ID002 Every Student Gets a Personal AI Tutor

The Idea

Keele should introduce a bold, institution-wide initiative: every student is provided with a personal AI tutor. This would be a secure, University-governed AI tool embedded within our digital learning environment, designed to support students throughout their academic journey. 

The AI tutor would not replace academic staff, but would complement teaching by offering 24/7 personalised support. It could help students understand complex concepts, plan their studies, reflect on feedback, prepare for assessments and navigate university systems and support services. It would be aligned to our curricula, assessment approaches and academic integrity principles, rather than relying on unregulated external tools. 

The AI tutor should exist primarily to provide an academic learning support function. Here it could help explain concepts and give challenging questions in various ways to help ensure they are understood. AI is also good at interpreting feedback and suggesting study patterns to help students stay on track academically. It can also help with signposting students to the many and varied services on offer. The clear boundary is with academic judgement and pastoral support, which are firmly in the domain of the qualified and skilled human specialist. 

It is not envisaged that Keele would build an in-house AI model. Evidence suggests a “foundation model” based on, for example, MS Copilot, and university-curated data that currently exists in modules (for example Blackboard), handbooks, policies and procedures.  

It’s suggested that the AI should be designed to support the thinking process, not give answers. The “Socratic model” is relevant here – it should act as a thinking partner that challenges user assumptions, promotes deep learning, and facilitates personalized, iterative dialogue: i.e. promote questioning over answering, encouraging the student to do the deep thinking. Guardrails should exist to prevent the AI from being an active source of answers, which should mean that it is never the source of content that gives rise to safeguarding concerns. 

A particular guardrail would be for the AI to always signpost the student to human-centred support at the first sign of any complex situation being discussed. It is envisaged that students would be trained in its use and would be thoroughly versed in the fact that it is an assistant to encourage deep thinking and provide signposting, not to be the solver of challenging situations and problems. This initiative would position the University as a sector leader in ethical, purposeful use of AI in education. It would also directly support student success and inclusion by reducing hidden barriers to learning, particularly for students who may lack confidence, academic capital, or flexibility in when and how they study. 

As outlined below, some students arrive lacking confidence or the family background to properly make a good start to their time at university. Having a non-judgemental assistant that can interpret questions thoughtfully and signpost the student to agreed answers and solutions could be an extremely valuable facility. Rather than reacting to student use of AI, the University would actively shape how AI is used to enhance learning, develop critical digital skills, and prepare graduates for a future where working effectively with AI is a core professional capability. 

It is acknowledged that AI approaches are being trialled elsewhere in the sector, however this idea is grounded in the deliberate and intentional use of AI, embedded in the student experience, to help give equitable AI-supported learning and to democratise the process. We are moving away from standalone, generic tools to something that is built into the curriculum, recognising how this approach mirrors the new world that students are now emerging into on graduation. Whilst some university approaches appear to be reactive, this would hopefully demonstrate an active, supported and governed use of AI. It gets away from students being uncertain how acceptable AI tool usage is, to something that is embedded in the strategy as a universally available resource. It is argued that this approach signifies to the sector that we are leading responsible and meaningful use of AI in the curriculum. AI usage will happen anyway, so it’s important that we take the initiative and put in place guardrails to make sure it is properly educationally grounded. 

It has been suggested that this may disadvantage widening participation students, however I would argue the opposite. As a first-in-family to attend university myself, I would have loved the opportunity to ask questions that to others may have seemed obvious, in a way that does not signal “basic ignorance”. It would be a risk-free way to make all those fundamental enquiries that other students seem to arrive at university already knowing much of the necessary information. On that basis, it is asserted that there are substantial positive aspects to this for students coming from a background that is non-traditional. 

A recent OfS article highlights a particular current challenge that this idea would go a long way to solve: "How does your approach support the AI literacy required by the labour market without displacing opportunities to develop deep skills and acquire knowledge critical to a student’s discipline?" 

The primary focus of the academic aspects of the AI tutor would be to thoughtfully question and probe to encourage deeper thinking in the mind of the student. This, combined with impactful and meaningful training would help to ensure students adopt a critical and ethical approach. 

Why This Idea Should Be Considered

Students are already using AI tools, but in uneven, unregulated, and often unsupported ways. By providing a Keele-designed AI tutor, we can ensure consistent, ethical and pedagogically sound use of AI. 

Comprehensive training to ensure students derive the best outcomes would be provided, in the form of a compulsory induction module that is common across all routes. Similar ground-up training would be indicated for staff, so that any concerns may be mitigated at an early stage. 

This idea helps the University stand out in a crowded higher education market by offering a tangible, distinctive benefit to students. It supports strategic priorities around student success, widening participation, digital innovation, and future-ready graduates, while reinforcing Keele’s role as a responsible leader rather than a passive adopter of technology. 

The model suggested here is something to complement, not replace, the human tutor. It means students have 24/7 access to a support facility that can help them answer questions by thoughtfully guiding their thinking through a process of constructive questioning. Yes, the environmental impacts of AI are well documented – the challenge here is to participate in what will clearly be a competitive domain whilst being mindful of the impact. As all universities have to do, we must be up-front in recognising the environmental impact of AI and be transparent in how we are taking a responsible, evidence led approach. We would need to ensure that we are mindful to consider providers that prioritise energy efficiency and sustainability.

How We Would Implement This Idea

Implementation could begin with targeted pilots in high-impact areas such as first-year transition, large cohort modules, and widening participation initiatives. The AI tutor initiative is an opportunity for co-design involving academic staff, professional services and students. It could be embedded within existing systems such as the VLE and the Keele App. 

Strong governance would be essential, covering data protection, transparency, bias, accessibility, and academic integrity. Ongoing evaluation should be built in from the start, using evidence to refine functionality and guide scaling. 

It is important that accessibility is considered at design stage, using a platform that is compliant with relevant standards. I am not an expert in accessibility, but as a minimum the solution should support input and output via text and speech, and have the facility to adjust the output style to suit an individual learner’s needs. 

As discussed, the AI tutor would be primarily signposting and questioning the student. Any deviation by the student from this into the area of pastoral care would be a particular “guard rail” scenario that would trigger a response by the AI for them to seek mentor / SESO support. 

What Success Would Look Like

Success would mean improved student confidence, engagement and academic outcomes; reduced attainment gaps; and students graduating with strong AI literacy and ethical awareness. Keele would be recognised as a sector leader in using AI to enhance learning in inclusive, responsible, and impactful ways. 

It is suggested that we evaluate this across learning outcomes (easy to evidence using assessment outcome data), student experience (NSS / end of module surveys), APP activities (are we adding value and closing attainment gaps, particularly across at-risk demographics and groups), and graduate outcomes (a slower-burn piece of work, but using HESA outcomes data to evidence where our students are going following graduation). 

Some of these are longitudinal activities where benefits may only become apparent after several years’ usage of the facility by cohorts.