Last year, Jisc launched a pilot to explore how AI could help reduce workload around marking and feedback. That work continues this year, with KEATH, Graide, and TeacherMatic being piloted across more than 30 participating colleges and universities.
We’re now launching a new strand of the pilot — one that will see participants using general-purpose AI tools such as ChatGPT, Claude, Gemini, and Copilot to support marking and feedback. You can learn more about this pilot in our announcement blog.
Central to this new strand will be the good practice toolkit, which will include guidance on how to create Custom GPTs and make them secure. In parallel, we are also publishing good practice case studies from colleges and universities whose innovative approaches have inspired and shaped this pilot.
In this post, you’ll learn about the pioneering work that London South Bank University has been doing around AI in marking and feedback. You may also be interested in a parallel post, exploring Kirklees College’s work in this area.
London South Bank University – Feedback Tutor in Your Pocket
Outcomes at a Glance
Quantitative Impact:
- Average grade in Therapeutic Radiography increased from 51% (2023/24) to 54% (2024/25).
- Fail rate reduced from 20% to 17%, while first-class marks doubled from 5% to 10%.
- Across Law & Technology and ICT Project Management in Practice, 68% of surveyed students (n=37) rated the Custom GPT as helpful or very helpful.
Qualitative Feedback:
- Students reported clearer structure, improved referencing, and greater confidence in meeting assessment criteria.
- Staff reported reduced workload during draft reviews, greater feedback consistency, and enhanced dialogue around learning.
- “It helped me understand what I needed to improve, like having a tutor on call 24/7,” (final-year Law student).
- “The tool has made feedback more consistent and less time-consuming,” (academic participant, pilot cohort).
Overview
What generative AI tools were developed and used?
LSBU developed a suite of Custom GPTs using ChatGPT Plus (and planning to move to Team licences in 2025–26) to support formative feedback and marking. Each Custom GPT is configured with module-specific documents – including coursework briefs, marking rubrics, and learning outcomes – to provide feedback aligned to academic criteria while maintaining ethical, human oversight.
How were they used?
Students uploaded coursework drafts to the Custom GPT, which provided constructive, rubric-aligned feedback. Students then revised their work and submitted a short reflection on how they used the feedback. Staff used similar GPTs to generate indicative feedback or marks under human review for moderation. The project spanned multiple disciplines, including Law, Computing, and Health.
Motivation and aims:
The initiative aimed to:
- Improve feedback quality and timeliness.
- Strengthen student engagement with feedback and assessment literacy (the ability to understand assessment criteria, use feedback effectively, and make informed judgements about one’s own work and performance).
- Reduce staff workload through AI-assisted consistency.
- Embed ethical, policy-aligned AI practice in teaching and assessment.
Impact
The Feedback Tutor in Your Pocket pilot produced measurable improvements in student outcomes and satisfaction. Statistical gains were supported by survey evidence and qualitative interviews. Staff reported faster turnaround times and greater fairness, understood as improved consistency and reduced influence of individual assessor style or assumptions, when AI tools were used as a supportive mechanism.
Evaluation metrics combined grade analysis, student surveys, and reflective feedback logs. Analysis demonstrated improvement in clarity and alignment with criteria, supported by LSBU’s legal and ethical review confirming compliance with GDPR and the (currently in draft) LSBU AI Policy 25–26.
The success of the pilot informed LSBU’s business case for institutional ChatGPT Team licences (Sept 2025), now under implementation.
Rollout and Engagement
The pilot was embedded in selected modules during 2024–25, including:
- Law and Technology (L6)
- ICT Project Management in Practice (L6)
- Therapeutic Radiography (L5)
Engagement was fostered through:
- In-class demonstrations and Moodle links.
- Mandatory student and staff guidance documents.
- Staff training workshops (15 delivered between March and July 2025).
- Peer support networks and AI champions within each School.
Over 200 academic and professional staff received training, with staff feedback collected via Microsoft Forms. Both student and staff experiences informed subsequent revisions to guidance and policy.
Challenges and Resolutions
- Data privacy and GDPR:
Challenge: Student and staff use of personal ChatGPT accounts risked data exposure.
Resolution: The LSBU AI Working Group introduced strict guidance on disabling model training, anonymising data, and transitioning to institutional licences with admin oversight. - Staff readiness and variation in confidence:
Challenge: Differences in digital literacy and uncertainty around AI ethics.
Resolution: Delivery of customGPT workshops across the institution, AI literacy courses via Moodle for staff and students, plus one-to-one mentoring. - Academic integrity and student misuse:
Challenge: Ensuring AI use enhanced, not replaced, critical thinking.
Resolution: LSBU’s AI literacy “Driving Licence” for all staff and students and Originality and AI Statement form within assessments, requiring reflection and disclosure of AI use. - Resource constraints:
Challenge: Limited ChatGPT Plus licences during pilot.
Resolution: Business case submitted for 20 institutional Team licences for 2025–26, with scalability plan.
Next Stages
Future actions include:
- Deployment of ChatGPT Team licences with central IT and legal oversight.
- Expansion of Feedback Tutor in Your Pocket across LSBU Schools.
- Evaluation of sector replication potential via Jisc partnerships.
- Further research on the effect of GenAI tools on students’ analytical and evaluative skills.
In Action
From a student perspective:
A Law student uploads her coursework draft to the Custom GPT. Within moments, she receives structured feedback aligned with her module’s marking rubric: comments on argument structure, referencing, and clarity. She revises her essay, reflecting on the GPT’s suggestions before submission. Her tutor later uses similar GPT to structure the feedback and compare marking consistency, ensuring fairness and transparency.
From a staff perspective:
A lecturer in Computing uses the GPT to draft initial feedback summaries for 60 final-year students. Each AI-generated comment is reviewed, edited, and verified before release. The lecturer reports reduced marking fatigue and greater consistency. Students appreciate the clearer rationale behind marks and improved turnaround times.
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
