Vocational undergraduate classrooms have long faced the dual pull of “academic drift” and “vocational narrowness,” and the intervention of generative artificial intelligence has further complicated this contradiction: the abundance of technological tools does not automatically translate into a deepening of competency cultivation; on the contrary, it may obscure the “logic of competency” with the “logic of efficiency.” Grounded in competency-based education theory and in light of the overlapping “academic” and “vocational” characteristics of vocational undergraduate education, this paper proposes a reconstruction approach for a “competency-based—digitally empowered” dual-wheel-drive teaching model: anchored on a job-competency map, driven by a three-level project chain of “basic functions—compound functions—system integration,” closed by an AI-assisted evaluation of “process-based competency evidence,” and supported by teacher competency transformation and institutional safeguards, so that the use of digital tools is always subject to the test of whether it has promoted the generation of complex vocational competencies. The study argues that the essence of dual-wheel drive is not a simple superposition of “competency” and “technology,” but rather the disciplining of technological logic by competency logic, so as to prevent generative AI from pushing vocational undergraduate teaching into a new trap that “looks efficient but is hollow at its core.”