Generative AI is often associated with producing written text, but its role in education is becoming much broader. It can help students turn complex reading into audio, allow a class to question an AI through spoken conversation or enable an educator to create an immersive simulation that would previously have required a production team.
These possibilities are opening up new ways to approach teaching and assessment. They also raise questions about inclusion and the continuing importance of human judgement.
This collection brings together case studies from colleagues at the University of Northampton who are exploring those questions through their own practice. The examples developed from the In Practice: Multimodal AI at UON Padlet, where colleagues were invited to share how they were combining AI with different forms of communication and learning activity.
The Padlet began as a space for exchanging emerging ideas. It has grown into a lively record of experimentation across the University, encompassing practices that are already established alongside approaches that are still being tested. The case studies presented here take some of those contributions further by exploring the educational challenge behind each activity and what colleagues learned from trying it.
About Study Smart 3
The collection forms part of Study Smart 3: Using AI for Multimodal Composition to Foster Inclusive Practice and Belonging, led by Dr Helen Caldwell and Rob Howe with colleagues from the Centre for Active Digital Education.
The project explores how AI can support students and tutors to work across different modes of communication. It is particularly interested in whether multimodal approaches can help students build confidence and express their understanding in ways that are meaningful to them.
This matters within a university community that includes students with varied educational experiences and different levels of confidence in academic English. It is also important for students who may find conventional learning materials difficult to process or who communicate their ideas more effectively through visual and spoken forms.
Multimodal AI does not simply mean adding an image to a piece of text. It involves moving between modes as part of the learning process. A student might use AI to turn a dense document into an audio explanation before returning to the original source with greater understanding. Another might develop an idea through a spoken exchange and then express it through creative practice. The mode changes, but the learner remains responsible for interpreting what the AI produces.
Find out more about Study Smart 3 on the University’s Research Explorer.
Responding to different ways of learning
One of the most powerful contributions to the Padlet came from a student who was struggling to process recorded lectures and extensive reading materials. The volume of information was creating cognitive overload and making it difficult to distinguish essential content from additional material.
After being introduced to AI projects and agents, the student began using NotebookLM to transform selected documents into audio and video resources. Being able to control the source material was particularly important because it allowed the student to work within a defined collection rather than search across unknown information.
The technology did not remove the need to engage with the course content. It changed its form, making it more manageable for a student with ADHD and dyslexia. The student described the tool as making a ‘massive difference’ to the feasibility of continuing their studies.
This example captures an important idea running through the collection. Inclusion is not always about simplifying the learning; sometimes it is about giving students another route into it.
Making learning experiences possible
Other examples show how AI can help educators create learning experiences that would previously have been too time-consuming to produce.
For a large multidisciplinary simulation at Silverstone, AI was used to create the fictional communication environment surrounding the Nexus Festival. Students encountered a live social media feed in which events unfolded throughout the day. They had to respond to disruption and manage communications under pressure.
Creating the volume of material needed to sustain that environment manually would have been difficult for one educator. AI made it possible to produce realistic posts and supporting imagery, all coordinated around the developing scenario.
A different simulation invited Marketing students to make strategic decisions for a live client project. An app created with Google Gemini allowed teams to select customer groups and test decisions about their proposed route to market. Students received immediate feedback and could see the consequences of their choices. The educator was then able to refine the model between teaching sessions.
In these examples, AI did not supply students with an answer. It created an environment in which they could make decisions and examine what followed.
Changing the conversation
Several of the case studies explore how AI can become part of a conversation rather than remaining a tool used privately by an individual student.
Fashion students included an AI team member in a live industry project. Each group gave its AI a role and interacted with it through voice as their ideas developed. The students did not accept every response. They questioned its suggestions and became frustrated when its contributions were repetitive. Those moments often prompted the most useful reflection.
In another Fashion workshop, students investigated emerging trends on TikTok before presenting a short forecast. ChatGPT’s voice mode listened to each presentation and responded aloud, sometimes adding context and sometimes challenging the group’s assumptions.
The spoken interaction changed the atmosphere. AI felt less like a search box and more like another voice in the room. This created opportunities for discussion, but it also made the need for critical distance more visible. A fluent or confident response was not necessarily a reliable one.
Supporting students without taking over
The Padlet also contains examples of AI being used to scaffold academic work. Students have practised assessed email conversations through an AI role-play activity in NILE, while others have used structured prompts within reflective journals.
Marketing students preparing for a client pitch received AI-generated summaries of verbal feedback from their dress rehearsals. These written records helped them revisit comments that might otherwise have been forgotten after an intensive feedback session.
Elsewhere, AI-assisted translation gave students access to a relevant research article available only in Turkish. The translated material could then be compared with another source and considered within a UK context. This did not remove the need to evaluate the research. It widened the material available for students to question.
Across these examples, AI is most useful when it supports the next stage of learning. It may help a student begin or make feedback easier to revisit. What matters is that it does not become the endpoint.
Creative possibility and human judgement
Creative practice is a strong thread throughout the collection. AI has helped Marketing students visualise campaign ideas even when they do not have advanced design software skills. It has supported Fashion students as they explore visual concepts alongside sketchbook and studio work.
These activities can create a more level starting point. A student with a strong campaign idea is less likely to be held back because they cannot produce a polished visual mock-up. At the same time, easy image generation can make authorship and originality harder to see.
The Digital Tools Logbook provides one response to this challenge. Students record when they use AI and reflect on why they chose it. This makes the development of an idea visible and encourages open discussion about the decisions that remain the student’s own.
The emphasis is not on presenting AI use as inherently innovative. It is on helping students recognise when the technology has added something useful and when its contribution needs to be rejected.
A collection of developing practice
The case studies do not present multimodal AI as a finished solution. Some activities produced strong engagement, while others had limited uptake. Colleagues describe tools that did not work as expected and approaches they would change next time.
That openness is one of the collection’s strengths. It allows us to see the practical work involved in moving from an interesting AI capability to an educationally worthwhile activity.
Together, the cases suggest that multimodal AI has the greatest value when it gives students another way to enter into learning or enables an experience that has a clear purpose. The technology can change the form of an activity, but it is the surrounding educational design that determines whether the change is meaningful.
Study Smart 3 is exploring that space between technical possibility and inclusive practice. This collection shows what that exploration looks like across the University of Northampton.
