Framework for AI Adaptation
Framework for AI Adaptation
Generative Artificial Intelligence Leadership Task Force Report
Unlike other kinds of institutions or sectors of the economy, universities need to consider several frameworks simultaneously, each with their own distinct concerns and considerations. In an op-ed piece, James L. Norrie argues that,
Effective governance requires a clearer distinction… between…academic and administrative uses of AI, each of which introduces different strategic, operational, and ethical considerations….These domains overlap, but they are not interchangeable. Conflating them produces reflexive over-reaction, weak policy, uneven implementation, and institutional confusion.4
We can go further and distinguish within the academic context between the domain of learning and teaching on the one hand, and the domain of research on the other.
The task force proposes three areas of strategic focus as the foundation of a framework for adapting to AI at TMU3:
TMU’s approach to generative AI in learning and teaching is grounded in a set of existing principles: fostering AI literacy, supporting ethical and transparent use, encouraging learner-centred innovation, protecting academic rigour, advancing equity and accessibility, and connecting learning to disciplinary and professional contexts. These commitments are important because the university’s response to AI cannot remain primarily reactive. The task now is to move toward a more intentional approach that helps instructors, students, support staff, departments/schools, and programs make thoughtful decisions about how AI should, and should not, shape teaching and learning.
A central principle is responsible experimentation. AI is already reshaping how students learn, how instructors design learning experiences, and how knowledge work is practiced across professions. For this reason, TMU’s response should not be limited to risk mitigation or academic integrity enforcement, though those concerns remain critical. The university also needs to create the conditions for ethical, evidence-informed, and pedagogically purposeful experimentation. Designated Teaching Fellowships focused on emerging applications of AI in teaching and assessment, along with Students as Partners in Generative Artificial Intelligence grants, can play an important role in this work. These targeted programs will help ensure that innovation is grounded in the practical realities of teaching and learning, while also reinforcing a culture of trust and shared responsibility.
At the same time, we must recognize that the educational impact of AI varies significantly across disciplines. Broad university-wide principles are necessary, particularly in areas such as privacy, transparency, equity, accessibility, academic integrity, and responsible use. However, the implementation of those principles needs to be discipline-specific. Programs need to be supported to develop guidance that reflects their own learning outcomes, accreditation requirements, professional standards, assessment cultures, and emerging workplace expectations. This approach allows local experimentation to remain connected to shared institutional values without reducing AI guidance to a one-size-fits-all model. CELT can play a critical role in sharing emerging strategies across the institution, and building a repository of effective pedagogical practices and interventions.
AI literacy and competency should also be integrated into curriculum and program design where appropriate rather than treated as separate or supplemental skills. CELT can support this work through the Periodic Program Review process, targeted department/school consultations, and individual course-design support. TMU Libraries’ liaison librarian program can also be leveraged to scaffold AI literacy classes and/or asynchronous learning modules into the curriculum in alignment with program and course learning outcomes. These processes can help programs consider where AI-related competencies belong in the curriculum, how they connect to existing learning outcomes, and how they should be developed across courses. In this sense, AI competencies should become part of the university’s academic fabric, relevant to the learning outcomes of our programs, not an additional training requirement that sits outside the curriculum.
Perspectives on AI across our teaching community are diverse and our approach to these technologies in learning and teaching must be inclusive of this broad spectrum of views. Community consultations revealed both a high degree of skepticism, concern, and resistance and a high degree of ambition, optimism, and innovation concerning the use of AI in the learning environment. Guidelines, policies, training, and resources must be respectful of academic freedom, recognizing the multitude of approaches that our teaching colleagues will take toward the use of AI in the curriculum, which will vary by discipline, level of study, and local course contexts.
Assessment reform is especially important. To secure assessments and support authentic learning, TMU will need to prepare for growing demand for in-person assessments. While some secure and identity-verified assessments will continue to have a place, “AI-proofing” alone is increasingly insufficient as a long-term strategy. The more important question is what kinds of learning, judgment, expertise, and professional discernment students need to demonstrate, and how programs can design assessment systems that provide confidence in those competencies. This may involve a program-level “assurance spine” of key assessment points, such as oral defences, in-class labs, applied demonstrations, supervised exams, or other forms of authentic verification. It may also involve more creative and authentic assessments that ask students to demonstrate critical judgment, authentic voice, disciplinary reasoning, ethical decision-making, and complex problem-solving.
The goal is not solely to prevent inappropriate AI use. The deeper goal is to ensure that students continue to develop meaningful academic, professional, and human capacities in a context where AI is becoming part of the broader knowledge environment. This requires programs to ask what students must be able to do independently, what they should learn to do with AI support, and what forms of judgment they need in order to use these tools responsibly.
It is also imperative that TMU approach AI literacy as a dynamic and evolving skill set. Students, faculty, and staff will require different forms of support. For students, this includes developing the capacity to make ethical and effective decisions about when, how, and whether to use AI, such as effective use, understanding and evaluation of AI technologies (see TMU Libraries’s AI Fluency Badges). For faculty, this includes support for assessment redesign, course design, data privacy, transparency with students, and the modeling of appropriate AI use. These supports need to recognize that members of the university community are at very different stages of readiness and comfort with AI technologies.
CELT will continue to play a central role in advancing the work of AI adaptation in learning and teaching in partnership with the Libraries, the Chang School, Academic Accommodation Support, Student Life and Learning Support, academic programs, and other institutional partners. Supports may include online modules, course design and assessment resources, facilitated departmental sessions, communities of practice, working groups, and hands-on collaborative design support. This approach recognizes that institutional capacity will not be built through one-time training alone. Expertise will need to scale through sustained professional learning, peer exchange, student partnership, and discipline-specific application.
It is also vital that AI adaptation in learning and teaching is supportive of the accessibility needs and benefits that students may yield through the use of AI as an assistive technology. The use of AI technologies for accessibility hold great potential for expanding access to learning and fulfilling the promise of higher education for all students.
In summary, TMU’s approach to AI in learning and teaching should be guided by responsible innovation, academic rigour, ethical use, disciplinary relevance, student partnership, and continuous capacity-building. Rather than imposing a single model of AI adoption, the university should create the conditions for thoughtful and context-sensitive experimentation across programs and departments. This will allow TMU to respond to AI not only as a challenge to be managed, but as an opportunity to strengthen teaching, learning, assessment, and the relationship between university education and the changing world of professional practice.
Encouraging innovation, creativity, and critical thinking in the use of AI in university research is crucial for advancing knowledge and addressing complex societal challenges. An open approach, however, must be carefully balanced with reflexivity and robust frameworks for security, privacy, and ethics to prevent misuse, protect sensitive data, ensure the equitable and responsible use and development of AI technologies, and protect the integrity and quality of research outputs. This includes recognizing the limitations of AI tools and identifying strategies to minimize these when using and developing AI tools. Striking this balance fosters an environment where academic freedom and groundbreaking research can flourish while upholding public trust and minimizing potential risks.
Ontario universities are driving the province’s AI innovation ecosystem. We are at the forefront of investing in and supporting the next generation of research talent, innovations, and intellectual property arising from AI technologies.
The recently released report on AI from the Council of Ontario Universities reinforces the requirements for research, including infrastructure, tools, secure computing and data sovereignty. It lays the groundwork for strengthening and sustaining AI research talent, skills development and collaboration across the sector on shared multi-modal and scaled platforms. The report addresses the opportunities to improve administrative efficiencies and the adaptation of AI tools that can improve the staff experience and allow institutions to redirect time and resources toward higher-value activities, including student support, strategic planning, partnership development and support for research.
To yield operational efficiencies from AI technologies, TMU will need to balance administrative progress with rigorous risk management to foster safe innovation and application of AI technologies to administrative work at the university. Central to this approach will be the establishment of safe experimentation spaces, which will offer hosted, expert-guided workshops (such as targeted sessions on utilizing centrally supported tools like Gemini and NotebookLM) to help staff safely experiment with prompt engineering and output verification.
Operational safety can be further reinforced through a tiered approval model that categorizes administrative tasks by risk level. While low-risk applications, such as formatting data, summarizing meeting notes, or drafting routine internal memos, should be permitted using approved tools without individual sign-offs, higher-risk use cases involving student records, human resources, or software development will require formal management review and explicit authorization. A university-wide community of practice would support these initiatives and encourage collaborative knowledge-sharing, provide an avenue to distribute peer-supported prompt libraries, and highlight successful innovative projects to promote continuous institutional improvement.
The safe use of AI tools for operational efficiencies requires five components: (1) Data governance and classification guidelines to explicitly detail which tiers of data (public, confidential, or highly sensitive) can be entered into approved tools without compromising information security; (2) a centralized use-case registry and tool repository to maintain complete institutional visibility and eliminate shadow IT procurement; (3) clear approval chains to define precise escalation workflows for evaluating new departmental AI initiatives, ensuring a human remains in the loop for high-risk deployments; (4) vendor AI assessments to enforce strict procurement controls, requiring third-party disclosures of AI integrations and ongoing vendor risk evaluations; (5) an AI incident response guide to address algorithmic errors or data security lapses, alongside comprehensive training and education programs that equip university staff with the skills to identify AI-related risks and champion responsible use.
Adaptation of university operations to AI should also ensure that efficiencies contribute to the broader institutional mission, including supporting student success. Special attention must be paid to areas of operations that are student-facing with particular focus on advising and counselling where human-interaction is essential and risks associated with data privacy and security are paramount.
3 See Appendix A for a summary table outlining this framework.
4 Norrie, James L, “Students are becoming AI fluent. Universities aren’t”, University Business, April 10, 2025. Accessed July 6, 2026. https://universitybusiness.com/students-are-becoming-ai-fluent-universities-arent/ (external link)