Background

Generative AI systems have rapidly transformed from experimental tools to essential workplace companions. ChatGPT, Claude, and Gemini are no longer just impressive demos—they’re becoming integral to how we work, think, and solve problems. As organizations worldwide integrate AI assistants into their workflows, a critical question emerges: How do people actually interact with these systems in real-world settings?

While we know AI adoption is accelerating, we have surprisingly little insight into the nuanced ways employees engage with chatbots day-to-day. What tasks do they tackle? How do they craft their prompts? Which interactions lead to satisfaction versus frustration? Understanding these patterns is crucial for improving AI systems and helping users collaborate more effectively with AI.

This thesis offers a unique opportunity: You’ll work with exclusive real-world data from our industry partner Trelleborg, analyzing actual employee interactions with enterprise chatbots. This isn’t simulated data or lab experiments—it’s authentic workplace AI usage that has never been studied before.

In this thesis you will:

• Apply cutting-edge language models to automatically analyze, cluster, and classify human-chatbot interactions at scale
• Design and implement a configurable analysis pipeline that can adapt to different types of interaction data
• Discover novel insights about how employees integrate AI assistants into their workflows, including task patterns, prompting strategies, and adoption behaviors
• Evaluate user satisfaction and identify factors that lead to successful versus unsuccessful AI interactions
• Generate actionable recommendations for improving AI assistant design and user training programs

Research impact

Your findings will have immediate practical applications:

• User empowerment: Develop evidence-based guidelines for more effective AI collaboration and prompt engineering. 
• Model improvement: Create targeted synthetic training data and fine-tuning strategies based on real usage patterns.
• Methodological advancement: Establish best practices for automated analysis of human-AI interaction data and validate automated approaches that can uncover meaningful patterns from thousands of real conversations.

We are looking for candidates who:

• Passionate about AI’s future: You’re excited by generative AI and natural language processing’s potential to transform work
• Technical foundation: Experience with Python and machine learning frameworks (transformers, pandas, scikit-learn, etc.)
• Self-motivated researcher: You thrive on tackling real-world problems independently while contributing creative ideas
• Strong communication: Excellent English skills for writing and presenting your research

Details

Start: October 2025
Duration: 6 months
Language: English
Industry collaboration: Direct access to Trelleborg’s interaction data and domain experts

How to Apply

We offer you a cutting-edge research topic at the intersection of AI and human-computer interaction, close mentorship from experienced researchers, and the rare opportunity to work with proprietary enterprise data. You’ll develop both theoretical insights and practical skills that are highly valued in today’s AI-driven job market.

Ready to dive into the future of human-AI collaboration?
Please send your current transcript of records, a short CV, and a brief motivation (3-4 sentences) to:

• daniel.hendriks∂kit.edu
• jonas.liebschner∂kit.edu