Editorial: Open the university door wider, raise the bar inside

British and Irish universities spend an extraordinary amount of effort deciding who gets in, and far less asking what students can actually do when they leave. The order is the wrong way round. Young people pay for it in waiting and in worry, and employers pay for it when a degree certificate turns out to say less than it promised.

A Finnish polytechnic lecturer recently said this out loud. Kimmo Haanpää teaches electrical power engineering at Metropolia University of Applied Sciences in Helsinki, and in a Kuuntelija podcast interview (in Finnish) he argues that the system mostly produces diplomas. In his view students arrive less prepared than they used to, standards have drifted, and funding rewards the number of graduates more than what those graduates know. That is one teacher’s observation, not a research finding, but it deserves a hearing because he is the person who signs off the grades. The conversation is about Finland, yet anyone who has sat on a UK or Irish admissions panel will recognise the pattern.

▶

Source: Kuuntelija podcast, interview with polytechnic lecturer Kimmo Haanpää, 27 September 2026 (in Finnish)

Start with how the gate works in England, Wales and Northern Ireland. Applicants name up to five courses through UCAS, and most receive conditional offers. As UCAS explains, a conditional offer means you still have to meet the requirements, usually exam results. Those offers are made months before the exams are sat, on the basis of grades that teachers predict. Predictions are a guess about the future, and guesses are often wrong. A 2013 research paper for the Department for Business, Innovation and Skills found that the share of accurate predictions fell from 52 per cent to 42 per cent when the A* grade was introduced in 2010, and that 55 per cent of predicted A* grades were over-predictions. The system is older than that study and the rules have moved on, but the principle stands: a great deal of the selection happens before anybody has shown what they can do at university.

Those who miss out scramble through Clearing or wait a year. Gap years can be wonderful when they are chosen. A year spent waiting for a place, or waiting to resit an exam to hit a threshold, is different: it does not build skills, it wears them down. The better arrangement is to turn the order around, with a wider door and tougher standards inside.

Estonia has tried a version of this. Entry there rests largely on national exams and school-leaving results, although some fields run their own tests or interviews, as ERR has reported. The decisive change was the higher education reform of 2013. Since then, studying in Estonian has been free for full-time students who complete their curriculum at the required pace, and students who fall behind may have to pay back credits, as the EU’s Eurydice database describes. Selection moved from the gate to the lecture hall, and the rule is simple: complete or pay.

The results are encouraging but not miraculous. According to Estonia’s education minister, the share of students dropping out fell from 15.6 to 13.6 per cent between 2013 and 2018, while the share graduating within the nominal time rose from 43.2 to 52.7 per cent between 2016 and 2019. Estonia’s National Audit Office, Riigikontroll, judged in 2019 that the results of the reform were patchy. So the Estonian model is not a finished answer. Its principle, however, holds up: a place at university is earned by progress, not by queueing. Britain and Ireland could admit more students and let selection happen during the first year, when the measure is real progress rather than one exam day or a teacher’s forecast.

There are fair objections. A first-year cut can be harsh on a student who works, cares for a relative or needs longer to settle, so the rules must measure progress rather than circumstances, and a part-time route must stay open. Cost is the other worry, since more students need more teaching. That is true, but the current arrangement already pays for it in non-completion. In England, the Office for Students reports a completion rate of 90.3 per cent for young full-time first-degree entrants in 2019-20, yet the gap for mature entrants was 11.3 percentage points lower. Money is spent either way. The question is whether it goes on waiting or on learning.

A tougher first year is fair only if the equipment works. Universities should give every student AI tools instead of banning them and hoping for the best. They also need a shared knowledge bank in which lectures, coursework, tutors’ feedback and earlier cohorts’ solutions accumulate, so that the AI can retrieve the right passage at the right moment. And students should earn extra credit for building new skills. Someone who builds a measuring tool for the lab, automates a repetitive calculation or teaches a language model the vocabulary of their field is doing precisely the work that employers will soon expect. The debate around what ChatGPT can really do in a daily workflow shows the same thing: producing output gets cheaper, while understanding the problem does not.

For the student, this means personal-trainer-level teaching. One-to-one coaching has always been a privilege of the few, because a tutor cannot give thirty students undivided attention at once. When AI knows each student’s mistakes, progress and blind spots, coaching scales. The teacher is freed to do what only a human can: demand, encourage and say plainly when the work is not yet good enough. The same logic works in business, as the case for rethinking the process before adding AI agents argues: the machine carries the repetition, the person owns the direction.

Assessment should go back to paper, and that is progress rather than nostalgia. When a language model can write an essay in seconds, coursework no longer shows what a student knows. A proctored paper exam does. It shows what is in the student’s own head when no tool is within reach. Only then does AI step in and analyse the answers: where the whole group’s understanding breaks down, which mistake recurs year after year, who needs extra practice before falling too far behind. The teacher gets a diagnosis, the student gets feedback, and the course improves next time. Real ability is measured without the machine, and learning is improved with it.

The boldest part of the idea concerns graduation. A leaver would take away, on a one-terabyte hard drive, the shared understanding of the whole school: course materials in their field, solved problems, tutors’ explanations, their own notes, and an assistant tuned on top of all of it. Copyright, data protection and keeping the material current must be handled carefully, but those are solvable problems, not reasons to do nothing. This starter pack would reshuffle who gets the best seats in working life. Employers would no longer hire a certificate but a person who has passed a real filter and arrives with a library they know how to work with from day one.

The proposal asks more of universities than the present model does. Funding should reward competence, not only completed degrees; staff time should shift from routine to coaching; and students must accept that the first year is a test rather than a prize. The price is lower than the price of the alternative, though. The same worry about how well technology is absorbed runs through our piece on million-token context windows and AI agents, and it is a reason for universities to act now instead of waiting.

Teachers like Haanpää have done the hardest part by saying that the emperor has no clothes. Now it is the turn of universities, ministers and employers to look in the mirror. Higher education does not have to choose between quantity and quality. It has to stop measuring in the wrong place.