Artificial intelligence (AI) may be selected before engineers determine whether a task requires inference or the consistent execution of explicit rules. This study investigates how industrial engineering can guide technology selection, assess system-level consequences, and translate these judgements into learning outcomes. It combines an implemented case of rule-based vehicle-transport allocation with a prospective natural-language analytics scenario. The rule-based system began deployment on 15 June 2026. Weekly allocation counts and recorded allocation-stage duration, covering system execution and human handling of referred exceptions, were analysed using confidential, normalised indicators. Relative to 23 pre-deployment weeks, recorded time per allocation was 68.6% lower across 16 weeks after deployment began, and 71.3% lower during July–September (13 weeks). These descriptive comparisons do not quantify employee labour hours released, establish a causal effect, or measure end-to-end transport performance. The prospective scenario illustrates how data coverage, governed metric definitions and human review may constrain an otherwise appropriate use of generative AI. The contribution is a four-step framework for diagnosing tasks, selecting technical mechanisms, evaluating sociotechnical outcomes and assessing student learning. A case-based exercise and assessment rubric are proposed, but their educational effectiveness has not been tested.