This article addresses a persistent challenge in economics education: students often struggle to reconcile abstract theoretical models with intuitive understanding. To bridge this gap, we propose an augmented pedagogy approach that systematically integrates large language models (LLMs) with visual programming tools (e.g., Mermaid diagrams) through a structured four-phase design framework. Our methodology emphasizes human-in-the-loop content creation, combining pedagogical expertise with AI-generated dialogue and diagrams. In Phase 1, instructors define clear learning objectives and scaffold the economic concepts to be taught. In Phase 2, we apply targeted prompt engineering to elicit instructive dialogues and narratives from LLMs. Phase 3 involves expert validation and iterative refinement of the resulting textual and visual materials, ensuring accuracy and pedagogical clarity. Phase 4 outlines strategies for curriculum integration and sharing of modular resources. We demonstrate the framework with case studies in microeconomics (market adjustment dynamics), macroeconomics (the Keynesian multiplier), and game theory (strategic interdependence). The proposed approach promises to enhance cognitive processing by leveraging dual-coding and reducing cognitive load, while also enabling personalization, ongoing assessment, and open-resource creation. We discuss implications for the teacher’s role (as content designer and facilitator), and potential pitfalls such as technical biases, ethical issues (authorship and integrity), the digital divide, and risks of pedagogical homogenization. Our article concludes with a call for empirical evaluation of learning outcomes and expansion of the framework.