Beyond Prompting: Advanced Reusable AI Workflows with Codex and Agentic AI
Paul Gavini (CSIRO)
This workshop is designed for adept users of generative AI who are ready to move beyond basic prompting into structured, reusable AI system design. It is intended for educators already own tools such as ChatGPT Plus, the upcoming EdChat AI (DfE schools only), Claude, or similar platforms, and who want to increase consistency, scalability, and control in their AI-assisted workflows.
While many educators rely on one-off prompts to generate lesson materials, assessments, or resources, this often leads to variability and limited reuse. This session introduces a more advanced paradigm: treating AI interactions as modular systems rather than isolated exchanges.
Participants will explore the design and implementation of reusable AI workflows through Markdown-based instruction files, structured project rules, reusable skills libraries, and agent-based orchestration patterns. While demonstrated in Codex, these approaches are platform-agnostic and transferable to tools such as ChatGPT Plus workflows, EdChat AI environments, Claude, and other emerging educational AI systems.
A key focus is the integration of rubric-driven AI evaluation. Participants will learn how to embed assessment criteria directly into workflows, enabling AI systems to self-evaluate outputs, identify gaps, refine responses, and iterate until predefined quality thresholds are met. This introduces a controlled feedback loop that improves reliability and pedagogical alignment.
Although aimed at adept users, the session provides clear conceptual scaffolding for those transitioning from intermediate practice. The emphasis is on developing durable, portable skills that can be applied across multiple AI ecosystems rather than reliance on any single tool.
By the end of the workshop, participants will have a practical, advanced framework for designing, deploying, and refining AI workflows that are consistent, auditable, and reusable. They will be equipped to apply these systems immediately to complex professional tasks such as curriculum design, assessment development, differentiated resource creation, and subject-specific content generation—while remaining adaptable to future AI platforms and capabilities.