Skip to main content

Prompt Alchemy: Designing AI Assistants with Prompt Chaining

Presented by James Cantonwine, Kris HagelConference WorkshopFor WERA Conference

Prompt Alchemy: Designing AI Assistants with Prompt Chaining

This advanced workshop, presented by Kris Hagel and James Cantonwine from the Peninsula School District, moves beyond single prompts to teach "prompt chaining"—a powerful technique for designing reliable, multi-step AI assistants and agents for complex educational tasks.

Key Takeaways:

  • Prompt Chaining as the Core Technique: The session introduces "prompt chaining" as a method to break down a complex task into a series of distinct, connected prompts. Unlike a single large prompt, this approach creates a clear, auditable workflow (e.g., Step 1: Clean Data → Step 2: Analyze Patterns → Step 3: Draft Communication), which is ideal for tasks like MTSS data analysis or board summaries.
  • The PSD AI Ecosystem: The presentation showcases the district's integrated approach:
    • psd401.ai (The Hub): The public-facing site for all AI principles, policies, and guidance.
    • AI Studio (The Tool): A self-hosted, open-source, and privacy-first AI platform that allows educators to build their own tools in a secure environment without sending data to third-party vendors.
  • Future-Proofing Skills Over Tool Mastery: The presenters emphasize an agile strategy, focusing on learning transferable logic like prompt chaining, which works across any AI platform (AI Studio, ChatGPT, Claude, etc.). This approach mitigates the risks of vendor lock-in and model updates ("model drift").
  • Hands-On Building Experience: This is a "build-along" workshop. After a live demonstration, participants engage in a hands-on lab to storyboard and construct their own multi-step AI assistant using the AI Studio's "Assistant Architect."
  • Rubric for Evaluating AI Assistants: A key deliverable is a rubric for evaluating chained prompts and AI assistants based on four critical dimensions: Accuracy & Reliability, Transparency, Ethics & Safety, and Impact & Efficiency.

Actionable Insights for Educators and Leaders:

  • Adopt Prompt Chaining for Complex Tasks: For multi-step processes, move away from single "mega-prompts." Instead, break the work into a chain of smaller, verifiable prompts where the output of one step becomes the input for the next.
  • Prioritize Auditable and Transparent Workflows: Focus on creating AI assistants where each intermediate step can be inspected and verified by a human. This is crucial for maintaining accuracy and ethical oversight.
  • Explore Self-Hosted, Privacy-First Solutions: Investigate the potential of a self-hosted platform like PSD's open-source AI Studio to create safer, more controlled environments for working with student-adjacent data.
  • Use the Provided Rubric to Evaluate AI Tools: When building or adopting AI assistants, use the four-part rubric to assess whether they are accurate, transparent, ethical, and truly efficient.
  • Focus on Principles, Not Just Platforms: Center your district's AI strategy on enduring principles like "Human in the Loop" and data privacy, which will remain relevant even as specific AI tools and models change.

Looking Ahead:

"Prompt Alchemy" shifts the paradigm from being a simple AI user to becoming an AI designer. The presentation empowers educators and leaders to move beyond basic chat interactions and start building their own custom, privacy-aware AI assistants that are tailored to solve real-world educational challenges. This approach fosters a culture of innovation, agility, and responsible AI use.