May HE AI community meetup - Artificial intelligence

Generative AI continues to raise new questions for higher education institutions, particularly around assessment, governance and implementation. At our May HE AI community meetup, colleagues from across the sector came together to explore these issues using our usual lean coffee format. This post summarises some of the main themes and reflections shared during the discussion.

Assessment redesign in response to AI

Assessment was once again a major focus, especially how institutions are redesigning assessment in response to generative AI.

Participants shared examples of more reflective and process-based assessment, including portfolios, mini-vivas and iterative designs. Several participants stressed that the most effective changes came when assessment was redesigned as a whole, rather than by simply adding a short viva to an existing essay task.

One example focused on programmes where students use AI tools throughout the learning process. Assessment focused on students’ thinking, decision-making and reflection rather than the final output alone. In these cases, students are asked to explain why they chose particular tools, how they used them and what difference those tools made to their work.

That emphasis on process rather than product came up repeatedly throughout the meetup.

Several participants discussed institutional experiments with mini-vivas. Experiences varied. Some described approaches that increased student anxiety and created practical challenges around workload, scheduling and assessment delivery, while others shared more positive examples where vivas formed one part of a broader portfolio-based model.

A particularly important point was the need to communicate clearly with students. Participants noted that vivas can easily be seen as high-stakes checks on authenticity unless their purpose, weighting and place within the wider assessment strategy are explained well.

The conversation also highlighted how much this varies by discipline. Some subjects, including nursing and other practice-based areas, already use observational or practice-based assessment. In other disciplines, large-scale assessment redesign may be much more difficult to implement.

Participants from art and design also pointed out that dialogic and reflective approaches are already common in many programmes through crits, reviews and peer discussion. Even here, though, new challenges are emerging as students use AI to support presentations, translation and reflective discussion.

Across the meetup, there was broad recognition that assessment redesign takes significant institutional effort. Participants raised concerns about workload, quality assurance, revalidation timelines and the challenge of scaling new approaches across large cohorts.

AI assessment scales in practice

The meetup also returned to the ongoing conversation around AI assessment scales and frameworks.

Some institutions said they are embedding AI assessment scales into formal assessment processes. While the idea was seen as helpful in principle, participants were clear that implementation is proving much harder in practice.

Some described models where module leaders select acceptable levels of AI use for each assessment. Even so, participants repeatedly reflected on how difficult it can be for staff and students to interpret those boundaries consistently, especially in unsupervised work.

Questions of enforceability also came up repeatedly. Participants noted that, in many contexts, it may be practically impossible to distinguish clearly between different levels of AI use.

Several attendees felt the scales work best as prompts for discussion rather than as rigid enforcement tools.

Some participants described using these frameworks to support conversations within programmes, helping staff and students discuss acceptable forms of AI use together. Even then, they reflected on how easy it is for the process to become a box-ticking exercise unless it is grounded in meaningful discussion about teaching and assessment.

Several participants also suggested that institutions may be reaching the limits of what broad institutional guidance can achieve on its own. Increasingly, discipline-specific examples, local experimentation and shared use cases are needed to show where frameworks genuinely support teaching and assessment practice and where they create confusion.

A recurring theme was that these scales are still relatively new. Many participants felt institutions are still working out how useful they are in practice and where further refinement is needed.

Acknowledging AI use in teaching materials

Another significant thread in the meetup was whether and how staff should acknowledge their use of AI in teaching and learning materials.

Participants shared a range of institutional approaches. Some institutions already have formal guidance for both staff and students on acknowledging AI use.

One institution described guidance that encourages staff and students to acknowledge:

The tool used
The version of the tool
How prompts were developed
How outputs informed the final work
Where human editing and verification took place

Several participants argued that this kind of transparency helps students understand what responsible and ethical AI use looks like in practice. Others noted that when staff model good use themselves, it can support wider AI literacy across the institution.

At the same time, there was still uncertainty about how far acknowledgement should go.

Participants reflected on how institutional language is shifting from referencing AI, to disclosure, and now increasingly to acknowledgement. Some questioned whether words such as disclosure unintentionally suggest wrongdoing, even when AI use is part of a normal and iterative working process.

Others highlighted how hard it is to define when AI use becomes substantial enough to warrant acknowledgement. In some cases, AI may only play a small part in developing teaching materials, which makes consistent expectations difficult to set.

The meetup also raised wider questions about quality and trust. Some participants noted that poorly used AI-generated materials, including inaccessible slide decks or low-quality imagery, can affect students’ perceptions of teaching quality.

Managing non-approved AI use and institutional risk

The meetup also explored growing concerns about governance, procurement and the use of non-approved AI tools.

Participants discussed the tension between encouraging innovation and meeting data protection, procurement and security requirements.

Several attendees noted that some staff are experimenting with a growing range of external AI tools and services, often outside formal procurement routes.

At the same time, participants acknowledged that trying to restrict AI use completely is unlikely to work.

Several participants questioned whether tightly controlling personal AI use is realistic. Others warned that overly restrictive approaches could leave staff and students disconnected from the tools and practices emerging in industry.

The conversation also explored the growing use of AI tools outside institutional systems and approval processes.

Participants highlighted how quickly AI meeting bots and automated note-taking tools have appeared across teaching and meetings. Many institutions are now updating guidance and settings at pace. Participants also raised concerns about data being processed outside the UK and staff understanding of wider governance responsibilities.

The meetup also highlighted the importance of AI literacy, data literacy and media literacy. While policy can sometimes be developed quickly, building the confidence and shared practice needed to apply it well.

AI sustainability and environmental impact

The final major theme was sustainability and the environmental impact of generative AI.

Participants discussed growing interest and concern around the energy, water and infrastructure demands associated with large language models and generative AI systems.

The conversation also moved beyond simple narratives about AI on its own.

Several participants argued that conversations about AI sustainability need to sit within wider questions about digital sustainability, including cloud storage, streaming, online meetings and wider digital habits and online activity.

Some also reflected on the growing volume of automatically recorded online meetings and questioned whether institutions are paying enough attention to the environmental impact of digital storage more broadly.

Others noted that AI may be acting as a prompt for wider institutional conversations about sustainability that go beyond AI tools themselves.

Participants also highlighted how hard it is to access clear and transparent environmental impact data from major AI providers. Several participants reflected on the challenge of moving beyond headlines and towards more evidence-informed discussion.

At the same time, there was broad agreement that AI should not be treated as a separate sustainability issue. Instead, participants felt it needs to be considered alongside wider questions around digital sustainability, including how institutions use cloud storage, online meetings and other digital tools.

Looking ahead

June’s meetup will also be our final HE AI community session before the summer break. Throughout the year, participants have shared a wide range of discussion topics on our Padlet boards, many of which we have not yet had time to fully explore during the sessions themselves. The community has already requested that we revisit these topics during June’s meetup and use them to help shape the discussion.

Thank you to everyone who contributed to this month’s discussion and shared examples, experiences and resources.

Our next higher education AI community meetup will take place on Tuesday 16 June 2026. If you’d like to join the next session, please Join the AI in higher education community meetup list