The rapid rise of frontier technologies like Large Language Models (LLMs) and Generative AI in Local Digital Twins (LDTs) presents a dual challenge: the potential to improve public administration through enhanced data access and interaction versus concerns over accuracy, governance, and cost. Cities must determine whether these cutting-edge AI tools warrant investment or if traditional, proven methods suffice. The solution involves a thoughtful evaluation framework that prioritizes value propositions over hype. Trainers employ concepts from the Triple Loop Learning framework, ensuring AI adoption drives resilience and decision-making improvements for citizens. This includes leveraging the EU’s Local Digital Twin Toolbox for structuring AI use, particularly LLMs for interaction and Generative AI for synthetic data. Learners will explore how these solutions are taught through practical, case-based exercises. They will apply the Triple Loop framework to assess AI roles in LDTs, practicing with real-world scenarios such as evaluating LLM-generated policy briefs. Emphasis is placed on fostering inclusivity, transparency, resilience, and ethical reflections within AI implementations, ensuring responsible, human-centric governance in smart cities.