There is no shortage of interest in AI across higher education. Strategy documents reference it, conferences discuss it and senior leaders talk about transformation.
What is harder to find is evidence of it working at the point where it would make the most difference.
The gap between stated ambition and practical implementation is wide, and it is not difficult to understand why. Procurement processes were not designed for tools that move as quickly as AI does.
IT governance frameworks require sign-off at multiple levels. Legal and compliance teams have legitimate concerns about data, liability and accuracy. By the time an institution has worked through all of that, the landscape has shifted again.
The result is that AI investment in higher education tends to concentrate in the areas where it is easiest to deploy rather than where it is most needed. Back-office automation, content generation, administrative tools. These are useful, but they are not transformative.
The area that has seen least progress is also the one with the most direct commercial consequence: the point at which a prospective student first makes contact.
Most institutions still handle that moment with a contact form, a delayed email response, or a chatbot that has not moved on much from the dropdown logic of five years ago. That is the digital front door, and for a sector competing internationally for students who are comparing options across multiple countries at once, it is not good enough.
The barriers to fixing it are real - teams are stretched, integration with existing CRM systems is not straightforward, and there is a reasonable institutional caution about AI that might get things wrong in front of a prospective student.
Those concerns are worth taking seriously, but they have also become a reason not to act, and the cost of not acting is visible in the data. Research from ICEF Monitor found that nearly six in ten prospective international students disengage when communication feels slow or unclear. That disengagement happens before application, before conversion is tracked, and before anyone notices.
Voice AI has reached a point where it can handle real conversations, in any language, at any hour, without sounding like a robotic system. The deployment barrier is lower than most institutions assume. The risk of getting it wrong is manageable if the tool is drawing on the institution's own verified content rather than generating freely.
The ambition is there, what is needed now is the willingness to close the gap between the strategy document and the website.
By John Crick
