Optimierung des IT-Supports durch einen KI-gestützten Chatbot

  • Alexander Michael Vollovec

Student thesis: Master's Thesis

Abstract

IT support faces the daily challenge of managing a wide variety of systems. Accessing relevant information can be especially time-consuming for rare or complex requests, as knowledge sources are often scattered and unstructured. The growing popularity of AI chatbots offers a promising solution: they can act as a central interface to a company’s knowledge databases. This study focuses on the development of such a chatbot for IT management systems. The goal is to make IT support processes more efficient, particularly by improving knowledge retrieval and access to relevant information.
Commercial systems are often associated with high costs and privacy risks. A locally operated, open-source based alternative presents an attractive solution. This study demonstrates that powerful results can be achieved through careful data preparation and optimization of Retrieval-Augmented Generation technology. Even with small, locally operated models like Llama 3.1 (8B), it was possible to generate precise answers. An automated performance analysis confirmed the effectiveness of hybrid search methods and the reranker, which further improved the relevance and accuracy of the answers. Additionally, a user study with seven participants was conducted, which also delivered very positive results. The chatbot’s ease of use and the high quality of the generated answers were particularly praised. The results of this work provide a solid foundation for future developments. Next steps include the integration of multimodal content and the optimization of prompt engineering to further enhance the chatbot’s functionality.
Date of Award2024
Original languageGerman (Austria)
SupervisorSebastian Pimminger (Supervisor)

Cite this

'