U of T outlines its approach to AI infrastructure
The university’s AI-ready future could reshape how students learn and navigate university.

Artificial intelligence (AI) is becoming integrated into university practices, especially by 2030, whereby it will play a big role in how students study, brainstorm and work through university as a whole. 

In July 2026, the University of Toronto (U of T) announced a multi-year partnership with Cohere, a Toronto-based AI company founded by former U of T students. According to U of T News, the Cohere North platform is expected to support the university’s upcoming enterprise-wide platform that connects information across university systems while keeping data all within its control. 

For students, this change could impact multiple parts of the university experience, including classrooms, policies, and the Office of the Registrar. 

In 2024, U of T established an AI Task Force to develop a strategy for responding to the rise of generative AI (GenAI). According to U of T’s AI Task Force report, the university’s approach centres on educational elements including active learning, critical thinking development, experiential learning, and responding to technological changes. 

The report highlights students’ access to personalized and interactive support systems for learning, as well as changes in the way instructors redesign assignments. 

In an interview with The Medium, Dr. Christopher Eaton, who studies AI literacy and emerging technologies in education, said, “AI-ready universities should be adaptable.”

Part of that adaptability, he explained, means giving instructors and students room to experiment as the technology changes. It also means recognizing that policies can change rapidly and that AI may make sense for some tasks and have little place in others. Dr. Eaton argues that universities will increasingly have to account for this as part of the learning environment itself, saying that it should be “woven into the fabric of the university structure.”

That could eventually mean moving away from the same chatbot being used for every course. Dr. Eaton envisions smaller and more specialized models, tailored to the culture and students’ needs, as an important benchmark for 2030. 

These specialized models might look like interactions with chatbots that are designed around the materials and learning objective of a course. The skills that students may need to develop may therefore shift from knowing how it should be used to deciding whether it should be used. 

Dr. Eaton highlights that the “ability to evaluate AI outputs will be essential to its evolution,” alongside students’ judgement on how it fits into their learning processes. 

According to the Task Force, AI could help university staff process transfer credits and manage scheduling and enrolment. Undergraduate students who may need additional academic support could also benefit from these systems. Students could also interact directly with university provided AI systems to explore academic opportunities and career paths, find answers to routine questions, or connect with the appropriate staff member when making inquiries.

Accessibility is another point that falls within U of T’s AI vision, as it imagines a wider use of AI for tasks like note taking, lecture transcription, synthesizing text to speech, screen reading and image recognition. 

The report argues that providing AI systems institutionally could make access more equitable, while allowing U of T to evaluate tools for privacy bias and environmental impacts before they reach students. 

Research could also experience changes, as it may involve researchers using AI for tasks such as literature analysis and data management. The report emphasizes that researchers would still need to establish appropriate boundaries for AI use and assess the quality of its output.

U of T currently facilitates its technological adaptations through AI kitchen, which the university describes as a secure environment where vetted AI tools and data can be tested for teaching, research, and administrative projects.

Dr. Eaton stressed that students should have a meaningful role in determining how AI enters their education. “It is easy for learners to become left out of the conversations,” he said. This exclusion could undermine the eventual uptake and success of the technology.

One way to prevent this is to include students as partners in some of the curriculum design. Instructors would continue to determine how a course meets its learning outcomes, but students could contribute to conversations about where AI belongs within that structure. 

Dr. Eaton also pointed to flexibility at the individual level, highlighting that when appropriate, students might have the opportunity “to opt out but with specific caveats” for AI use when it is incorporated into course work.

The Task Force recommends a university-wide AI advisory body that would consult faculty, librarians, staff and students when developing institutional guidance and evaluating new applications. 

U of T’s decision to partner with Cohere adds another dimension because it asks who controls the technology and the data behind it. Cohere AI describes its own technology around the idea of sovereign AI systems that give organizations greater control over their data and infrastructure. Through partnerships, Cohere’s technology is intended to operate within U of T’s own AI environment.

Dr. Eaton sees the Canadian connections as significant for higher education. He pointed to potential advantages for data security, individual property and the university’s ability to build AI systems around its own context. He also stops short of treating sovereign AI as a guaranteed solution. 

“I am skeptical that this is a bit of an idealised vision for AI and sovereign built AI,” he said. Rather, Dr. Eaton described it as a potential “North Star” for how the university approaches AI. 

Becoming AI ready means that the university has to determine how those systems operate in classrooms and how they might operate in student services. For students, the university envisioned that AI may eventually become a part of the infrastructure of university life. This also means that students need to learn when to assess AI and question what it produces.

Professors responses:

Chris answers (on email): 

1. Q: From your perspective as someone who studies AI literacy and emerging technologies in education, what should an AI-ready university look like in practice for students and faculty? 

A: An AI-ready university should be adaptable. Adapting to the technology and its evolution is important, yes. But being adaptable to the changing ways that learners and employees use these technologies is essential. Universities can often be rigid and slow to adjust. Increased flexibility in policy making (on a macro level) and more space for experimentation by teachers and learners will make it easier to put those policies into practice. This will also enable space for people—learners, teachers, staff, administrators—to use the technologies and experience how they may logically fit into their work (or don’t fit). This flexibility also allows space for folks to opt out or, at least, identify and articulate areas where AI is inappropriate within our system.

Another dimension of being AI ready involves AI being woven into the fabric of the university structure, not just as a cool or threatening tool on the horizon. AI being an assumed influence on the curriculum—learners are likely using AI to support learning even in environments where AI is not permitted—that is thoughtfully accounted for in reimagined assessments, active learning opportunities, learning management systems, and course administration will lead to a more naturalised and deliberate AI uptake across the university.

2. Q: The AI Task Force report looks ahead at how AI could be integrated into teaching and learning by 2030. How do you think this could change the way students learn and engage with their coursework, and what skills do you think will become important for students to develop? 

A: An important benchmark for 2030 would be smaller, more specific AI models integrated into university courses to support students. This means that AI models will be more tailored to courses and student needs within those courses. It will contribute to AI becoming more naturalised and deliberate at the university. It will also lead to more thoughtful AI integration in course and programmatic curricula. In this scenario, which is likely given the university’s partnership with Cohere AI, learners will have clearer understanding of how AI can and should be used in specific scenarios. AI can become a more contextually specific tool that they can use to experiment, build AI skills specific to tasks, and—perhaps most crucially—determine where AI enhances their work and where it is best left out of their processes. The ability to evaluate AI outputs will be essential to this evolution, as will be the skills necessary to evaluate how AI fits into their own learning processes. The ability to communicate how AI is integrated and how that integration impacts the final creation will be equally essential.

3. Q: As UofT expands its use of AI through things like partnerships with Cohere AI, what do you think the university should prioritize to make sure that students are involved in how these tools are introduced and used, both in courses and beyond?

A: This is challenging given how difficult it is to procure these partnerships, vet the technologies, and develop a roadmap that shows instructors can thoughtfully integrate Cohere’s tools into their curricula (where appropriate). It is easy for learners to become left out of the conversation, which means there is a risk that an essential—if not THE essential)—part of the conversation and uptake can be overlooked. This will have negative impact on the uptake and success of the technologies.

Now for a potential solution. Learners becoming partners in designing parts of the curriculum can be a useful and overt way to ensure their voices are heard and help guide some of what happens in a classroom. Of course, the instructor will need to ensure that decisions align with learning outcomes, but a thoughtful design can invite useful conversations that can mediate, guide, and enhance the way AI is used in a classroom. At the same time, it is useful to think of flexibility within the curricular structure in a way that gives individual students quiet ways to participate in AI discussions. Flexibility in how AI tools are taken up in class and in assessments gives learners space to opt in, opt out, or opt in but with specific caveats. This does not give learners a voice overtly, but it does give them a lot of control and say in how AI shapes their learning while still meeting the learning outcomes for a particular task or class.

One final important note: Cohere AI is a sovereign (i.e., Canadian) company. This is huge for education. It ensures security of data, continuity of the systems regardless of regulation and geopolitical situations, context-specific learning analytics to inform AI and curriculum decisions, and a way to control data/copyright/intellectual property internally. It may help us partner with other universities to support and collaborate on AI across the higher education sector. It should also give the university and its people more specific ways to build AI into the fabric of university programming. Well-built sovereign products will enhance confidence and encourage experimentation in courses. I am a bit sceptical that this is a bit of an idealised vision for AI and sovereign-built AI, but it may act as a useful North Star that can enhance AI uptake, student voice, and the security and access of AI regardless of whether the full idealised objective is accomplished.