In Machine Learning (ML), heuristic search—though essential for managing computational complexity—may often yield results that appear unpredictable, opaque, or inconsistent, particularly to nonexpert users. Nevertheless, most studies focus on improving algorithmic interpretability, with comparatively less attention devoted to whether direct instruction in non-analytical search methods can enhance enduser understanding of outputs produced even by eXplainable AI (XAI) systems. This study introduces KIWI (Knowledge Inferencing Web-based Interface), a gamified platform designed to strengthen AI literacy, by immersing Educational Technology (EdTech) practitioners in diverse (sub)optimization strategies. Using the Traveling Salesman Problem (TSP) analogy as a conceptual framework, it engages users with different search techniques ranging from brute-force (exhaustive) enumeration to advanced meta-heuristic paradigms (evolutionary algorithms). Field-tested in Cyprus with 90 in-service educators, KIWI was evaluated using a mixed-methods approach, assessing perceived usability (System Usability Scale – SUS), cognitive demand (Cognitive Load Questionnaire – CLQ), and search efficiency (analytics on accuracy and engagement patterns). Preliminary findings suggest that KIWI demonstrates strong usability, sustains cognitive engagement, and fosters modest yet measurable improvements in heuristic reasoning and problem-solving skills, with implications for inclusive adoption schemes that extend beyond system transparency to support practitioner engagement with the internal mechanics of AIEd. More broadly, this research contributes to an ongoing longitudinal investigation into the interplay between perceived trustworthiness and experiential trust in user-centered AI EdTech.