The application of AI in university admissions needs oversight, not just optimism

Everyone is talking about AI (and sometimes automation dressed up as AI) and this in itself isn’t unsurprising. With any new revolutionary technical capability, it excites us to consider the potential applications, use-cases and benefits.

Whilst AI adoption and application has moved quickly in many industries which are both complex and highly regulated, reshaping businesses and workforces in the process, this hasn’t been the case for university admissions – at least not at a ‘revolutionary’ scale or speed to date. And, there is good reason for this.

AI and the technical capabilities it brings if applied correctly has enormous potential. But, only within scope of what AI is good at and the quality of any data set that any underlying LLMs are trained on.

Historical admissions data is not the starting point – at least not for ‘interpretative’ decision making

For any senior leader managing admissions functions, the cornerstone of successful delivery in any given cycle is the regular review of data, data, data and then some more data. However, it’s not the data that delivers successful outcomes. It’s the business acumen, interpretation and understanding the operating and regulatory context within each cycle that informs the necessary decision-making to deliver successful outcomes for the institution. If it was as simple as ‘data’ alone, and that predictable, every institution would meet it’s enrolment targets each cycle – the reality, this task in practice feels more like landing a jumbo-jet on a postage stamp.

It’s for this reason that institutions’ historical admissions data cannot be the starting point for any application of AI. No matter how much admissions data an institution has, it carries no context of the rules, regulations, policies or requirements at the time of admission and these cannot be inferred through data points or patterns alone. The value in an institution’s historical data lies in the potential to train an LLM against appropriate (current) rules, whilst validating outcomes & accuracy.

Potential application and benefits, but start afresh and not without a controlled rule-based engine!

AI works best with structured data, but it needs a controlled rule-based engine to mitigate risks, execute what is desired with accuracy and handle exceptions appropriately.

AI capabilities have the potential to deliver huge efficiencies for admissions professionals, if these are deployed in a discrete and purposeful way, with robust controls, rules and audit.

Intelligent document processing, data conversion, computation & calculation with interpretation against a controlled rule-based approach all have the potential to reduce non-value-add activity in admissions. But, checks and balances are essential. The potential here is an AI automated review with a transcribed audit for an admissions professional, who ultimately remains the ‘decision maker’ on that process (or part of).

Assessment of eligibility, fee status, scholarship eligibility, applicant profiling (widening participation or disadvantage), fraud detection (documents) and interview proctoring are all potential areas of application (to name but a few), providing they are controlled and managed in a robust manner.

Flexibility is paramount, without which the stakes are too high

The application of AI within the admissions process, even for discrete processes against controlled rule bases, need to remain flexible and easily controlled. Otherwise, the risks exponentially increase. Particularly, if you have relied on efficiency gains for a period of time, adjusted your resourcing accordingly and then find yourself in a position of having to operate without those gains.

Many of my counterparts across the sector will have experienced this at certain points in their careers (prior to AI), but where automations or integrations fail, or institutions implement new systems that do not deliver prior functionality once relied upon without a comprehensive resource impact assessment. Same impact and risks, just new drivers in technological terms.

This might seem like a contradiction; flexibility vs defined rules, and on the one hand it is. But, both need to co-exist. Which is to say that a responsible and successful deployment of AI in admissions will always require a controlled rule base, that multiple rule bases for the same process(es) will need to be able to coexist and operate accurately and that all of these will need to be easy to maintain efficiently. Why? Because, the applicant journey isn’t linear and the rules, policies and regulations that govern admissions and compliance aren’t static either.

In any given admissions cycle, teams are managing admissions to multiple intakes, years or points of entry concurrently, that can operate different; entry requirements, international equivalencies, immigration requirements, fee status regulations, academic portfolios, fee rates, and the list goes on. On top of that, applicants will apply and then may make a host of changes, amending entry backward or forward, or changing programme or entry level and this can occur at almost any point throughout their admission journey. Careful consideration & mapping of all of these potential scenarios in the application of AI, automation, workflow and audit are therefore essential.

You can’t buy a degree – which is why university admissions isn’t transactional

It isn’t the case that university admissions is necessarily more complex than other areas of highly regulated industries (e.g finance) that have made significant inroads into the adoption and deployment of AI, but what does differ is that university admission and student engagement isn’t transactional, and that means it is still inherently more nuanced. This is where the risks & limitations of deployment are situated and why unsurprisingly the EU AI Act regulates the use of AI in educational admission as high risk activity.

Even if we assumed that applicant behavior followed a linear process (which it doesn’t) we’re instantly presented with challenges and pitfalls - I’m noting two, but there are many!

• Meeting entry requirements doesn’t guarantee an offer. And, the more selective an institution is, the higher the burden of proof on holistic review, additional selective measures and a divergence from application-available ‘data points’ in decision making. It’s simply not that ‘simple’ and it’s for that reason that so much consideration and careful curation is put into establishing fair & effective admissions policies, where exception and nuance is provisioned for in a manner that doesn’t contradict the aims or compliance or fair admission.
• Admissions professionals assess and treat applicants on academic merit (with compliance requirements & regulations always assessed) and this is in line with institutional commitments to the UUK and GuildHE fair admissions principles. But, external policy changes can drive differential treatment of individuals, eroding and compromising institutions’ ability to meet these values and principles in full. This creates an increasingly nuanced landscape of process and communication for institutions to navigate, both in terms of the systems that they use to support the process and the stakeholders who interact with these; agents, schools, partners and prospective applicants.

Transparency avoids hypocrisy because trust cannot be automated

Those with responsibility and accountability for admissions, access, compliance and participation work tirelessly to deliver the best outcomes, not only for the individuals they support (prospective students), but their institutions alike.

When an individual applies to a university, it’s a personal endeavor. It is an intention with a serious financial & personal commitment. As recipients of those applications, institutions look to consider those applicants fairly, but regulatory shifts that drive divergent consideration and processes continue to impact the endeavour of fair admissions.

Authenticity, genuineness & credibility, whilst not new concepts are increasingly becoming a focus for governments with respect to international student mobility, broadening attention and administrative burden in the process.

As the process of university admissions becomes increasingly burdensome and high stakes for both applicants and institutions alike, this creates the conditions to increase motivations to utilise AI for both parties also. However, applicants’ use of AI in the admissions process is still largely undesired by institutions and this is unlikely to change any time soon. So institutions adopting AI in their admissions process will need to be incredibly transparent on their use and application of AI to avoid establishing hypocritical standards.

Trust is built and cannot be automated. For many applicants, admissions is their first real interaction with a university and it sets the tone for the entire student journey and experience with the institution. While automation can improve speed and consistency, it can also influence how personal the experience feels. If every interaction seems too technical or system-driven, it can create distance at the very moment institutions are trying to build connection in an increasingly competitive recruitment landscape.

Admissions decisions shape not just institutional reputation but also long-term academic communities and student outcomes.

What does the future look like?

AI is now firmly embedded in day-to-day life for many and it will continue to develop and shape industries and workforces alike.

But university admissions is a trust-based function. Applicants, agents, partners, schools and regulators expect decisions to be fair, explainable, and grounded in judgement.

Applied appropriately, AI can help institutions manage complex, high-volume and standardised tasks whilst reducing error. But scale alone shouldn’t be the starting point for adoption, as a homogenous cohort is not the goal of any institution, nor would it deliver the desired student experience.

Systems may generate recommendations, but institutions must remain clearly accountable for the outcomes.

The real question is not whether AI should play a role in admissions. It is to what extent universities are currently prepared to adopt it, govern it thoughtfully, explain it clearly, and mitigate the risks when they arise. The credibility of AI-assisted admissions will depend less on technical capability and more on transparency, with a continued commitment to the value of fairness at the core of our profession.

About the author

David Parrott is Director of Admissions & Conversion Services at UniQuest, with over 20 years of admissions experience gained across UniQuest, Imperial College London and the University of Roehampton.

His work focuses on leveraging the latest technology coupled with the best application of people to meet pressing admissions demands, whilst enhancing the applicant experience and developing new solutions that ensure world-class support for our partners whilst supporting them in fulfilling their strategic aims.