From Panic to Purpose: Leading the AI and Assessment Conversation
Academic leaders are fielding urgent questions about AI, assessment, and student learning. Practical responses include shifting from securing everything toward intentional assessment design, AI literacy, and local experimentation.
By Gwenna Moss Centre for Teaching and LearningThe questions have shifted
Academic leaders are hearing a different set of questions about generative artificial intelligence (AI). Instructors are no longer only asking whether students are using AI; they are asking what programs, departments, and colleges should do now that AI is part of the learning environment.
Some questions are about academic integrity, some about workload. Most are really about learning: what students still need to know, practice, and judge for themselves when AI can produce a fluent answer in seconds.
The most useful response is not a single rule for all courses. It is a better set of questions: What learning matters most? Where is AI off-limits, permitted, or required? Where must an assessment change because AI can now complete the task without the student demonstrating the intended learning?
Are our assessments still doing what we need them to do?
This is the question behind most concerns about AI, and it is the right one. Some assessments still work because students must demonstrate the intended learning under appropriate conditions. Others need to change because AI can produce the submitted product while leaving instructors uncertain about what the student understands.
The work is to redesign the right assessments, clarify expectations, and help students build judgement. Academic leaders can help by moving the conversation from “Can we secure everything?” to “What do we most need students to show us before they move on?”
USask instructors are already showing what this looks like. In Pharmacy, Dr. Ed Krol and Dr. Anas El-Aneed redesigned a group assignment so students completed their own work first, then used Microsoft Copilot and the grading rubric to evaluate AI-generated feedback. In an online English course, Dr. Carleigh Brady rebuilt her assessment around checkpoints that make student thinking visible, with feedback possible, across the writing process.
What skills do students need now?
Students need more than rules about when AI is allowed. They need practice using AI responsibly, questioning its output, identifying errors, and deciding when human expertise matters. USask has recently released faculty and student AI Literacy Tutorials.
Dr. Kyle Anderson’s redesigned exam in Biochemistry, Microbiology and Immunology is a strong example. His “fact-check” questions present AI-generated passages containing deliberate errors; students must correct the inaccurate statements and leave the accurate ones alone. That is what graduates will increasingly need to do in an AI-enabled workplace: verify fluent, confident, mostly-right output before it affects a patient, dataset, or decision.
The example works because it treats AI literacy as part of disciplinary learning, not a separate topic. Academic leaders can help programs ask: What can AI produce in this discipline? What must students still judge for themselves? Where could an AI error cause harm?
Do instructors need to redesign everything?
No. And that answer matters.
Not every course needs an overhaul, and not every assessment is equally affected. A manageable first step is to identify the key program outcomes students must demonstrate to progress, then decide where those outcomes need to be assessed more securely. Oral assessment can be useful here because it asks students to explain their reasoning in the moment. This shifts the work from protecting every task to designing milestone assessments that show students are ready for the next stage.
Other assessments will still support practice, feedback, learning, reflection, collaboration, and AI literacy. They do not all need the same level of security.
Anderson’s exam is reassuring for another reason: it began with materials he already had, including lecture transcripts, slides, past exam questions, and his own rules for good questions. AI drafted questions; the academic judgement stayed with him.
What can academic units do together?
Academic leaders do not need to solve every course-level problem. They can create conditions for instructors to make better decisions together.
At the program or department level, that might mean identifying the assessments most affected by AI, agreeing on common syllabus language, and deciding where students will encounter explicit AI literacy development. It should also mean naming the milestone assessments where key outcomes must be demonstrated before progressing. Colleges at USask have already built custom frameworks for determining acceptable levels of AI use and communicating them to students.
Shared approaches help students see AI expectations as part of learning, not only compliance.
How can AI improve teaching?
The conversation should not be only defensive. Expanded access to protected AI tools gives instructors more room to model responsible use, create practice opportunities, and help students engage with complex material.
USask is exploring these questions through structured AI pilots, including TrackPoint and Gemini Notebook. Instructors can also now purchase access to protected AI tools, with student access continuing to develop. Together, these developments create better conditions for teaching students how to use AI well, not just warning them where they cannot use it.
The point is not uncritical adoption but careful experimentation with appropriate supports. Local examples help instructors move past abstract claims and ask better questions: Does this tool improve learning? Does it reduce or increase inequity? Does it help students practice the skills they need?
A practical stance for academic leaders
AI creates a chance to be more explicit about the learning that matters most. Integrity, assessment design, privacy, workload, and teaching innovation all need attention, but not all at once.
A useful starting point is to help instructors move from panic to purpose. What learning matters most? What should students be able to do without AI, and what should they learn to do with it? Which outcomes must be demonstrated securely before they move on?
USask already has examples to ground those conversations: an exam that measures verification and disciplinary judgement, a Pharmacy redesign that has students evaluate AI feedback, an English course that makes student thinking visible across the writing process, college-level frameworks, AI Literacy Tutorials, and pilots such as TrackPoint and Gemini Notebook.
The goal is not to make every assessment AI-proof. It is to make assessment more intentional about what students must know, do, judge, and verify, and to secure the milestones that matter most for progression.
Title image credit: website/author
This article was created with the assistance of AI tools, as described in the GMCTL AI Disclosure Statement.