Day 2: The Next Frontier, Creating Custom AI Tools for Your Classroom

This hands-on learning lab is designed to help educators move beyond simply using AI tools and begin designing customized AI supports for their own teaching practice. Participants will choose between two differentiated learning pathways:

Beginner Builder Track — for educators new to creating AI tools and interested in learning foundational prompting and Google Gem design skills.
Advanced Agent Design Track — for educators with prior experience using tools such as Google Gems, Custom GPTs, Claude Projects, or multi-step prompting workflows.

Presenters

Clare Berends-Eitel: Director of Learning Design, Digital Futures Institute
Joel Lederer: High School Educator, Brooklyn Technical High School

Objectives

Analyze classroom challenges to determine which instructional and planning tasks are appropriate for AI-supported workflows.
Design customized AI supports using prompting strategies, structured workflows, and educator-created knowledge resources.
Evaluate the ethical, instructional, and practical limitations of AI systems when building tools for teaching and learning.

To earn the Badge for this session, you must:

Attend the Learning Lab
Submit your Learning Lab Journal
Interact with the Padlet below

Day 2: Learning Lab: Shared Foundations Session

AI Design Journal

Utilize this AI Design Journal throughout this session. At the bottom of your journal, be sure to read through and fill out the submission checklist before submitting it below.
- Your AI Design Journal

Workshop Slides

TST: AI for Everyday Teaching Tasks
What AI Can Reliably Help You Build in the Classroom — and What It Can't (Yet)

Build Tracks - Beginner Build

Session Materials

Getting Started with Google Gems
Edu Gems

Build Tracks - Advanced Build

Session Materials

Audience: Educators already comfortable building custom systems — familiar with Gems, Custom GPTs, or Claude Projects, and with multi-step (multi-shot) prompting.

Focus: Systems thinking + context engineering.

Topics to cover

From prompting to context engineering — managing everything the model sees, not just wording one request; moving from a single prompt to a system you can trust
What's worth building — the three tiers (closed/bounded / open)
The research-first method — Identify → Research → Distill → Build; why grounded, domain-specific agents outperform generic ones
Giving an agent memory — the foundation files (orientation, state, continuation) that keep sessions from starting over
Knowledge documents & conditional loading — what to put in an agent's knowledge section to make it flexible and task-specific, and loading only what a task needs
Keeping an agent accurate — structured instructions (the agent writes the XML/markdown, not you), anti-sycophancy, and the Recursive Protocol two-pass review to prevent drift and hallucination
Beating the context limit — continuation prompts and multi-thread workflow for work that outgrows a single session

Session Links:

AI Classroom Task Tiers for Teachers (PDF)

Building Folder-Based Agents (Slideshow)

Building Folder-Based Agents (PDF)

Building Folder-Based Agents (Downloadable HTML Mini-site)
Prompting Language guide (PDF)

Submit your AI Design Journal

Learning Lab Async Activity

To receive your CTLE/CEU credits for the Learning Lab, please submit the Learning Journal that you utilized throughout the day.

Submit your AI Design Journal