Universities have spent years trying to make digital services feel more relevant to individual students. Generative AI creates a new possibility: an experience that can respond to a student’s context, explain information in a useful way and help them navigate institutional knowledge through conversation.
But “personalised” can describe two very different models.
In one, the system quietly assembles a profile, makes assumptions and changes the experience without making those choices visible. In the other, the student provides relevant context, can see what the AI uses and retains control over the experience.
The second model is the stronger foundation for higher education.
The important question is not simply whether universities can personalise AI. It is what the experience should adapt, what must remain consistent and who controls the context.
Students want useful AI, but not an unbounded one
Student use of generative AI is no longer confined to producing answers. Jisc’s research with more than 200 students found that learners increasingly view it as a collaborative tool for coaching, active learning and critical thinking. Students also raised concerns about equity, bias and accessibility and expected institutions to provide competent support and fair policies.
That combination matters. Students do not only want access to powerful tools. They want an experience that helps them learn while giving them confidence about the boundaries.
The sector is still working out what good looks like. In May 2026, the Office for Students and Advance HE began a research project involving leaders, academic staff and students. The work will examine how AI affects student outcomes, what students expect from their institution and which barriers stand in the way of effective adoption. Findings are expected later in 2026.
Our view is that personalisation should sit inside this wider institutional response. It should not be a bolt-on feature or a substitute for inclusive teaching and student support. It should give each student a better route into approved information and appropriate help.
A responsible approach to AI personalisation
A university can adapt an AI experience without creating a different set of facts or standards for every student.
This distinction is essential.
The experience may change how it explains an academic policy, which examples it uses or how it structures a response. The underlying policy, source material and rules should remain consistent. The same applies to academic integrity, assessment requirements, safeguarding boundaries and decisions that require professional judgement.
A useful principle is: personalise the route through knowledge, not the institutional truth beneath it.
That gives universities a practical dividing line.
What can adapt
- The level of explanation and amount of background provided
- The structure of a response, such as a short summary followed by detail
- Examples relevant to a student’s course, role or stated goal
- Communication preferences and, where supported, preferred language
- Suggested next steps within the services and resources available to that student
What should remain controlled
- The approved sources used to answer the question
- Access permissions and boundaries between audiences
- Academic regulations, policies and assessment standards
- Safety, academic-integrity and communication guardrails
- Routes to qualified staff when a question needs human judgement
Personalisation becomes safer and more useful when institutions design both lists deliberately.
Layers of responsible personalisation
1. User-provided context
The most transparent place to begin is information that the user chooses to provide: their area of study, priorities, goals or preferred communication style.
Kalisa’s Personal Settings allow users to specify key facts about themselves, what matters to them and how they want AI to communicate. Kalisa states that this background information is stored securely and used only in agents where personalisation is enabled.
This is materially different from hidden profiling. The student has an active role in shaping the context. The institution can also decide where personalisation is appropriate rather than applying it to every interaction by default.
The examples below show how user-provided context can shape an AI response. Two students ask the same question, and both answers draw on the same sample academic-integrity policy. The only difference is the information each student has chosen to add to Personal Settings.
.png)
.png)
The underlying guidance remains consistent, but the route into it changes. One answer reflects a Business Management student’s preference for clear explanations and practical examples. The other uses shorter sections, stronger signposting and a law-related example for a student with dyslexia. Kalisa adapts the experience without changing the policy or academic standard.
*Fictional profiles and sample policy created for demonstration.
2. Relevant, approved knowledge
Personalisation without reliable knowledge produces a more fluent experience, not necessarily a better one.
A student-support agent should work from current policies, guidance and service information. A learning agent may need teaching materials and course-specific resources. Each experience should have a defined knowledge boundary and a named owner responsible for keeping that material current.
3. Audience and access boundaries
Students are not one audience, and students and staff should not automatically see the same information.
A university may need different experiences for applicants, current students, educators and professional-services teams. Within those groups, permissions may vary by course, campus, service or role.
Personalisation cannot replace access control. It operates after the platform has established what the user is entitled to see and do. This protects sensitive information and prevents a helpful tone from disguising an inappropriate answer.
4. User control and clear explanation
People should know that an experience is personalised, which information it uses and how they can change it.
The UK Government’s Data and AI Ethics Framework, updated in December 2025, recommends privacy by design and default, data minimisation and clear information about what data is collected, why it is needed, how it is used and how long it is stored. It also says users should understand how to manage their privacy settings.
What this looks like in practice
The University of Law and Kalisa announced a multi-year collaboration in July 2026. From September, ULaw is progressively rolling out a Kalisa-powered environment to 17,000 students and 3,000 staff.
The platform is designed to let ULaw create secure AI agents grounded in its teaching materials, policies, guidance and institutional knowledge. Students can tailor the experience around their preferences and needs, while the university retains guardrails aligned with its standards. Our aim is not merely to give students another general-purpose AI. It is to create a university-led environment that supports learning, accessibility, responsible use and preparation for professional practice.
The example illustrates the wider design choice facing the sector. Institutions can leave personalisation to consumer tools whose context, sources and rules they do not shape. Or they can create an approved experience in which personal relevance sits inside institutional boundaries.
A six-question test for university leaders
Before introducing personalisation into a student-facing AI experience, ask:
- What user need will personalisation solve? Name the improvement rather than starting with the feature.
- What may change between users? Define the adaptable elements, such as explanation, format, examples or navigation.
- What must stay consistent? Protect sources, policies, standards, permissions and safety rules.
- What context does the user provide and control? Make collection proportionate, visible and editable.
- When should the experience hand over to a person? Build clear routes for judgement, support and exceptions.
- How will we know it helps? Test usefulness, comprehension, accessibility, source quality and differences between user groups.
If a team cannot answer those questions clearly, it is not ready to personalise the experience.
The goal is relevance with boundaries
Personalised AI in higher education should not turn every student into a data profile or every interaction into a prediction.
At its best, it does something more modest and more valuable. It lets students explain what they need. It helps them approach trusted institutional knowledge in a way that works for them. It keeps the university’s standards visible. And it knows when to bring a person back into the conversation.
That is how personalisation becomes part of a responsible student experience: not by removing institutional control, but by using it to make AI more relevant, inclusive and trustworthy.
See how Kalisa can help your institution create secure, personalised AI experiences grounded in your own teaching materials, policies and guidance.
Book a free demo to see Kalisa in action
* This articles' cover image is generated by AI
%20(1).webp)