AI in Higher Education: Use Cases, Risks and Priorities
Artificial intelligence is becoming part of everyday university life, with students experimenting with generative AI, researchers finding new ways to accelerate their work and institutions exploring where automation, analytics and AI could improve services.
The opportunity is significant, but universities also face important decisions about how and where to use the technology. Where can AI in higher education create meaningful value? Which processes should institutions prioritise? And how can universities innovate without introducing unacceptable risks around data, bias or high-impact decision-making?
To explore those questions, we spoke to Tom Cadman, Digital Transformation Director at Crimson. Tom works extensively with universities in the UK and internationally, helping institutions improve student recruitment, admissions, retention and the wider student experience while navigating digital transformation and AI adoption.
From his work across the sector, Tom sees universities at very different stages of their AI journey. Some remain cautious about its impact, while early adopters are already exploring how AI could reshape university operations, research and the student experience.
For Tom, the answer is not to resist AI or adopt it indiscriminately, but to apply it appropriately while retaining human judgement where it matters.
"You've got to embrace it but equally utilise it and keep that human in the loop."
Tom Cadman, Digital Transformation Director, Crimson
Watch the Digital Lounge
In this episode of the Digital Lounge, Tom Cadman joins Mark Britton to explore AI in higher education, including where universities are finding practical use cases, the risks institutions need to consider and the priorities that should shape responsible AI adoption.
Why is AI different for higher education?
Higher education faces a particular challenge with AI because universities are considering the technology from two perspectives at once.
On the one hand, institutions are exploring how AI could improve the way the university operates, from admissions and student services to research and administrative processes. On the other hand, they need to decide how students should use AI within learning, assessment and research.
Tom argues that universities cannot simply ignore the technology when students are already using generative AI to research and develop their ideas. Instead, institutions have an important role in helping students understand what AI produces, critique its outputs and apply critical thinking.
That creates a more nuanced challenge than simply deciding whether to adopt a new technology. Universities need to consider how AI can improve institutional outcomes while also determining how to integrate it responsibly into education.
Crimson’s Universities Don't Have an AI Adoption Problem. They Have an AI Consistency Problem. develops this idea further, highlighting research presented at UCISA that found students are already developing AI behaviours through peer networks. At the same time, institutions face the challenge of providing clearer and more consistent guidance.
Read: Universities Don't Have an AI Adoption Problem. They Have an AI Consistency Problem.
Where are the most valuable AI use cases in higher education?
Tom sees opportunities for AI across the university, from teaching and learning to admissions, the student journey, research and day-to-day operations. Within those areas, potential use cases range from handling routine applicant enquiries and supporting admissions decisions to identifying students who may need additional support, accelerating research and improving administrative efficiency.
The challenge is not finding possible applications for AI, but deciding which ones are worth pursuing. Tom recommends starting with the outcomes the university is trying to achieve and working backwards from there.
For many institutions, those priorities include attracting the right students, improving applicant conversion and strengthening retention. Universities can then assess their existing processes, technology and data to understand where AI could make the greatest difference and whether the foundations are strong enough to support it.
Taking this outcome-led approach helps prevent an AI programme from becoming a collection of interesting experiments with no clear connection to institutional priorities.
How can AI improve university admissions?
Admissions is one of the clearest operational areas Tom identifies for AI, with potential applications extending from initial enquiries through to application processing and decision support.
At the enquiry stage, AI can help answer more straightforward applicant questions by drawing on relevant institutional information, reducing the volume of routine enquiries handled manually and allowing admissions teams to focus their time on cases requiring greater attention.
Further along in the application process, universities may need to assess qualifications, English language requirements, and other applicant information before making a decision. AI can help structure and standardise information and support recommendations, while keeping people involved where cases require judgement or additional review.
Faster processing also has implications beyond efficiency because, for universities competing for applicants, the time between receiving an application and communicating a decision can influence the applicant experience.
Tom explores this issue separately in Speed to Offer: The Strategic Advantage Universities Can't Ignore, which examines why response times are becoming strategically important for university admissions.
Read: Speed to Offer: The Strategic Advantage Universities Can't Ignore
AI could also support document processing across admissions, including identity documents, degree certificates and transcripts. The technology can help classify documents, extract relevant information, verify that applications are complete, and flag uncertain cases for additional review.
The important principle is that AI supports the process rather than owning the outcome. It can help structure information and identify where further attention is required, while admissions teams retain responsibility for the decision.
How ready is your student recruitment operation for AI?
Before applying AI to admissions, it helps to understand how effectively the wider recruitment operation is working today.
Crimson’s Student Recruitment Scorecard assesses how effectively recruitment supports institutional objectives, where applicants experience friction or delays, whether teams, processes and information are aligned and how prepared a university is to embrace new ways of working with AI. It also identifies areas to prioritise around conversion, efficiency and applicant experience.
The assessment takes around four minutes and provides a personalised report.
Take the 4-minute Student Recruitment Scorecard.
How can AI help universities convert offers into enrolments?
The opportunity continues after a university makes an offer, with data and AI creating new ways to support more relevant engagement throughout the applicant's decision-making journey.
A prospective student is not simply choosing a course. Students may consider the wider university experience, cost of living, accommodation, employability and what studying in a particular location could mean for their future.
Using insight to communicate those strengths more effectively could help universities create more relevant engagement and nurture applicants towards enrolment.
This moves the conversation beyond AI simply reducing administrative effort and towards supporting an important institutional outcome: helping the right students choose and successfully join the university.
Can AI help universities identify students at risk of dropping out?
Retention is another significant AI use case in higher education, particularly when universities can identify changes in student engagement early enough to offer appropriate support.
Signals such as attendance, coursework performance and financial circumstances can help indicate when a student may be at risk of disengaging or dropping out. Used appropriately, AI and analytics can bring those signals together and help identify changes in behaviour that warrant further attention.
The purpose should not be to replace human judgement, but to identify where an intervention may be needed so that an appropriate member of staff or student services team can investigate further. This could range from an initial prompt or adviser outreach through to more targeted support, depending on the individual circumstances.
Human contact and transparency remain important, particularly when universities use student information to identify potential risks. Identifying a signal is not the same as understanding an individual student's circumstances, so determining the appropriate response still requires human judgement.
Where should universities draw the line with high-risk AI?
Some technically possible use cases warrant much greater scrutiny because the consequences of getting them wrong could materially affect individuals.
Tom gives a particularly important example from work exploring student interventions, where a university considered whether information relating to extenuating circumstances could be used to identify signals indicating potential suicide risk.
Although technology might identify relevant language or warning signals, Tom explains that, following discussion with the university and its legal team, the idea was not pursued because of the potential consequences if a system failed to identify somebody at risk.
That example crystallises one of the central questions around AI risks in higher education:
Just because an AI use case is technically possible, should a university implement it?
The answer depends on much more than a model's accuracy. Leaders need to consider the consequences of false positives and false negatives, the sensitivity of the information involved, accountability for decisions and whether an appropriate human remains responsible for the outcome.
For a deeper look at this, Crimson’s What is AI governance? Why the best governance frameworks don't feel like governance explores how organisations can create practical guardrails around AI without making governance a barrier to useful innovation.
Read: What is AI governance? Why the best governance frameworks don't feel like governance
What role could AI play in teaching and curriculum design?
The impact of artificial intelligence in higher education extends beyond administration, with universities beginning to explore how AI could support the design and development of courses, modules, learning objectives and teaching activities. This creates an opportunity to rethink how institutions construct the curriculum of the future and reduce some of the effort involved in bringing the different components of a programme together.
AI could support academics by helping to draft learning outcomes and initial course content, map curricula to identify gaps or duplication and check alignment with relevant requirements. It could also help bring together information that informs programme design more quickly, giving academic teams a stronger starting point for discussion and refinement.
Academic ownership should remain central to that process. AI-assisted curriculum design raises questions about academic freedom, ownership, content accuracy, intellectual property, and the risk of programmes becoming overly standardised or reliant on automation. A stronger model is therefore human-led and AI-assisted, with academics retaining responsibility for shaping, challenging and validating what is ultimately delivered.
The opportunity is not to hand curriculum design over to AI, but to use the technology to gather and structure information more efficiently while academics provide the judgement, expertise and context needed to create the right programme.
What about AI in research?
AI can help researchers explore and validate models before progressing to physical experimentation, allowing different approaches to be tested more efficiently. Those outputs still need to be compared with real-world evidence, so the technology augments rather than replaces research expertise.
The opportunity also extends to activities such as literature review and synthesis, analysing large volumes of information, data analysis and coding support. Used appropriately, these capabilities can help researchers work through information and explore potential approaches more efficiently.
Domain expertise remains essential throughout, as accelerating analysis is only valuable when researchers can assess whether the outputs are credible, relevant, and worth pursuing.
Why good data foundations matter for AI in higher education
The quality of an AI solution depends heavily on the information that supports it, creating a challenge for universities whose institutional data is distributed across legacy platforms and departmental silos.
Many universities are still working with long-established student record systems, while data teams rely on spreadsheets and manual processes to consolidate information. Before introducing more sophisticated AI, institutions need to understand whether they are capturing the right data, whether that information can be trusted and whether sensitive data is appropriately secured.
This is where the relationship between data strategy and AI strategy becomes particularly important. AI needs access to appropriate, trusted information within a managed environment, supported by clear data governance and controls.
Microsoft Fabric is one example of how universities can bring information from operational systems and existing data silos into a more unified data environment, creating a stronger foundation for reporting, analytics, and AI.
For readers who want to understand that technology relationship in more detail, Crimson’s Microsoft Fabric vs Power BI: What's the Difference? explores how the two technologies work together and why the data sitting behind reports is just as important as the reporting technology itself.
Technology is only part of the answer, however, because data quality, ownership and governance still need attention before universities can place confidence in AI-generated recommendations.
What are the biggest risks of AI in higher education?
The risks associated with AI in higher education will vary considerably depending on the use case. Universities need to consider issues including academic integrity, bias and fairness, data privacy, inaccurate outputs, unequal access and over-reliance on AI, particularly where the technology could influence decisions that affect individual students.
Three principles are particularly important when universities consider how to manage those risks.
Keep humans accountable
Where AI contributes to decisions that materially affect students, human oversight should remain central. This is particularly important in areas such as admissions and student interventions, where AI can help bring information together or highlight potential issues, but an appropriate member of staff should remain responsible for reviewing the circumstances and making the decision.
Protect institutional information
Universities also need to understand where information is being processed and how AI tools interact with institutional data. Sensitive information should remain within appropriately managed environments, with clear governance around the data being used and controls over how it is accessed.
Establish clear guardrails
AI governance in a university needs to consider two distinct but connected areas: how staff and institutional teams use AI to support university processes, and how students use AI within learning and research. Both require clear policies, appropriate controls and transparency around responsible use.
The objective is not to eliminate every possible risk before experimentation begins. Universities need to assess the risk associated with each use case and implement controls that reflect the potential consequences, particularly where decisions could significantly affect individual students.
What should universities prioritise when developing an AI strategy?
With so many possible AI use cases in higher education, prioritisation becomes essential and should begin with the university's business priorities and desired outcomes.
Is the priority to increase applications, improve conversion, reduce administrative effort, strengthen student retention or improve the student experience?
Once leaders understand the outcome they want, they can assess whether the existing processes, technology and data can support it.
That approach also reflects themes explored in Crimson’s Digital Lounge episode, Unlocking AI's Potential in Higher Education: From Data to Impact, featuring Ian Tilsed, Director of Digital Technology and Operations at the University of Surrey. The episode explores where universities can realistically apply AI and the data, governance and organisational changes needed to make adoption stick.
Watch: Unlocking AI's Potential in Higher Education: From Data to Impact
The priorities need to extend beyond technology and consider:
- Clear institutional outcomes
- Trusted data and appropriate governance
- Prioritised use cases
- Human accountability for high-impact decisions
- Clear guidance for staff and students
- Skills and AI literacy
- Testing and evidence before wider implementation
The objective is not to deploy the greatest number of AI tools, but to identify where AI can create meaningful value without introducing risks that the institution cannot manage.
How can universities move from AI experimentation to practical value?
Universities do not need to identify every possible use case before starting, but they do need a clear understanding of where the institution currently sits and what it is trying to achieve.
Tom recommends examining institutional priorities and talking to peers about what they are seeing across the sector. External expertise can add practical experience, technical understanding and lessons from both higher education and other industries.
A sensible progression is to identify opportunities, evaluate readiness and test clearly defined use cases where outcomes can be measured. This discipline matters because enthusiasm around AI can otherwise make it tempting to start with the technology and work backwards.
The question university leaders should ask is not "Where can we put AI?" but “Which outcomes matter most, and where can AI realistically help us improve them?"
AI in higher education: focus on outcomes, evidence and trust
There is no shortage of potential applications for AI in higher education, from more responsive admissions and efficient handling of routine enquiries to student-support interventions, research acceleration and new approaches to programme design.
None of those opportunities remove the need for judgement, and the higher the potential impact on an individual student, the stronger the case for human accountability, clear governance and careful evaluation.
Universities need to balance ambition with pragmatism by starting with the institutional outcome, understanding the data and processes behind it, identifying where AI can add value and establishing appropriate guardrails before testing and measuring the use case.
Ultimately, AI should help universities make better decisions and deliver better experiences, rather than simply enabling them to make decisions faster.
Turn your university's AI priorities into a practical roadmap
Understanding the potential of AI is one thing. Deciding which opportunities are right for your university, where to focus first and what needs to be in place to deliver them successfully requires a more structured approach.
Crimson’s AI in Higher Education Art of the Possible workshop brings university leaders together with Crimson’s technology specialists to explore how Microsoft AI could support your institution's strategic objectives. The session examines practical use cases across university operations and identifies both immediate opportunities and areas that could form part of a longer-term AI roadmap.
The engagement includes a half-day workshop and playback, followed by an Art of the Possible assessment report that captures the identified opportunities and provides a strategic roadmap to inform your AI business case.