Utilising AI in Teaching and Learning in the CBE

College of Business and Economics, ANU | Updated June 2026
Prepared by the Associate Dean (Education), CBE

This document provides guidance for CBE staff on the use of Generative AI (GenAI) in teaching, learning and assessment[1]. It covers:

What GenAI is and how it works
Tips for talking about GenAI with students
CBE’s approach to assessment in an AI-enabled environment
How AI can be used in teaching and course design
Academic transparency about AI use
Data privacy and security

What is GenAI?

Generative Artificial Intelligence (GenAI) refers to a category of AI models and algorithms designed to generate new, original content based on patterns learned from existing data. Unlike traditional AI systems that focus on recognising patterns or making predictions within fixed categories, GenAI can produce new content — text, images, software code, music, video — that reflects the style and characteristics of the data it was trained on.

Most GenAI tools rely on large language models (LLMs), which use predictive algorithms to determine the most statistically likely sequence of words, given a prompt. This approach performs well in areas where extensive, high-quality training data exists, but it can also produce “hallucinations” — confident, plausible-sounding outputs that are factually incorrect. Because LLMs generate text by predicting likely word sequences rather than retrieving verified facts, hallucination cannot be entirely eliminated, even as accuracy continues to improve with each new generation of models.

GenAI has significant potential to transform education through personalised learning, administrative efficiency, and improved outcomes. It also introduces challenges — most acutely in the integrity of assessment and assurance of learning — that require deliberate and transparent responses from academic staff.

The ANU principles on AI use and institutional information is available here. Additional resources include the University of Sydney’s AI and Education site (co-created by academics and students) and a Digital Education Council AI certificate.

GenAI and Our Students

An important entry point with students is an early, honest conversation about how GenAI intersects with your discipline and with their learning. Key points to cover:

Employability. GenAI is increasingly embedded in professional practice. Developing confidence and critical literacy with these tools is now a genuine graduate capability.
Responsible use. In the absence of consistent regulation, students need to develop their own ethical framework: understanding the security and intellectual property (IP) implications of what they input; recognising and correcting for bias in AI output; questioning or override AI-generated content.
Hallucination. A good classroom activity is to ask an AI to write a biography for a real person known to the students, or to answer a discipline-specific question where the model is likely to struggle. This makes hallucination concrete rather than theoretical. [NB: This is becoming increasingly rare, so test beforehand].
Critical thinking. Emerging research suggests that over-reliance on AI can erode the recall and critical thinking skills that deeper engagement with material builds. The act of working through a problem — not just receiving an answer — remains pedagogically important and critically important to employers.
Equity. The quality of AI output is partly a function of prompt quality, which is in turn shaped by prior exposure and experience with AI tools. Access to advanced AI versions also varies. Providing all students with access to prompt training resources helps mitigate these inequities. Useful starting points: the University of Sydney AI site, LinkedIn Learning (available to all ANU staff and students), and the Digital Education Council X ANU AI literacy for all .
Academic Integrity. Increasingly we are expecting students to use AI in non-invigilated assessment. However, the key point is that they are responsible for the output submitted. In part this pertains to the information in ‘responsible use’ above, but this also means that they should check and interrogate all references. If references are required but not provided or incorrect/not real – this is a breach of academic integrity that the student is responsible for.

Be explicit with students about the framework for AI use in your course — which assessments permit it, how they are expected to use, it, and what referencing or disclosure is required.

GenAI and Assessment

The CBE position: embed AI or invigilate

CBE current guidance: All assessment that is not conducted under invigilation should, by default, permit and actively incorporate the use of AI. Where a specific learning outcome requires students to demonstrate capability without AI assistance, that component should be assessed in an invigilated setting.

The practical reality is that any work completed outside a supervised environment is very likely to be completed with some AI involvement. Rather than treating this as a compliance problem, the CBE approach is to treat it as a design opportunity: if a task can be completed adequately by an AI without meaningful student input, the task needs to be redesigned.

For the majority of courses, assessment design should therefore follow a two-stream structure:

Non-invigilated assessment (for learning): Redesign these tasks to embed AI use deliberately. Students should be expected to use AI, to critique its output, and to demonstrate what they have added. Tasks that simply ask students to explain or summarise content — which AI can do adequately — should be replaced or supplemented with tasks requiring engagement with in-class discussion, a guest speaker’s perspective, or a specific dataset or case that requires genuine analysis[2]. Tasks can also include options such as quizzes where you encourage students to complete without the use of AI for their own learning/development/mastery. However, no actions can be taken if you suspect students use AI to complete.
Invigilated assessment (of learning): At least one assessment per course should be conducted in a supervised setting — an exam, tutorial quizzes, oral defence, or in-person task. This is where the University’s confidence in student achievement against learning outcomes rests. A weighting of up to 70% is reasonable for this component; making it a hurdle requirement is also worth considering for foundation courses[3].

Redesigning non-invigilated assessment

Some approaches that make AI use a feature[4] rather than a loophole:

Anchor tasks in in-class activity. Require students to connect their submission to a specific lecture discussion, guest presenter, or workshop activity. AI cannot replicate engagement that has to have happened in real time.
Require disclosure of AI use. Ask students to submit the prompts they used, explain how they refined the AI output, or reflect on where the AI was inadequate. Prompt quality itself can be evidence of student understanding.
Bring AI into the room. Provide a question ahead of time, allow students to bring a GenAI draft, and use class time to refine and extend it. Students submit the handwritten final version alongside the original AI output and must explain what they changed and why.
Use AI to individualise tasks. Generate personalised datasets, case scenarios, or question variants. Generic AI prompts are less useful, and individualisation can work to increase engagement.
Critique AI output directly. Set tasks that ask students to identify the biases, errors, or gaps in an AI-generated answer and produce an improved version, allowing a demonstration of a student’s critical thinking skills and/or creativity.

Assessment for learning: embedding AI as a study tool

For students

Use AI as a personalised tutor — input your answer to a question and ask for feedback or ask it to generate revision questions from your notes.
In group tasks, use AI to find a starting point or draft an initial answer, then critique, refine and extend it as a group. Submit the group’s work alongside a personal reflection on what you changed and why.
Discuss and correct for bias in AI output — who is represented in the training data? What perspectives are missing?

For staff[5]

Use AI to generate step-by-step worked solutions or practice question sets, freeing tutorial time for group discussion and application.
Use AI to develop real-world case studies that ask students to apply concepts to complex situations rather than solve formulaic problems.
Generate personalised assessment datasets using AI, reducing the possibility for collusion.

How AI Can Support Teaching and Course Design

AI tools can meaningfully reduce the administrative and preparation burden on academic staff, and their use in course design is encouraged. Examples include:

Generating marking rubrics for assessment tasks as a starting point for academic review and refinement.
Developing H5P[6] interactive content from course materials uploaded to AI.
Creating lecture slide drafts, case study scenarios, or practice question banks
Generating individual or group datasets for assessment tasks

Academic Transparency About AI Use

CBE expectation: Academics are expected to be transparent with students about their own use of AI in developing course materials. This includes, but is not limited to, the use of AI in drafting marking rubrics, developing course slides, generating practice questions, or producing case studies. You may also wish to be clear about what AI is NOT used to do (i.e. marking).

There are two good reasons for this. First, transparency about AI use is consistent with the academic values of honesty and integrity that we ask students to uphold — it is difficult to credibly ask students to reference and disclose their AI use if staff do not model the same practice. Second, being open about how AI has been used in course design creates a productive teaching moment: it gives students a concrete example of AI in a professional context, and an opportunity to discuss what value was added by human expertise and judgment.

In practice, transparency might look like:

A sentence in the course introduction or first lecture noting that AI was used in drafting the marking rubric or generating practice questions, and that the output was reviewed and refined by the course team
A brief note on a slide or worksheet that AI contributed to its development
A course-level disclosure statement on the Canvas site alongside the AI-use policy for students

This is not about exhaustive documentation of every AI interaction — it is about modelling the kind of straightforward professional disclosure we want students to develop as a habit.

Data Privacy and Security

The use of AI tools raises important data privacy and security considerations. The following are non-negotiable:

Do not input student data[7] into any open-source or third-party AI platform.
Do not upload student work[8] into AI tools for marking or feedback purposes.
Do not upload your own unpublished work into open-source AI tools if you intend to retain intellectual property over it.
Choose reputable ANU approved platforms that adhere to data protection regulations and implement robust security measures. Noting that even if your use of an AI platform is ‘approved’ for use by ANU, uploading private information is still prohibited.

ANU’s currently approved platform for staff use is Microsoft Copilot Enterprise. However, the precise privacy protections within Copilot are not yet fully clarified at the institutional level — exercise caution with sensitive material, do not upload private material and watch for updated guidance from the University with regards to approved platforms and usage restrictions.

Document last updated: June 2026 | Associate Dean (Education), CBE, ANU

[1] Further guidance and specifics from ANU will be forthcoming and will supersede these. However, in the meantime, these will be our guiding principles and are consistent with where the ANU is heading.

[2] These assessment tasks are not foolproof to AI, but it may be harder. It might also be advised, depending on the assessment option being used, to ensure devices aren’t on in the classroom i.e. not recording through AI for later summarisation of guest lecturer.

[3] Hurdle requirements to have specific information needed in the Class Summary and options to account for supplementary assessment. Please speak to your DDE and/or the ADE if you are considering this approach.

[4] Note, this doesn’t remove the ability of students to use AI to answer, but potentially changes expectations.

[5] The key to remember is to always have oversight over output. And remember – due to privacy regulations, no student work can be uploaded into any AI platform outside of Canvas.

[6] ChatGPT tells me: H5P, short for HTML5 Package, allows users to create rich, interactive web content such as quizzes, interactive videos, presentations, games, timelines, and more, all directly in a web browser without needing advanced programming skills. H5P is supported by Canvas at ANU. See the EdTech team if you’re interested in finding out more.

[7] There are privacy use conditions around student data as collected by ANU.

[8] The IP of student work submitted is retained by students and as such can not be uploaded without their explicit consent.