Document Type
Article
Publication Date
3-9-2026
Abstract
As large language models become broadly accessible, questions about how to integrate them productively into engineering education have moved from speculative to urgent. This paper presents the design rationale and early deployment experience of 247officehours.com, a faculty-created, faculty-managed platform that grounds AI tutoring in instructor-selected course materials and connects it to assessment, engagement tracking, and targeted instructional intervention. Drawing on deployment across undergraduate and graduate engineering courses, we describe six core platform features, analyze how student interactions map onto Bloom's revised taxonomy, and identify recurring patterns of productive and unproductive AI use. A tutor–tutee inversion activity, in which students must identify and correct deliberate chatbot misconceptions, emerges as a particularly effective structure for driving close reading and deepening conceptual engagement. Early observations suggest the platform may reduce the probability of very poor course outcomes more than it raises the ceiling of high performance. We discuss implications for instructional design, AI fluency as a professional formation goal, and the limits of current evidence.