Artificial Intelligence (AI) is reshaping society by enabling new opportunities in data processing and interpretation. However, the trustworthiness of AI systems – especially opaque “black box” models like ChatGPT – remains uncertain. As trust is subjective yet potentially crucial for user acceptance, this thesis examines and replicates a conceptual path model in Trust in AI to provide insights for trustworthy AI design and future research.

A quantitative, cross-sectional survey was conducted among 98 UBS professionals. Using structural equation modelling (SEM), the study replicated Karg’s (2024) path model based on the Trust in Automation framework. Constructs included Propensity to Trust, perceived system performance, process transparency, purpose, and behavioral Intention to Use ChatGPT:

The study confirms that trust in AI remains a central mediator for user acceptance, especially in applied business contexts. Purpose and propensity to trust were the strongest predictors of AI trust. On the contrary, process transparency showed no significant effect, challenging assumptions about its relevance in black-box systems like ChatGPT. Psychometric limitations, particularly around reverse-coded items, were observed, highlighting the need for adapted instruments in professional environments. For the client, the findings provided actionable insights: to foster AI adoption, focus should be placed on clearly communicating AI purpose, demonstrating system reliability and aligning deployment with users’ trust dispositions. These results support more effective, trust-oriented strategies for integrating AI tools like ChatGPT into business processes.