AI Agents for Teaching - A Short Course

By the end of the seminar, you will have developed practical, course-ready tools you can immediately apply in your own teaching, including:

• • A course-specific teaching agent workflow.
  • Redesigned assignments and classroom activities adjusted for AI-enabled learning environments.
  • Practical strategies for improving grading consistency, feedback quality, and rubric clarity.
  • Student-facing AI policy language and transparent AI use guidelines.
  • A clear framework for deciding where AI can support teaching and where instructor judgment should remain central.

The workshop is not about tricks with prompts, automated grading, or replacing instructors with AI systems. The focus is on building controlled, transparent, and pedagogically defensible workflows.

Come with an AI coding agent installed and working — Warp, Claude, or Codex all work, so use whichever you prefer. A premium plan is preferred, since we run the agents hard and free tiers tend to hit their limits mid-exercise.

API keys are helpful but optional; everything core to the course runs through the agent interfaces.

We will provide up-to-date setup guidance before the seminar, including recommended AI tools, account requirements, and any estimated API costs for course exercises.

This seminar is designed for faculty, graduate instructors, instructional designers, and program directors who want to thoughtfully integrate AI into teaching, assessment, and course design. It is especially relevant for instructors teaching writing, methods, statistics, policy, ethics, and other courses where AI is changing how students learn and complete assignments.

Session 1: From chatbots to teaching agents

• • What AI agents are and how they differ from chatbot use
  • Which teaching tasks agents can support and which should not be delegated
  • How students are already using agents in coursework
  • Why detection-first responses fail and what to do instead

Session 2: Building human-controlled teaching workflows

• • Building multi-step teaching workflows
  • Writing agent instructions for teaching tasks
  • Preserving instructor voice and course coherence
  • Documenting AI-supported teaching decisions

Session 3: From static assignments to interactive learning

• • How AI changes the value of traditional assignments
  • Designing activities around decision-making
  • Creating role-specific agents for simulations and tutoring
  • Building branching scenarios and structured debriefs
  • Making simulations assessable and avoiding gimmicks
  • Exercise: Build a virtual teaching game

Session 4: Using agents to improve course materials

• • Generating examples at different difficulty levels
  • Translating abstract concepts into concrete cases
  • Building flawed examples for students to diagnose
  • Creating practice problems and synthetic teaching datasets
  • Mapping likely student misconceptions
  • Testing assignment instructions before release
  • Exercise: Build a misconception-driven activity with a diagnostic question, worked example, and feedback plan

Session 5: Designing assessments for an AI-saturated classroom

• • Why AI detection is a weak foundation
  • Redesigning assignments around process and judgment
  • Using staged submissions, oral defenses, and revision memos
  • Assessing AI-assisted work transparently
  • Writing student-facing AI use rules
  • Exercise: Redesign an assignment that AI has weakened

Session 6: Debiased grading and feedback workflows

• • Why fully automated grading is risky
  • Rubric clarity audits and feedback consistency checks
  • Bias checks and blind review workflows
  • Detecting rubric drift and uneven feedback
  • Human override and transparency
  • Privacy and data minimization in grading workflows
  • Exercise: Build a grading-audit workflow with clear limits on what the agent may evaluate

Session 7: Moodle, Canvas, MCP, and LMS agents

• • What an LMS agent is and what APIs and Model Context Protocol (MCP) actually enable
  • Read-only versus write-enabled agents and why you start read-only
  • Permission boundaries and student privacy
  • Sandboxes, test courses, approval gates, and audit logs
  • Exercise: Design a safe LMS agent

Session 8: Governance, course policies, and final build sprint

• • Course-level AI policy design and student-facing transparency
  • Accessibility, equity, and data protection
  • Academic integrity and instructor responsibility
  • Evaluating whether AI improved learning
  • Building a sustainable workflow and deciding what not to automate
  • Final deliverable: Develop a teaching agent playbook for one course