The advent of Generative Artificial Intelligence (GenAI) has led to a rapid expansion of AI use in educational contexts. This new technology is profoundly transforming students’ learning experiences. However, a limitation of current research is its predominant focus on how AI influences student performance, concentrating on quantitative indicators such as test scores. In contrast, research on how AI tools are perceived and used by students remains limited, and it is still unclear how collaboration with AI supports students’ knowledge-building processes. This study addresses this gap by exploring how students interact with AI and how they feel during these interactions in design processes. A mixed-methods study was conducted to investigate this question, involving 16 students enrolled in a bachelor’s degree program in Industrial Design Engineering. The students were divided into groups and participated in a two-day workshop, during which they worked on product development following the Double Diamond process and supported by KOBI, an AI tool specialized in the field of arts and design. During the workshop, conversations between the students and the AI tool were recorded, along with data collected through questionnaires administered before and during the workshop activities. The data were analyzed through descriptive statistics and qualitative content analysis. The results show that students interact with AI in different ways depending on the cognitive nature of the task. Two main modes of interaction were identified: exploratory and executive. These modes differ in both the form of interaction and the type of request addressed to AI. The findings also show that these different modes of AI use influence students’ perceptions of the tool’s usefulness and lead to different emotional experiences. Overall, the study contributes to the literature on generative AI in education by deepening understanding of how students interact with AI systems. Furthermore, it proposes an interpretive model, inspired by the concept of absorptive capacity, that links the identified modes of AI use to students’ knowledge-building processes.