Module 1: AI and Agentic Engineering for Education - Worked Examples

Agentic Engineering for Teaching Professionals

These examples move from bounded AI assistance toward reusable teaching workflows. Adapt every prompt to the participant's actual subject, institution, learner level, evidence, and authorized tools rather than treating any template as universally correct.

Improving a Weak Classroom AI Request

An instructor asks, "Explain photosynthesis." The approved Grade 8 notes state that plants use light energy to form sugars from carbon dioxide and water. An AI draft instead says, "Plants obtain all their food from soil." Repair the request and decide whether that sentence can be released.

Step-by-Step Solution

0 of 3 Steps Completed
1

Comparing AI Systems Without Ranking Them

Apply this exercise to a specific teaching context and preserve the learner level, governing evidence, and institutional constraints that make the result valid.

Step-by-Step Solution

0 of 3 Steps Completed
1

Detecting a Hallucinated Explanation

Apply this exercise to a specific teaching context and preserve the learner level, governing evidence, and institutional constraints that make the result valid.

Step-by-Step Solution

0 of 3 Steps Completed
1

Making Uncertainty Visible

Apply this exercise to a specific teaching context and preserve the learner level, governing evidence, and institutional constraints that make the result valid.

Step-by-Step Solution

0 of 3 Steps Completed
1

Adapting One Concept for Different Learners

Apply this exercise to a specific teaching context and preserve the learner level, governing evidence, and institutional constraints that make the result valid.

Step-by-Step Solution

0 of 3 Steps Completed
1

Assistant Versus Decision Maker

Apply this exercise to a specific teaching context and preserve the learner level, governing evidence, and institutional constraints that make the result valid.

Step-by-Step Solution

0 of 3 Steps Completed
1

Selecting the Minimum Sufficient Architecture

Apply this exercise to a specific teaching context and preserve the learner level, governing evidence, and institutional constraints that make the result valid.

Step-by-Step Solution

0 of 3 Steps Completed
1

Designing a Human-in-the-Loop Workflow

Apply this exercise to a specific teaching context and preserve the learner level, governing evidence, and institutional constraints that make the result valid.

Step-by-Step Solution

0 of 3 Steps Completed
1

Hands-On Lab — AI Use Boundary and Workflow Selection

Activity goal: Choose one real teaching task and design the minimum sufficient AI workflow while preserving evidence, verification, and educator authority.

Required evidence and deliverables:

  • A one-paragraph teaching task and intended learner group.
  • A source-of-truth list identifying what AI may and may not treat as evidence.
  • A consequence classification: low, moderate, or high.
  • A selected architecture: prompt, file-aware assistant, workflow, or agent.
  • A human-review checklist and explicit stop condition.

Assessment rubric — 25 points

CriterionPointsRequired evidence
Task and evidence definition5The educational objective and controlling evidence are specific and realistic.
Architecture selection5The chosen AI architecture is no more complex than the task requires and is justified.
Risk and authority boundary5Consequence level, prohibited actions, and educator authority are explicit.
Verification design5Checks target plausible failure modes rather than generic confidence.
Usability and transfer5The workflow can be reused in an actual teaching context.

Step-by-Step Solution

0 of 5 Steps Completed
1