Creating Course Materials with AI

Generative AI can help brainstorm or draft cases, slide outlines, examples, questions, and reading-list annotations. It should not decide what belongs in a course or supply facts and references that an educator cannot verify.

A reusable production workflow

Case studies

Case-based learning can support application and decision-making when the case is aligned, credible, and followed by structured analysis.¹ AI can propose scenarios and draft fictional details, but realistic-sounding data and stakeholders require careful review.

Case-study prompt

Draft two alternative case outlines for [learner group].

Learning outcome:
[approved outcome]

Students should have to:
[decision, analysis, or skill]

Use only the supplied source pack for factual claims. Clearly label invented
names and fictional data. Each outline must include:
- setting and decision point;
- information students receive;
- two plausible options with genuine trade-offs;
- missing information students should ask for;
- three discussion questions mapped to the outcome; and
- an educator verification checklist.

Do not resolve the dilemma or portray a protected group as the source of the
problem. Flag anything the sources do not support.

[source pack]

Select an outline, then verify and refine it. Ensure the case gives students enough evidence to reason without embedding the “correct” conclusion in the narrative. Label synthetic numbers so they cannot be mistaken for real data.

Lecture slides

AI can turn an educator-approved outline into a slide plan, suggest examples, or identify where a visual might help. It cannot know what students already understand, whether a slide is overloaded, or whether a generated image is accurate and appropriately licensed.

Slide-outline prompt

Create a slide outline for a [duration] session with [learner group].

Use only the supplied learning outcomes and source material. For each slide,
return:
- one message-style title;
- no more than three concise points;
- a suggested visual only when it supports comprehension;
- speaker notes with the source passage for factual claims; and
- the student activity or learning check, if any.

Include an opening retrieval question and a closing check mapped to the
outcomes. Do not invent citations or image licences. Flag content that requires
the educator to source or create a visual.

[outcomes and source material]

Before use, remove decorative clutter, split overloaded slides, add meaningful alt text, check reading order and contrast, and verify that spoken explanation does not carry information unavailable to students using another format.

Reading lists

Language models are unreliable bibliography generators: they can invent a publication or combine details from several real sources.² Use library databases, publisher records, DOI registries, disciplinary indexes, or an approved search system to discover and verify literature.

AI can still assist after sources are verified. Supply the confirmed bibliographic records and abstracts or full text you are permitted to use, then ask it to compare coverage, propose an order, or draft annotations.

Reading-list audit prompt

Using only the verified source records below, audit this reading list for:
- coverage of the stated learning outcomes;
- prerequisite difficulty and weekly workload;
- duplicated contribution;
- relevant perspectives or methods that are absent;
- accessibility and lawful availability; and
- records that lack enough information for a judgment.

Return a table with source identifier, contribution, outcome, difficulty,
availability, concern, and educator action. Do not add publications.

[learning outcomes]
[verified records and abstracts]

Open each item before publishing the list. Confirm authors, title, venue, year, DOI or stable link, access conditions, and that the source supports the annotation. Asking the same model “are these references real?” is not verification.

Final human review

• The material serves a named learning outcome.
• Facts and references were checked against authoritative sources.
• Fictional and synthetic content is clearly labelled.
• Examples do not reproduce stereotypes or exclude relevant perspectives.
• Reading level, terminology, workload, and activity timing fit the learners.
• Accessibility was checked in the final delivery format.
• Copyright, licences, attribution, and image provenance were reviewed.
• No protected information was entered into an unapproved tool.
• Students are told when AI use is relevant to interpreting the material.
• The educator remains able to revise and teach the material without the tool.

References & Footnotes