Description

(Gen-)AI-based systems are transforming the IT landscape of organizations, with Large Language Models (LLMs) driving efficient information retrieval through their advanced language capabilities. While challenges like the generation of false or fabricated information, commonly referred to as hallucinations, have been mitigated by integrating LLMs with external knowledge repositories, these Retrieval Augmented Generation (RAG) systems provide domain-specific information. However, despite their proven benefits, user adoption of RAG systems remains low. This thesis investigates barriers to adopting LLM-based information retrieval systems in enterprise settings. Conducted in collaboration with hsag (https://hsag.info), a German energy-sector service provider, it focuses on ENNA, a chatbot leveraging curated industry knowledge and internal project documents. Despite past evaluations demonstrating ENNA’s efficiency and effectiveness, user adoption is sparse.

Tasks

1. Identify key barriers to adoption
2. Suggest possible solutions
3. Implement & evaluate one or more solutions

Requirements

• Proficiency in Python/JavaScript ;
• Strong English writing & presentation skills
• Interest in Human-Computer Interaction