While clinical trials show that computer-aided detection (CADe) has the potential for increasing adenoma detection rates, real-world practice has produced inconsistent improvements. Many endoscopists report frustration with false positive notifications and workflow disruptions yet remain optimistic about its future use. This paradox challenges established information systems theories of post-adoption use. This study explores: (1) The role of expertise in endoscopists’ adoption of CADe technologies, and (2) How expertise influences the reinvention of CADe in practice. 

We adopt a qualitative research design based on 13 semi-structured interviews with Canadian endoscopists having used a CADe platform. Participants represent a spectrum of experience levels from early-career to senior clinicians. Interviews explored motivations for adoption, experiences with CADe integration, perceptions of reliability, and workflow adaptations. Data were transcribed and coded using NVivo, with analysis guided by abductive reasoning. Themes include cognitive learning, and information systems (IS) use. The analysis draws on absorptive capacity, exploratory vs. exploitative technology use, and technology reinvention theories to examine how expertise influences post-adoption adaptation. 

1) The influence of clinical expertise on CADe adoption and application: More experienced endoscopists tend to be more resistant to using CADe with only 1/5 with >20 years practice continuing to use the system versus 7/8 with <15 years. Moreover, the 3 novice endoscopists (< 3 years practice) are more likely to follow AI-generated prompts rigidly (including unnecessary resections of diminutive rectosigmoid hyperplastic polyps). In contrast, the 5 more experienced endoscopists (>10 years) adapt the system to fit their own needs and workflows, demonstrating a process of "reinvention" in practice. Overall, those with less experience use CADe as prescribed in their practice, while more experienced endoscopists envisioned different roles for the technology than its principal intended use. Clinicians also actively modulate the influence of CADe, negotiating a balance between trust in AI and professional judgment.

These findings will refine technology reinvention theory by foregrounding expertise as a key determinant of user-driven adaptation, contributing to post-adoption information systems research, technology reinvention theory, and provides actionable insights in attempting to optimize CADe integration into practice.