Contextual AI Coding Assistants in Professional Practice: A Mixed-Methods Investigation of Perceptions, Benefits, and Experience-Based Challenges

The integration of contextual AI coding assistants, such as GitHub Copilot and ChatGPT, promises productivity gains, yet their adoption depends on complex human factors. This mixed-methods study examines how professional developers perceive, adopt, and integrate these tools, with a focus on experience-related differences. Guided by UTAUT2 and trustin-automation theory, we surveyed 219 developers across four experience tiers and interviewed 14 practitioners. Adoption intentions were associated less with perceived ease of use than with trust calibration centered on output quality. Active verification and strategic integration were positively associated with technical friction, suggesting an exposure effect. Although core perceptions were similar across experience levels, mid-career developers reported stronger performance motivation and friction sensitivity, while senior developers used more refined friction-management strategies. We propose the Trust-Calibration Adoption Framework, extending UTAUT2 by positioning output trustworthiness and experiential learning as central mechanisms. The exploratory findings require replication.