Interdisciplinary, project-based learning in higher education requires student teams to self-organize, collaborate across disciplines, and engage in structured reflection – challenges that intensify in digitally mediated contexts. To facilitate these processes, a generative AI-based coaching chatbot grounded in systemic coaching principles was developed and deployed during a 2024 interdisciplinary project week at TH Köln – University of Applied Sciences, engaging 149 participants from diverse disciplines. The coaching chatbot supported daily team reflections using resource- and solution-focused questions. A mixed-methods approach was chosen to evaluate acceptance, integrating a questionnaire at the start and end of the intervention with a qualitative analysis of chatbot conversation histories and open-ended student feedback. Quantitative results indicate that chatbot usability was rated consistently high, while attitudes toward social influence and privacy-related concerns remained stable over the intervention period. Notably, perceived usefulness showed a statistically significant decline, suggesting unmet expectations regarding the chatbot’s practical benefits. Qualitative findings highlighted the value of the chatbot’s structured and accessible facilitation, but also revealed limitations in conversational depth, contextual adaptivity, and technical reliability. The group-based interaction further limited opportunities for quieter members to be directly included. Taken together, these results suggest that while AI-based coaching chatbots can effectively support structured team reflection and lower usability barriers, their sustained educational impact will depend on clearer communication of system capabilities, inclusive design, improved adaptivity to group needs, and stronger transparency in data handling to foster user acceptance in interdisciplinary learning environments.