AI in Higher Education: Lessons from Students, Educators, and Institutions
What does responsible, practical AI use look like in higher education?
That was the question at the heart of our recent webinar, AI in the Open: Real-World Practices in Teaching and Learning, which brought together educators and AI practitioners from the University of Leeds, Red River College Polytechnic, and the Idaho State Board of Education.
The conversation covered student experiences with generative AI, institutional AI strategy, accessibility, agentic browsers, assessment, and the practical work of helping educators and students navigate a rapidly changing landscape.
But before getting into what AI can do, there is another question worth asking: what does using AI cost?
Understanding the environmental impact
One audience member asked something that we want to bring to the foreground:
How do we work together to minimize who has to do this work, given its cost and environmental impact?
Rebecca Molly Hiebert shared that Red River College Polytechnic has a working group specifically looking at the environmental impact of AI use. The institution is exploring how it can use its IT data to estimate the water and electricity associated with different types of AI interactions, with the goal of eventually creating a dashboard that makes that impact easier to understand.
The goal is to give people enough information to make more informed choices about when AI is useful, what tasks are worth using it for, and whether a particular task could be handled locally rather than through a cloud-based tool.
For anyone who wants to explore the environmental impact of their own digital habits, the panel also shared Your Digital Life, a resource for thinking about the environmental impact of AI and other digital technologies.
This question set an important tone for the rest of the conversation: AI literacy isn’t just about learning what these tools can do. It’s also about understanding their implications and making intentional choices about when and how to use them.
Three projects, three approaches to working with AI
1. Learning with AI: Listening to students
At the University of Leeds, Chrissi Nerantzi and a team of student editors are taking a student-centred approach to understanding AI. The featured text, Learning with AI, is a crowdsourced, co-created open book that captures students’ experiences of using generative AI in their learning.
The project grew out of earlier crowdsourced work that primarily captured educators’ experiences with AI. The team noticed that, despite actively seeking student perspectives, relatively few students were contributing. That led to a dedicated project in which students could share their own experiences and work as student editors.
The 2024–25 data included 28 contributions across 21 academic disciplines and four countries, from undergraduate students, taught postgraduate students, and postgraduate researchers.
Their analysis identified four themes in how students were using generative AI:
- Sensemaking — using AI to interpret complex information
- Academic work planning — organizing, structuring, and mapping direction
- Self-efficacy — building confidence in their own abilities
- Creative exploration — finding innovative solutions to problems
Perhaps most interestingly, the team said the contributions did not suggest students were simply using AI to bypass thinking or learning. Instead, they saw students using AI to diversify opportunities for learning and inquiry in critical and creative ways.
The project continues to evolve. The team has launched Roundabout, a framework and research project exploring how educators across different sectors can work with AI based on principles and trust, alongside tools such as the Thinking Thermometer.
Learning with AI – from the University of Leeds
2. Building institutional capacity: Red River College Polytechnic
Rebecca Molly Hiebert shared two connected pieces of work from Red River College Polytechnic: an open educational resource on AI created with instructors, and the institution’s ongoing AI strategy.
The original resource was developed in 2023–24, when AI was rapidly entering educational practice and instructors were looking for practical guidance. Project lead Troy Heaps consulted instructors across a wide range of programs—including culinary arts, nursing, early childhood education, electrical engineering, library and information technology, and medical laboratory services—to understand how they were actually using AI.
The resulting resource covers using AI to develop courses, lesson plans, activities and assessments; enhance student learning; teach students to engage with AI effectively, ethically and responsibly; and support administrative tasks.
The resource has reached more than 10,000 visitors and 25,000 page views and is used as a core reading in courses at the institution.
That work has since expanded into a broader institutional AI strategy. Red River’s approach includes AI commitments and principles, cross-functional governance, AI readiness work with academic programs and industry partners, and training for students, staff and faculty. Two particularly interesting faculty initiatives focus on using AI as a collaborative partner for deeper learning and critical thinking, and redesigning assessments so students can more clearly demonstrate what they have learned.
The bigger lesson is that institutional AI readiness isn’t just about adopting tools. It involves curriculum, professional development, policy, assessment, governance and ongoing conversations with the people who will actually use the technology.
AI in Education — from Red River College Polytechnic
3. Accessibility by Design: Using agentic browsers for practical work
Liza Long approached AI from a different angle: what happens when we let AI perform some of the repetitive administrative work that takes up our time?
Her presentation focused on agentic browsers—tools that allow an AI system to interact with websites and perform actions rather than simply generating text.
Liza compared an agentic browser to a self-driving car: it can do some of the work, but you still need to keep your hands on the wheel and your eyes on the road.
Her examples focused heavily on accessibility. One workflow used an agentic browser to check and correct heading structures in a Pressbooks book. The key was to give the AI very specific constraints: check the headings, don’t rewrite the content, don’t make unrelated changes, and provide an audit report showing what was changed.
She also demonstrated workflows for standardizing image attribution and alt text, converting large numbers of PDFs, and even creating a Canvas course from a Pressbooks book. In one example, 758 PDFs were converted using code she developed with Google Colab.
But perhaps the most important part of this presentation wasn’t the automation itself. It was the guardrails around it.
Liza recommended downloading an XML archive of a Pressbooks book before allowing an agentic browser to modify it, and emphasized breaking complex tasks into small, clearly defined steps rather than simply asking an AI agent to “fix” something.
Accessibility by Design — by Liza Long
What can we take away from these projects?
Despite looking quite different, the three presentations shared some striking themes.
1. Start with the learning problem—not the AI tool
The Leeds project starts with a simple question: How are students actually using AI?
Red River’s work asks: What do students and educators need to become AI-ready?
Liza’s workflows ask: What repetitive task could be done more efficiently while preserving human oversight?
In each case, the technology follows the problem rather than defining it.
2. AI literacy needs to include critical use
The Leeds research is particularly useful here. Students weren’t only using AI to produce answers. Their reported uses included sensemaking, planning, building self-efficacy, and creative exploration.
That suggests AI literacy can be about much more than knowing how to write a good prompt. It can involve understanding when to use AI, how to question its output, how to incorporate it into a learning process, and when not to use it at all.
3. Human oversight doesn’t disappear when AI gets more capable
This came through especially clearly in Liza’s presentation.
Agentic AI can click, edit, organize and transform content, but the human remains responsible for the result. During the Q&A, Liza described this as “asymmetric accountability”: the AI isn’t accountable for the work it produces. The human is.
That means automation needs its own literacy: understanding permissions, backups, privacy, verification and appropriate boundaries.
4. Institutional AI work is a people-and-process challenge
Red River’s experience demonstrates that AI strategy isn’t something that can be solved by purchasing a tool or publishing a single policy.
Their work brings together governance, training, curriculum, assessment, industry partnerships and feedback from students, staff and faculty.
5. Open resources can help institutions learn together
All three projects demonstrate the value of sharing practical work openly.
The Leeds team is documenting student experiences in an open, living book. Red River shared its research and teaching resources openly. Liza has made her accessibility workflows available under a Creative Commons licence. The result is that educators don’t have to start from scratch every time they encounter a new AI challenge.
Resources shared during the webinar
Student perspectives and AI literacy
- Learning with AI — University of Leeds
- 101 Creative Ideas to Use AI in Education — Zenodo
- Towards AI Literacy: 101+ Creative and Critical Practices, Perspectives and Purposes — Zenodo
- Project Roundabout
- The Thinking Thermometer — Zenodo
- AI in Education — Red River College Polytechnic
- Learning with AI student survey/contribution form
AI and workforce readiness
Accessibility and AI
- AI Accessibility — Idaho State Board of Education
- Accessibility by Design — Liza Long
- Pressbooks Content Toolkit
Environmental impact
Keeping the conversation going
One of the strongest messages from the webinar was that there isn’t a single “right” way to approach AI in higher education.
Students are experimenting with AI in different ways. Educators are developing new teaching and assessment practices. Institutions are building policies, training and governance structures. And practitioners are finding new ways to use AI for the repetitive work that can otherwise consume valuable time.
The challenge is to keep experimenting critically, openly and with intention—while continuing to ask who benefits, what we’re asking AI to do, what it costs, and where human judgment needs to remain firmly in the loop.
Cover image created by Pressbooks