Artificial Intelligence (AI) is increasingly integrated into enterprise document management systems, with automatic summarization emerging as a valuable feature to help users quickly extract key information. While this enhances efficiency and reduces information overload, trust in AI-generated content remains a significant concern, especially in multilingual, multi-domain enterprise contexts where precision and transparency are essential.

This study adopts a qualitative, exploratory approach, guided by real-world interaction with a prototype built using the Mistral AI model. It investigates how users perceive trust in AI-generated summaries, focusing on accuracy, domain adaptation, and transparency. A prototype using abstractive summarization was developed as a foundational proof of concept for potential integration into Centuri’s document management system.

10 participants from various sectors provided real documents, which were summarized in both short and detailed formats. Participants then took part in semi-structured interviews, and their responses were transcribed and thematically analyzed.

The findings suggest that users are more likely to trust summaries that are clear, complete, and aligned with domain-specific terminology and structure. Transparent content selection, human-like reasoning, and explainable outputs were key factors influencing perceived trustworthiness. While many participants appreciated the efficiency and overall quality of the summaries, some expressed concerns about missing context and unclear summarization logic.

The study concludes with practical recommendations for improving trust in AI summarization tools and highlights the need for further research to validate these findings across larger and more diverse enterprise contexts.