March HE AI community meetup - Artificial intelligence

Our March HE AI community meetup brought members together from across the sector to explore the questions institutions are facing as AI capabilities continue to evolve. Using our usual lean coffee format, the conversation focused on agentic AI, student support, academic integrity and assessment practice. Here’s a summary of what you shared.

Agentic AI in practice

Agentic AI was the most voted topic this month, and it’s clear why. There was a strong sense of uncertainty about where the line sits between chatbots and genuinely agentic systems. Several members reflected that this confusion is being reinforced by inconsistent language from some vendors. Jisc’s agentic AI primer was also shared during the discussion, which helps to clarify how agentic AI is currently being described and understood.

Some institutions have begun piloting tools with agentic features. Early feedback was mixed. In some cases, there is clear potential. In others, the tools feel early-stage or require more structure and input than expected. The gap between expectation and reality came up more than once.

Participants also reflected on the difference in skills the shift to agentic systems may require, and how we might need to bring these aspects into AI literacy in the near future. Working with agentic systems is not just about prompting. It involves decision-making, oversight and delegation in a more active sense. There was some discussion about how these skills, particularly delegation and oversight, are not always emphasised in academic contexts, and what this might mean for both staff and students

There were also questions around governance, with a need for shared understanding and clearer guidance as institutions begin to explore agentic AI more widely.

Building student capability

The conversation then moved to student AI training, specifically how we can address reluctance to engage with student focused training. This remains an area where there is clear intent, but uneven delivery.

Many academics want to support their students but are unsure where to start. Time and confidence continue to come up as the main barriers. At the same time, where training is available, particularly through library or digital education teams, engagement tends to be strong.

There was a consistent message around relevance. Discipline-specific examples appear to land far better with students than more general guidance. Approaches such as short, modular content, collaborative delivery and the use of local champions are helping to spread the workload.

There was also an interesting counterpoint from some institutions, where students themselves are not always fully receptive to using AI, particularly in assessment contexts.

AI Slop

Concerns around low-quality AI-generated writing, sometimes referred to as “AI slop”, resonated strongly across the group. This is not just a student issue. Participants noted that it is beginning to surface in staff work as well.

One example raised was in doctoral contexts, where concerns around originality and authorship were discussed, although similar questions are emerging more widely. Some institutions are responding by introducing more opportunities for students to explain and defend their work, through viva-style elements, oral discussion or process-based assessment.

There was also interest in using AI itself as part of the solution. Asking students to critique or evaluate AI-generated outputs was seen as one way to build more critical engagement. A recurring point here was that this is unlikely to be solved through policy alone. Ongoing, faculty-level conversations were seen as just as important.

Assessment approaches

The discussion then turned to AI in assessment more broadly. As in previous sessions, there were mixed views, particularly around feedback.

Some participants were cautious about the use of generative AI in feedback, raising concerns about losing the relational and personal aspects of the process. Others shared examples of more targeted use, particularly where tools support efficiency but do not generate feedback themselves. One example discussed was the use of classification-based feedback tools such as Graide, which draw on existing feedback rather than generating it from scratch. In this approach, academics remain fully in control of the feedback. Participants noted that this can help maintain student trust.

Where there is a clear human in the loop, and transparency about how tools are used, there appeared to be greater acceptance.

AI detection tools were also discussed. The general view remains that they are not reliable enough to be used in isolation, particularly given the risk of false positives. Instead, many institutions are focusing on clearer expectations, setting out how AI can be used within each assessment and what students should declare. As with previous meetups, the conversation often returned to assessment design.

What’s next

Thank you to everyone who joined and contributed. As always, the value of the session came from the openness of the discussion and the willingness to share what is, and isn’t, working in practice.

Our next HE AI community meetup takes place at 15:30 on Tuesday 21 April. We look forward to continuing the conversation.

Links shared during the call

Find out more by visiting our Artificial Intelligence page to explore publications and resources, learn more about our communities and sign up for our AI Literacy training.

For regular updates from the team sign up to our mailing list.

Get in touch with the team directly at AI@jisc.ac.uk