This page offers context for understanding what agentic AI is and why it matters in relation to student learning.
Agentic AI tools can autonomously take actions on a user’s behalf. These tools are designed to complete multi-step tasks with minimal user direction
In educational settings, agentic AI can navigate a learning management system, consult linked course resources, open quizzes or activities, select answers and submit them, all without ongoing user input
This activity is currently undetectable within Brightspace (Avenue to Learn) or other learning management systems.
As AI tools become more capable of autonomously completing academic tasks, approaches that rely primarily on restricting tools or detecting misuse become increasingly difficult to sustain at scale.
This does not mean academic integrity no longer matters. Instead, it highlights the limits of detection-based approaches on their own and points to the need for broader, more durable strategies.
A growing body of higher education research and commentary argues that supporting academic integrity will require structural changes to how assessments are conceived and implemented (e.g., Corbin, Dawson, & Liu, 2025).
Meaningful changes take time, vary by disciplinary and course context, must be balanced against workload, class size, and curricular constraints. There is no expectation that these challenges will be resolved individually or immediately. These shifts also call for coordinated action by the university.
In many courses, online assessments are intentionally designed to support practice, retrieval, and ongoing learning development.
When students use agentic AI to complete these activities, a significant consequence is lost learning opportunities. Students may appear successful in the moment while undermining their own preparation for later, more integrative, or invigilated assessments.
This can create a mismatch between performance on low-stakes or practice activities, and actual readiness for higher-stakes demonstrations of learning.
You are not expected to navigate these questions alone. Support is available for instructors, academic units, and programs working through how assessment can remain meaningful in the context of generative and agentic AI.
Consultations and program-level support
Teaching and Learning Consultations: support for course-level assessment design (you can find your Faculty key contact here)
Curriculum and Program Design Support: for coordinated, programmatic approaches
Authentic and meaningful assessment
Principles and examples that emphasize application, judgment, and context
Differentiating assessment of learning from assessment for learning
Guidance on aligning assessment purpose, integrity, and learning outcomes
A growing body of commentary argues for a more candid approach: being clear about what instructors and institutions can and cannot control.
This has meant reframing conversations with students around discernment and agency. Instead of focusing only on whether AI is permitted, these discussions emphasize:
These conversations are most effective when they reinforce assessment structures that clearly differentiate between when learning is being supported and when it is being evaluated. Talk alone cannot resolve integrity challenges. But when paired with thoughtful assessment design, these conversations can help align expectations, reduce confusion, and support students in making more intentional decisions about their learning.
Instructors may find it helpful to ask students questions such as:
Instructors who have found effective ways to discuss AI use with students are invited to share prompts or approaches that worked well in their context.
Agentic AI tools often require access to accounts, platforms, or documents in order to carry out automated tasks. In some cases, this can involve:
Instructors and students should remain mindful of privacy, data security, and institutional policies when considering the use of these tools.
Learn more:
Privacy and security guidance for generative AI at McMaster