Mark Avery, is a senior lecturer in Education and PGCE Secondary Computer Science at UWE. Having taught in and led computer science departments for more than two decades, he brings a wealth of experience in curriculum development, digital pedagogy, and the integration of emerging technologies into teaching and learning.
His research focus includes technology enhanced learning, online learning, and the use of GenAI within the curriculum across a myriad of teaching and learning workflows. His current research explores how GenAI chatbot assistants, provide critical support for teacher trainees while on school placements.
This case study highlights how the integration of GenAI-driven tools, like bespoke created ‘Virgil’, can complement traditional mentoring structures. By deploying AI in this way, we offer just-in-time support that improves confidence, encourages self-reflection, and enhances academic engagement. This project then, reflects a commitment to exploring emerging technologies to address educational challenges.
Reflections from this case study, show that initiatives such as this have a higher margin for successful deployment if there is a strong focus on safeguarding privacy, ensuring content accuracy, and the embedding of ethical AI practices.
The Problem
Recent changes to the Postgraduate Certificate in Education (PGCE) calendar have resulted in less time for trainee educators on-campus and more time in schools, increasing the need for timely support at a distance. While trainees have school-based mentors and university tutors mentor availability is limited due to workload constraints. Email communication is often delayed, and while MS Teams discussions offer collaborative support, there are moments when trainees need immediate, structured, and advice in a safe environment.
Key challenges for PGCE trainees include:
- Managing workload and well-being: Balancing teaching, marking, planning, and personal life.
- Understanding complex teaching concepts: Translating theory into effective classroom practice.
- Building subject knowledge: Strengthening understanding of curriculum content.
- Developing reflective practice: Engaging in structured self-reflection.
- Navigating school policies: Adapting to varying institutional expectations.
- Accessing support outside office hours: Ensuring timely assistance when planning in the evenings.
- Aligning with teaching standards: Meeting the Department for Education’s Core Content Framework and Teacher Standards.
To address these needs, I explored whether an GenAI assistant could provide structured, immediate, and resource-informed guidance. Thereby offering just-in-time support while escalating critical queries when and if necessary.
The Solution
The GenAI assistant, informally named ‘Virgil,’ was designed to aid progression against the Department for Education’s Core Content Framework and Teacher Standards. Inspired by Dante’s Inferno, where the character Virgil serves as a guide for a difficult journey through hell, helping navigate the torments of the damned.
He is a calm, rational figure who provides wisdom, protection, and encouragement as Dante embarks on his harrowing journey. Virgil does not simply lead Dante; he teaches him how to observe, interpret, and understand human sin and divine justice.
Perfect for Education!
Key features included:
- 24/7 availability: Providing support outside traditional office hours.
- Structured, evidence-informed responses: Reducing confusion in pedagogical advice.
- Support for learning theory, assignments, and pedagogical strategies: Ensuring alignment with training frameworks.
- Scaffolded reflection prompts: Enhancing self-assessment practices.
- Subject knowledge support: Addressing specific curriculum needs.
Initially, Virgil was developed as a ChatGPT prototype, accessible via a web link, before transitioning to a ChatGPT Assistant with Open WebUI. Later, it moved from GPT-3.5 to 4o, improving response accuracy and contextual understanding.
The assistant was structured with prompts defining its role, objectives, tone, key areas of focus, standards alignment, and communication guidelines. These constraints helped minimise GenAI hallucinations, ensuring responses were sourced primarily from provided materials such as programme handbooks, DfE guidance, Blackboard resources, and assignment support.
Development was halted due to UWE’s AI policy restrictions, particularly regarding data privacy and the use of unauthorised third-party platforms, but support from Digital Learning and IT Services has since enabled access to MS Copilot Studio, allowing further compliant development. For their support in this regard, I would personally like to thank Sam Crowley, Tom Peake, and Richard Manley, as gaining support and time for this exploratory work has been especially difficult.
Challenges
Several technical and logistical challenges emerged during development:
- Embedding issues: The chatbot did not render correctly when embedded into Teams, Blackboard, or Xerte, likely due to cross-browser protection.
- Documentation gaps: OpenAI’s documentation did not always align with rapid platform updates.
- Privacy and copyright concerns: Ensuring compliance with data security policies and ethical content use.
- Student expectations: Trainees valued the tool but were conscious of potential fact-checking, plagiarism risks, and GenAI biases.
- Content depth: Requests for expanded resources had to be managed within copyright limitations.
Despite these challenges, Virgil successfully provided just-in-time support, bridging gaps in traditional mentoring structures.
Feedback from Students
Trainees responded positively to Virgil, appreciating its immediacy and structured guidance. Worth noting is that while effective for quick queries, some found responses lacked depth for complex topics, suggesting follow-up questions and interactive dialogues as improvements, and highlighted the need for further curated resources.
“It was great to get an answer straight away rather than waiting for an email or a tutorial.“
Student A, PGCE
“Sometimes I just needed a quick reminder, and the chatbot was perfect for that, I felt more confident asking things I wouldn’t want to bother a tutor with.“
Student B, PGCE
“Having something available 24/7 meant I could get help whenever I needed it, even late at night.”
Student C, PGCE
Next Steps
The next phase involves rebuilding Virgil using MS Copilot Studio within UWE’s AI compliance framework, with the aim of creating a template model adaptable for other departments. Expansion possibilities include:
- Developing real-time, multimodal GenAI interactions incorporating text, visuals, and voice support.
- Supporting other PGCE secondary subject areas.
- Integrating GenAI into asynchronous courses for continuous student support.
- Exploring GenAI questioning on access to Bandura’s sources to highlight possible intervention strategies.
- The ability to process text, speech, images, and even video content to provide a wider range of media outputs as responses.
- Support for generated explanations through diagram, illustration, and concept maps.
Top Tips
- Identify key support areas and integrate within existing systems: Align GenAI tools with student needs and institutional platforms like Teams and the Power Platform.
- Encourage engagement and adapt to learning contexts: Showcase GenAI benefits and tailor usage for different educational settings (asynchronous, hybrid, in-person, etc).
- Balance AI with human support and set clear boundaries: Use GenAI as a supplement, not a replacement, defining its role transparently.
- Ensure accuracy and ethical compliance: Minimise GenAI errors through structured prompts and curated data while prioritising privacy and security.
- Continuously improve based on feedback: Regularly update GenAI responses and refine capabilities through student and tutor interactions.
Conclusion
Virgil demonstrated the potential of GenAI as an on-demand, structured, and scalable support system for PGCE trainees. While it did not replace human mentoring, and nor should it – it did enhance accessibility, efficiency, and self-directed learning. An unexpected benefit was gaining insights into trainee concerns by reviewing their GenAI interactions, allowing me to refine face-to-face teaching and programme support. This provided valuable insight into common trainee concerns and questions, helping shape future support strategies. This was akin to having a critical friend and shows how GenAI can serve as an ally, helping not only student teachers, but also lecturers in supporting trainees navigate their challenging journey while maintaining pedagogical integrity.
As GenAI integration in education advances, the question is no longer whether GenAI will impact learning but how educators will shape its role. By embedding GenAI as a supportive ally rather than an authoritative instructor, we ensure it enhances learning without diminishing the human elements of mentorship, reflective practice, and professional development.
Should you wish to connect and find out more about this work in greater detail, please do reach out – mark3.avery@uwe.ac.uk