There’s a lot of noise around AI in higher education right now. New tools are being announced every week, and the pressure to invest, to deploy, to keep up, is real. But beneath the headlines, a more important question is being asked by the institutions we work with: does any of this actually improve conversion?
At UniQuest, part of Keystone Education Group, we’ve spent the last decade building the expertise, the processes, and now the technology to answer that question properly. Last week, our Director of Digital Transformation and Innovation, Jakub Blackman, shared his thinking in a live webinar. What follows is a summary of the key ideas, and why we think the sector needs to hear them.
The AI landscape: useful, but not neutral
Before we talk about what AI can do for student recruitment, it’s worth being clear-eyed about what’s happening in the wider AI market, because it affects every institution making technology decisions right now.
The large language models most people use daily, including ChatGPT and others, are increasingly under commercial pressure. Models that once felt objective are beginning to show the influence of advertising and commercial weighting. As Jakub explained, it’s not difficult to manipulate how a model responds. Research has already demonstrated how artificially amplifying a single concept can skew a model’s outputs across every kind of query. As commercial models chase revenue, that kind of influence is likely to increase, not decrease.
The implication for universities is straightforward: if your AI-powered tools are built on off-the-shelf commercial models, you have limited visibility into what’s actually shaping the responses students receive, or the recommendations your team relies on. That’s a risk worth taking seriously.
The other shift worth understanding is in how students research universities. Search is giving way to conversation. Students increasingly turn to AI tools to ask questions, compare programmes, and make decisions, and the responses they receive feel authoritative, even when they’re not. Institutions that don’t think about how their information appears in these AI-driven environments will increasingly lose ground to those that do.
Where AI investment in HE is falling short
Higher education has been quick to adopt tools that look like AI, and slower to ask whether they’re delivering real outcomes. A few patterns come up again and again.
A lot of what’s being sold as AI is really automated workflow: scripted responses, decision trees, rules-based routing. These tools have their place, but they shouldn’t be confused with machine learning or genuine intelligence.
Student-facing chatbots are the most visible example. Done well, they can genuinely help prospective students get answers quickly and feel supported. Done poorly, which is too often the case, they damage trust and leave students more confused than before they asked. The difference usually comes down to data: whether the chatbot is pulling from a structured, current, institution-specific knowledge base, or whether it’s been given a static document and left to its own devices.
Then there’s the data problem that sits upstream of all of this. Machine learning, and particularly propensity modelling, only works if it’s trained on data that is rich, structured, and connected across the full student journey. That’s hard for a single institution to achieve. Most universities hold their data in silos, across different systems, with inconsistent structures and incomplete capture across channels. Without solving that foundation problem, sophisticated AI capabilities simply aren’t available to you.
The UniQuest model: four elements that work together
At UniQuest, we think about the student engagement journey in four connected components. Each one matters, and each one depends on the others.
The human adviser remains at the centre
This is worth saying plainly, because it’s easy to lose in discussions about AI capability. Student recruitment is still built on human relationships. Students want to feel heard. They want fast responses, yes, but they also want to feel that someone is genuinely invested in their decision. That’s something AI can support, but not replicate.
As Jennifer Parsons put it during the webinar Q&A: the goal isn’t to take humans out of the process. It’s to make sure they’re used where they matter most. The conversations that build trust. The moments where empathy and judgement actually influence a student’s choice.
AI handles the volume, the classification, the triage, the pattern-matching across millions of data points. Advisers handle the relationship. That’s the model we’re building toward, and it’s one we’ve seen deliver results across our partner institutions.
Why scale matters for data
One thing Jakub was clear about: the data advantage UniQuest has is not something any single institution can replicate. Over a decade, working with 85+ universities across the globe, we’ve handled more than 500,000 student enrolments, delivered over 70 million student communications, and now manage more than 60,000 daily interactions. We analyse over 150 million data records every week.
That scale is what makes our propensity engine work. It’s trained not just on one institution’s intake history, but on patterns across the entire partner set. It knows what worked last year and what the data says about this year’s cohort, shaped by a world that’s changed significantly between the two.
As Jakub noted, propensity scoring isn’t a new concept. It underpins credit decisions, mortgage applications, insurance. What’s new is applying it rigorously to the international student recruitment context, with the data depth to make it genuinely predictive. That’s what we can now offer our partners.
What to look for when evaluating AI tools
If you’re thinking about where to invest in AI for student recruitment, a few questions are worth asking of any provider.
-
What is the underlying model, and how much visibility do you have into how it’s been trained and weighted? Off-the-shelf commercial models carry real risks around objectivity and commercial influence that aren’t always obvious.
-
How is the knowledge base being maintained? A chatbot or engagement tool is only as good as the data behind it. If that data isn’t structured, current, and institution-specific, the experience it delivers won’t be either.
-
Is conversion actually the focus? There’s a difference between tools that look impressive and tools that are designed, from the ground up, to support students through to a decision. The former are easy to build and easy to sell. The latter require a much deeper investment in data, in process, and in understanding what actually drives enrolment decisions.
-
And finally: where are the humans? Any AI approach that removes your advisers from the process entirely should be treated with scepticism. The technology should amplify what your team can do, not replace the part of the process where judgement and relationship-building still matter most.
A note on where this is going
Jakub was asked during the webinar about artificial general intelligence and what it might mean for services like UniQuest. His answer was honest: it’s genuinely unclear when, or whether, we reach that point, and it’s a long way from the large language models we see today. The more useful question is what the technology in front of us can do, right now, if it’s applied thoughtfully.
There will likely be some kind of correction in the AI market. The patterns of overvaluation, circular investment, and revenue gaps that Jakub described are real. But as he pointed out, the dot-com crash didn’t destroy the internet. It built the infrastructure that made everything else possible. The underlying technology is genuinely useful. The question is whether it’s being applied in ways that hold up to scrutiny.
At UniQuest, we think it is. And we think the institutions that approach this carefully, grounded in data, with their advisers at the centre, will be the ones that earn students’ trust, and convert them.
To find out more about the UniQuest platform and how we work with partner universities, get in touch with our team.
