion frequently and recognized its ability to motivate critical thinking about requirements. While the prototype showcased significant opportunities to support the requirements engineering practices, there were also some limitations. Those include the need for human oversight to confirm the accuracy and completeness of the generated user story. Therefore, the intelligent requirements engineering companion should be considered as a supportive tool rather than a complete replacement of human expertise. Nevertheless, by using the RE-Companion during the experiment, the participants developed a better understanding of its functionalities and experimented with various inputs.