UNIwise Blog

Nobody is at level three: assessment readiness rethought

Rasmus Blok — Mon, 10 Aug 2026 07:30:24 GMT

Somewhere in your institution there is a slide with a number on it. Level two, perhaps, or “Intentional”, or an amber dot two-thirds of the way along a five-stage arrow. It was produced by a serious exercise, by serious people, and it is almost certainly wrong - not because the underlying work was bad, but because the question it answers is one no institution actually faces.

Higher education has spent 2026 measuring itself. In April, Times Higher Education published its AI & Digital Maturity Index, built on 4,950 responses from 1,353 institutions across 82 countries and sorting them into four levels: Incidental, Intentional, Integrated, Optimised. In May, Jisc opened a free digital maturity assessment to UK institutions - 24 questions drawn from a wider transformation framework - with anonymised sector benchmarks due this September. Vendors have published their own maturity models alongside them, each ending, reasonably enough, with an invitation to talk.

This is not a bad development. The sector needs a shared language for a genuinely hard problem, and these instruments give it one. But the ladder metaphor that almost all of them share carries an assumption that does not survive contact with assessment, and it is worth saying plainly what that assumption is: that an institution has a level. That readiness is a single property, held by the organisation as a whole, that moves in one direction.

It isn't, and it doesn't.

What one number hides

Take a pattern of the kind that turns up constantly. An institution runs coursework submission and feedback through its VLE with genuinely high adoption - nearly every module, nearly every academic, a decade of habit behind it. Integration to the student record system is solid. Accessibility has had proper investment because someone senior made it a priority in 2021. On any general digital maturity instrument, this institution scores well.

The same institution resolves second-marking disagreements by email. Its integrity policy is applied differently in three faculties because nobody has ever reconciled them. It has no consistent way of showing, after the fact, how a contested mark was arrived at. In the domains where assessment risk actually concentrates, it is nowhere near where its overall score suggests.

Now give that institution a level. Whatever number you choose, it is an average of things that should never have been averaged - and the harm is not that the number is imprecise. The harm is that averages hide exactly the cells that matter, because a strong cell and a weak cell cancel. The institution above will read “we're doing well” and invest in the areas it already leads, because those are the ones with internal champions and existing momentum. The gap that will end up in front of an appeals panel stays invisible, protected by the good score sitting next to it.

There is a second failure mode, quieter and more damaging. A ladder implies an end state, and end states become targets. Once “level three” exists as an institutional goal, work reorganises around producing the evidence of level three - the documented process, the policy on the intranet, the dashboard - rather than around the outcome the level was supposed to indicate. This is not cynicism about universities; it is the oldest finding in performance measurement, and assessment is unusually exposed to it, because so much of what makes an assessment defensible is invisible in a process document. You cannot tell, from a policy, whether two markers who disagreed actually reconciled their disagreement or simply split the difference.

Readiness is not a property of an institution

Here is the sharper version of the problem. Readiness is not a property of an institution at all. It is a property of a context.

A university might be entirely ready for formative, low-stakes assessment - fast feedback loops, generous accommodations, relaxed controls, exactly as it should be - and not ready at all for a high-stakes regulated exam where a contested outcome ends up with a professional body. Those are different capability requirements, different proportionality judgements and different evidence obligations. A single institutional grade cannot distinguish them, which means it cannot help with the only question leadership actually has: given what we assess, and what happens when we get it wrong, what should we fix first?

That question has four contexts hiding inside it, and they behave differently: formative and low-stakes work; coursework-based summative assessment; exams in controlled conditions; and high-stakes, regulated or accreditation-critical assessment. The control set that is proportionate in the fourth is oppressive in the first. An institution that applies its high-stakes posture everywhere has not become more mature; it has become worse at teaching. An institution that applies its formative posture everywhere has a problem it has not met yet.

Assessment readiness, then, is not a position on a line. It is a pattern.

The case for the ladder, which is better than it looks

The honest objection to everything above is that a pattern is harder to use than a number, and that using things is the point.

A level gets an item onto a board agenda in a way that seven domain scores do not. It permits comparison with peers, which is sometimes the only argument that releases budget. It gives a change programme a destination, and change programmes without destinations tend to become permanent. And there is a real risk that a nuanced diagnostic simply produces seven amber cells and a shrug - a more accurate picture of a situation nobody now knows how to act on. Anyone who has watched a maturity exercise land in a university knows that the single number is often doing more political work than analytical work, and that this is not entirely illegitimate.

So the answer is not to refuse to score. It is to score at the level where the score is meaningful, and stop there.

Score each capability domain, separately, on a scale short enough to be honest: 0 for ad hoc, 1 for baseline stable, 2 for scaled and documented, 3 for adaptive and continuously improved. Publish those scores as a heatmap, and do not roll them up. A heatmap is not a refusal to be assessed - it is a more demanding assessment, because it removes the place where weak areas hide. What it gives leadership is not a grade but a priority, and a priority is a more useful thing to take to a board than a grade is, because it comes with an obvious next sentence.

That is the design principle behind the UNIwise Assessment Trust & Readiness Model, and it is the reason the model deliberately produces no overall label.

Seven domains, and why these seven

The model scores readiness across seven capability domains: Integrity and Authenticity; Feedback and Learning Impact; Operational Excellence; Security and Assurance; Accessibility and Equity; Interoperability and Data; and Governance and Change Capability.

They are chosen so that the ones institutions tend to be good at and the ones they tend to be weak at cannot cancel each other out. Interoperability and Data is usually the strongest domain in the set, because integration work has had budget and attention for fifteen years. Governance and Change Capability is usually the weakest, because it has no natural owner and no obvious artefact - although the EU AI Act's 2027 obligations are, for once, giving it both. In a rolled-up score, the first hides the second. Side by side, the pattern is the finding: an institution with excellent plumbing and no decision rights is not immature, it is specifically and fixably stuck, and the fix is not technical.

Two of the domains are worth a word on framing. Security and Assurance is scored on whether controls match the stakes - over-control counts against you, as it should, because blanket surveillance applied to a formative quiz is a governance failure and not a safety margin. Proportionate control is layered control, and a readiness model that rewarded maximum control would be rewarding the wrong instinct. And Accessibility and Equity is a domain rather than a compliance footnote, because an assessment some students cannot sit properly is not a defensible assessment, whatever its audit trail says.

The result a readiness model must never assume

There is one more assumption buried in most maturity frameworks, and it is the one with the clearest commercial fingerprints on it: that the route upward runs through consolidation. Standardise the platform, unify the ecosystem, retire the local tools, and maturity follows.

Sometimes that is right. Often it is not, and the exercise that concludes it should always be treated with suspicion - including when the exercise is ours. If an institution's VLE is already the operational hub for coursework, with a decade of academic habit behind it, then the highest-value, lowest-risk improvement is almost always to extend capability where the work already happens: rubrics used as quality assurance rather than paperwork, a real moderation workflow, a marking record that survives a challenge - delivered into the VLE through LTI rather than through a procurement round. That is what Marking Studio is for. Where the risk profile genuinely demands more - controlled conditions, proportionate monitoring, defensible integrity checks - those capabilities layer on afterwards, by context, in the order the heatmap says.

If an institution's VLE adoption is thin or fragmented, the honest answer is different, and a readiness model that cannot say so is not a diagnostic. It is a brochure.

Fifteen minutes, and no verdict

Writing in HEPI this week about the cohort arriving in 2027, Roney Lima do Nascimento made an argument about students that applies just as well to institutions: diagnose on entry rather than assume. His point was that a department which can articulate what competence looks like in its discipline can teach it, while one that only knows what it wants to forbid cannot. The institutional version is the same. A university that can say precisely which of its assessment capabilities is weakest, and in which context, can fix it. One that knows only that it would like to be more mature cannot.

That is what the readiness check is for. It is live and free, at uniwise.eu/digital-assessment-readiness. It takes ten to fifteen minutes and runs in three steps - institutional context, then VLE and operations, then the seven capability domains - and it returns your heatmap on screen immediately. There is no single number at the end of it, and that is not a limitation of the tool.

What you get is a picture of where your readiness is uneven, which is the only shape institutional readiness ever actually has. What you do with it is a conversation, and a better one than “how do we get to level three” - because the answer to that question was never a level. It was a list, in order, of the specific things that would not hold up if someone asked.

External sources cited

Times Higher Education, AI & Digital Maturity Index 2026 (21 April 2026) - timeshighereducation.com
Jisc, Launching our digital maturity assessment to support transformation across UK higher education (7 May 2026) - jisc.ac.uk
HEPI, The cohort arriving in 2027 will not reason like the one in front of you. Are universities ready? (4 August 2026) - hepi.ac.uk

Onboarding is not an installation: why the right vendor matters as much as the right platform

Rasmus Blok — Wed, 08 Jul 2026 11:32:47 GMT

No institution buys an assessment platform because it wants more software. It buys a platform because it wants something to change: fairer and more secure exams, saner workflows in the exam office, richer feedback for students, less friction for academics. The software is the instrument. The change is the point.

That is why the most revealing question you can ask a vendor - any vendor, including us - is not “what can the platform do?” but “what will you do to make us succeed?” The two questions get very different answers, and in my experience the second one predicts the outcome of a procurement far better than the first.

THE COMFORTABLE FICTION OF THE STANDARD IMPLEMENTATION

A standard implementation assumes a standard institution. There is no such thing. Every institution arrives with its own regulations, its own assessment traditions, its own academic calendar, its own systems landscape, and its own history of change projects - some of them still remembered with a shudder.

Yet much of our industry still treats onboarding as a technical delivery: insert the proverbial disk, run the setup wizard, hand over the manual, close the project. The software gets installed. The institution does not get transformed. And when the playbook is identical for every customer, it is the customer who ends up bending to the tool, rather than the tool serving the institution.

The gap between those two outcomes - software installed versus institution succeeding - is exactly where onboarding lives. It is the same lesson we have drawn elsewhere on this blog: secure exams are not a feature you switch on, and neither is a successful implementation.

ONBOARDING IS A CHANGE PROCESS, NOT A TECHNICAL TASK

Look closely at what actually changes when an institution adopts digital assessment, and very little of it is technical. Exam officers build new routines for planning and running assessments. Academics change how they author, mark, and give feedback. IT integrates a new platform into the student information system and the wider architecture. Invigilators, administrators, and students all meet new ways of working - often in the most high-stakes moments of the academic year.

In other words: onboarding is the deliberate launching of new habits and practices, and the respectful retiring of old ones. It is also where the gains get harvested - better feedback quality, more consistent marking, cleaner data, faster processes. None of that arrives automatically with a login. It has to be worked for, together.

That is why good onboarding looks less like a deployment checklist and more like a supported change process: mapping stakeholders, aligning goals, training people in their actual roles, testing against real scenarios, and building confidence before the first live exam - not after it.

A VENDOR THAT ADAPTS TO YOUR WORLD - NOT THE OTHER WAY AROUND

So what should you look for in a vendor, beyond the feature list?

Look for a partner who starts by listening: who maps your existing processes before proposing new ones, and who adapts the plan to your regulations, your timeline, and your people. Look for genuine sector understanding - assessment in higher education is not a generic workflow, and a vendor who has it in their DNA will anticipate problems a generalist will not even recognise. And look for how success is measured: by their delivery milestones, or by your outcomes - adoption, confidence, and quality - long after go-live.

Structure still matters. A proven process is what makes adaptation safe rather than improvised. The distinction is between a vendor who is structured but rigid, and one who is structured and responsive. You want the latter.

THE LONG RUN: PARTNERSHIP, NOT PROJECT

Assessment is critical infrastructure. Once your exams run on a platform, you are not buying a product; you are entering a relationship that will span many years, many exam seasons, and many changes in both your institution and the technology around it - AI being only the most obvious example.

That changes what onboarding is for. It is not the final phase of a sales process; it is the first phase of a partnership, and it sets the tone for everything that follows. A vendor playing the long game invests disproportionately in those first months, because they know trust is built when the stakes are highest. They keep showing up afterwards too - with check-ins, webinars, product updates, and a genuine channel for your needs to shape the roadmap. And when the regulatory ground shifts, as it is doing right now with the EU AI Act, a real partner helps you get ahead of it rather than leaving you to read the legislation alone.

If a vendor treats go-live as the finish line, you have learned something important about the next ten years.

HOW WE ONBOARD AT UNIWISE

This philosophy is not abstract for us; it is codified in how we work. Every new institution is guided through six phases, each with a clear purpose.

It starts with Initiate, where a dedicated Implementation Consultant works with your team to align on goals, map existing processes, and set a clear project plan with the right stakeholders around the table - from IT to academic leadership. In Execute, we configure WISEflow to match the way you work, including integrations with your student information system and institution-specific workflows. Then comes Test & Train: structured testing in your environment and tailored training for administrators, academics, IT staff, and key users - because we do not go live until you are confident.

After launch, Hypercare gives you elevated, hands-on attention in the weeks when it matters most, with close monitoring and rapid response. From there, Support means a dedicated Customer Success team, professional support, and our comprehensive Service Centre. And finally, Continue to Improve: regular check-ins, webinars, and product update sessions, because onboarding is the beginning, not the end.

The approach works, and we can put numbers on it: a 98% onboarding success rate, more than 20 institutions successfully onboarded in 2025 alone, and an average time to go-live of under three months. Today 150 institutions across 14 countries work with us, and in 2025 WISEflow carried 4.1 million exams and assessments.

WHY SO MANY CUSTOMERS PRAISE OUR ONBOARDING

I could describe the effect of all this myself, but our customers say it better.

“The transition was intense and had to happen quickly, but UNIwise stayed very close to us throughout. The onboarding was thorough, and the continuous follow-up made us confident that we would succeed. This was not a standard delivery. It felt like a true collaboration.”

- Heidi Olsen, Examination administration, VID Specialized University

“For us, it’s a success story. The quality of the product was important, but the quality of the service was what really convinced us. We couldn’t have done it without UNIwise.”

- Jonas Debrulle, Director of Programmes, IÈSEG

“We dared to run large, high-stakes exams very early because we knew we had UNIwise’s support behind us. We knew they were available, we had their mobile numbers, and that made all the difference. It made us feel safe.”

- Lena Knudsen, Senior Advisor, Section for Examinations, Østfold University College

Notice what these quotes are about. Not features. They are about feeling seen, feeling safe, and being met as a partner rather than administered as an account. That - not the software alone - is what a successful onboarding produces.

THE POINT

Platform and partner are two halves of the same decision. The platform determines what is possible; the partner determines what actually happens. Onboarding is where you find out which vendor you chose - the one running a standard implementation, or the one committed to your success for the long run.

Choose the one who understands that onboarding is more than inserting the disk.

AI in Higher Education Assessment: Reliability Over Hype

Rasmus Blok — Tue, 30 Jun 2026 11:28:24 GMT

Generative AI has crossed a line that the sector is still catching up with. In barely two years it has gone from a fringe curiosity to ubiquitous infrastructure — available to every student, at near-zero cost, on demand. The comfortable assumption that a piece of written work represents a single person’s unaided thinking no longer holds by default. And unlike most technology shifts in higher education, this one does not allow for a multi-year policy cycle before institutions respond. The pressure is immediate.

That puts assessment, not teaching, at the centre of the AI conversation. Because if AI can reproduce the outputs long used to certify learning, then the credential itself is at risk. Employer trust, accreditation and public confidence all rest on one thing: the credibility of the assessment behind the qualification. This is the real story of AI in education — not productivity hacks or essay generators, but a structural test of whether a qualification still means what it claims to mean.

Here is how I see the next few years unfolding — three trends I think will define them — and why I believe the institutions that thrive will be the ones that choose reliability over spectacle, and demand the same of their technology partners.

The dual mandate: embrace and shield

Institutions face two imperatives at once, and they pull in opposite directions. On one side, they must embed AI literacy into the curriculum, because graduates are entering a workforce already restructured around these tools; pretending AI does not exist serves no one. On the other, they must protect the integrity of assessments that exist precisely to certify individual, unassisted capability.

These two mandates cannot be solved in separate places. They have to coexist within the same programmes, the same platforms and the same processes. That is uncomfortable, and it is why simplistic answers — “ban it” or “embrace it” — both fail. The honest position is that some assessment should welcome AI and some must deliberately exclude it, and an institution needs the infrastructure to do both, consistently, at scale.

Trend one: assessment becomes a system, not a format

For decades, assessment has often been defined by a single artefact — the essay, the multiple-choice test, the final exam. That model is breaking down. As confidence in unsupervised written work erodes, institutions are rethinking formats, and the response is converging on something more deliberate: multi-modal assessment.

Expect to see more controlled contexts where integrity is critical — on-campus and proctored digital exams. Expect the return of oral and hybrid formats, where a written submission is followed by an oral defence that validates authorship. Expect portfolios and continuous assessment that build evidence over time rather than betting everything on a single high-stakes moment.

The deeper shift is that these will stop being isolated events and start being designed as a coherent sequence — a system that intentionally combines open and controlled contexts to balance pedagogy, efficiency and integrity. The institutions that come out ahead will not be those clinging to a single assessment type. They will be the ones that treat assessment as a system they can design, control and evolve over time — and that choose the tools to support it accordingly.

Trend two: reliable beats impressive

It is easy to be impressed by AI right now. Demos are dazzling, and the temptation to bolt a clever feature onto an exam workflow is strong. But high-stakes assessment is unforgiving in a way that consumer software is not. A grading suggestion that is brilliant nine times out of ten and confidently wrong the tenth is not a 90% success — in an exam context it is a fairness, appeals and accreditation problem.

For AI in education, reliability matters more than brilliance. A feature that behaves predictably, explains itself, leaves an audit trail and fails safely is worth more than one that is occasionally spectacular and never quite accountable. This is the standard UNIwise holds itself to: AI that assists institutional judgement rather than replacing it, with a human in the loop and a clear line of accountability at every step.

Trend three: “vibes” do not scale to high stakes

A lot of early AI development runs on intuition. A model “feels” right, the outputs “look” good, and that is enough to ship in a low-stakes setting. That approach has its place — but it does not survive contact with summative assessment.

When the output determines whether someone qualifies as a nurse, an engineer or a lawyer, “it usually works” is not a standard anyone can defend to an external examiner, a regulator or a student lodging an appeal. High-stakes use cases demand something that intuition-led development cannot provide: measurable consistency, transparency about how a conclusion was reached, defensibility under scrutiny, and behaviour that holds up across millions of cases rather than a handful of demos. The gap between a promising prototype and a system you can stake a degree on is exactly the gap the sector now has to close — and it is far wider than the hype suggests.

Why UNIwise treats assessment as core infrastructure

This is the principle that shapes everything UNIwise builds. However an institution chooses to adopt it — as the full WISEflow platform or as a focused capability bundle such as Marking Studio — assessment is never treated as a cosmetic add-on to a learning platform. It is treated as core infrastructure: the load-bearing system that institutions, students and employers depend on, and that has to keep working under pressure, at volume, with full accountability.

Treating assessment as infrastructure has practical consequences. It means full audit trails, not opaque automation. It means EU-based data processing and a clear point of accountability, rather than fragmented tools that each add risk. It means supporting the full range of formats — written, oral, portfolio, proctored — within one trusted environment, so that as an institution’s approach evolves, the foundation does not have to be ripped out and replaced. And it means institutions adopt at their own pace, starting where the pressure is greatest, always retaining academic ownership and the initiative.

It is why WISEflow is used by more than 150 higher education institutions across 14-plus countries, with millions of users. Not because it is the flashiest tool in the room, but because assessment at that scale has to be dependable.

The bottom line

AI is not the product. Trust, integrity and control — amplified by AI — are the product. The institutions that come through this transition strongest will be the ones that design assessment as a system, insist on reliability over novelty, and refuse to let high-stakes decisions rest on intuition. That is the future UNIwise is building for, and in my view it is the standard the whole sector should expect of its technology partners.

If you want to see what AI-assisted, infrastructure-grade assessment could look like at your institution, UNIwise would be glad to talk.

Assist, don't dictate: where AI belongs in assessment - and where it doesn't

Rasmus Blok — Wed, 24 Jun 2026 09:42:04 GMT

There is a line running through every conversation about AI in higher education, and most of the disagreement comes from not naming it: is the AI assisting the human, or is it driving?

It is an easy distinction to blur, because the same technology can do both. The same model that helps a marker sharpen their feedback can, with a different switch flipped, be asked to produce the grade itself. The interface looks identical. The implications could not be more different.

I want to work through the general question first - because it is the general question that should drive the answer - and only then explain the choice we have made at UNIwise, and why.

ASSIST OR DICTATE - THE DISTINCTION THAT MATTERS

Assisting means the AI supports a person who remains in control. It drafts, suggests, checks, flags, and structures - but a human reads it, judges it, and owns the decision. The accountability stays where it has always been.

Dictating means the AI makes the call. It scores the script, ranks the candidate, decides the outcome - and the human, if they are still in the loop at all, is reduced to rubber-stamping a result they did not reach themselves.

Both are technically possible today. The interesting question is not “can we?” but “should we, here, now?” And the honest answer is that it depends entirely on where in the assessment process you are standing. Supporting an academic’s workflow is one thing. Letting a model determine a student’s result is another thing altogether - and we should not pretend they sit on the same footing.

MARKING IS WHERE IT GETS HARDEST

Nowhere is this clearer than in marking. This is the part of assessment where the stakes for the individual student are highest, where fairness is most scrutinised, and where the consequences of getting it wrong are hardest to undo.

It is also where the pressure on markers is most acute. We know markers want help, and for good reason: they face large volumes of scripts to assess in a short window, often on top of teaching, supervision and research. That is a genuinely hard set of conditions, and it is precisely why so many have started reaching for whatever tool is closest to hand - increasingly a general-purpose chatbot such as ChatGPT. The instinct is human and the need is real.

The problem is what happens next. Pasting student work into a general-purpose chatbot often sends personal data outside the EU, hands it to systems that may train on it, and quietly compromises both the privacy and the intellectual property of the students whose work it is. The student never agreed to that. The institution rarely knows it is happening. It is a data-protection and IP problem wearing the costume of a productivity gain - which is why the answer cannot simply be “stop using AI”. The need is real, so the responsible move is to give markers help they can actually trust.

WHY LETTING AI DRIVE IS PREMATURE

If the need is real, why not let the AI go further and take on the marking itself? Because we are still early, and several things have to mature first.

There is genuine and reasonable doubt across the sector, limited experience to draw on, and trust that has not yet been earned. On top of that sit real legal uncertainties: it is still unclear where a student stands, in law, when a grade has been produced with AI involvement — their right to a fair, explainable and contestable assessment is not yet settled ground. That uncertainty alone makes AI-led grading immature today.

The regulatory picture points the same way. Some providers offer systems that participate in or influence the assessment of students without meeting - or while actively side-stepping - the obligations the EU AI Act places on high-risk AI in education: transparency, human oversight, logging, and impact assessment. A tool that helps decide a student’s result is not a low-stakes convenience, and a tool that ignores those duties is a liability an institution inherits the moment it switches it on.

There is a design dimension too. Even where AI only assists, it has to be built to work in closed circles - with access to the relevant assessment criteria, the assignment and the student’s own submission, and nothing beyond. Kept inside that boundary, it stays anchored to the task. Allowed to roam, it starts borrowing from other sources and becomes far more prone to hallucinate. Without that discipline, “assistance” quietly becomes another integrity risk rather than a safeguard against one.

THE UPSIDE IS REAL - AND IT WILL GROW

None of this is an argument against AI in assessment. It would be a mistake to read caution as reluctance.

Used as an assistant, AI offers real and substantial gains - and feedback is where they show up most clearly today. It can help ensure consistency, so that the first script of the day and the last are held to the same standard. It can raise the quality and depth of feedback, giving students more than a mark and a sparse comment. It can help guard against the bias and fatigue that affect every human marker. And it can anchor feedback explicitly to the assessment criteria, turning it into something a student can actually learn from rather than a verdict they simply receive.

And the role will grow. Humans are not perfect markers either - fatigue, drift and inconsistency are real, and pretending otherwise helps no one. There will come a point when AI can responsibly assist in grading itself, not only in feedback, and as the evidence, the law and the sector’s experience mature, we should widen its role deliberately. The goal is not to keep AI permanently at arm’s length. It is to avoid throwing credibility out with the bathwater by moving faster than trust allows.

WHY WE’VE CHOSEN TO ASSIST - FOR NOW

That is the reasoning behind the line we have drawn at UNIwise, and it is why our AI assistant in WISEflow does what it does, and not more.

It assists with feedback, and it does so after the marking and the grade have been decided - never before or alongside them - so it cannot influence the result it is meant to support. The human marker stays in control of the decision and owns it. The AI works in a closed circle, drawing only on the assessment criteria, the assignment and the student’s own submission. When it is active, it is visible and traceable to everyone involved - institution, marker and candidate alike. It is also explainable: how the AI works and what it drew on to reach a suggestion can be surfaced and understood, rather than left as a black box - which is not only good practice but a requirement the EU AI Act places on high-risk AI in education. And all of it runs on EU-hosted, GDPR-aligned infrastructure with a defensible audit trail, so that supporting the marker never comes at the cost of the student’s privacy or intellectual property.

We will keep widening what our AI assists with as the evidence, the law and the sector’s experience mature. But we will not get ahead of the trust that makes assessment legitimate in the first place. That is what “assist” means to us in practice - not a limitation we apologise for, but a standard we have chosen.

THE POINT

The assist-or-dictate question will not stay still. As experience grows and trust is earned, the line will move, and some things that feel premature today will become responsible tomorrow. That is fine. What matters is that institutions move that line deliberately, with evidence and oversight - rather than having it moved for them by whichever tool a busy marker happened to open on a Friday afternoon.

For now, we are clear about where we stand. We assist. We keep the human in control. And we make sure the institution can show exactly how and why every assessment decision was made.

The closed-book exam comeback: why securing exams against AI needs layers, not a single setting

Rasmus Blok — Wed, 17 Jun 2026 13:23:48 GMT

Generative AI didn’t break assessment. It exposed how much we had quietly outsourced to the take-home essay.

For two decades, higher education moved steadily away from the invigilated hall. Take-home assignments, open-book papers and submitted projects were easier to run, kinder to timetables, and - we told ourselves - better aligned with how graduates actually work. Then generative AI arrived, and a format that rested on the assumption “this is the student’s own unaided work” could no longer carry that weight on its own.

The sector’s response has been visible and, in places, abrupt: a return to closed-book examination. Not everywhere, and not for everything - but for the assessments where authorship and identity genuinely matter, institutions are once again asking candidates to demonstrate competence under controlled conditions. The instinct is right. The execution is where it gets hard.

CLOSED-BOOK IS NOT ONE THING

The phrase “closed-book exam” hides an enormous range of risk and control. A reflective in-class essay on paper is closed-book. A high-stakes professional licensing exam taken remotely is also “closed-book”. Treating them as the same problem - and applying the same security to both - is how institutions end up either over-policing low-risk assessments or under-protecting high-stakes ones.

The useful question is not “open or closed?” It is “how much assurance does this specific assessment need, and at what cost to the candidate experience?” Security is a dial, not a switch.

WHAT MANY INSTITUTIONS DO - AND WHY IT DOESN’T SCALE

Faced with an urgent integrity problem, the common reaction is to reach for a quick fix: a lockdown browser bought for a department, a third-party proctoring tool trialled for a single high-stakes module, or a blanket rule that “all exams now use webcams”. These ad hoc measures share three weaknesses.

They don’t scale. A tool procured by one faculty becomes a support burden no one owns. A dozen point solutions across an institution mean a dozen contracts, a dozen data flows and a dozen different onboarding experiences for students.
They struggle on compliance. Many proctoring tools route biometric data - faces, voices, room scans - through servers outside the EU, with retention the institution never really controls. Under GDPR and the EU AI Act, that is precisely the high-risk processing that now demands a documented, defensible basis.
They create an inconsistent student experience. A candidate who meets a different surveillance regime in every module - some heavy, some light, none explained - is a candidate who, rightly, starts asking questions about proportionality and fairness.

Ad hoc works for a pilot. It does not work as policy.

WHAT GOOD LOOKS LIKE: PROPORTIONATE, SCALABLE, COMPLIANT

A serious approach to exam security has three properties. It is proportionate - the level of control matches the stakes of the assessment, chosen exam by exam rather than imposed institution-wide. It is scalable - one platform, one data flow and one student experience, whether you are running fifty candidates or thousands.And it is compliant by design - EU hosting, GDPR-aligned processing, defined retention and a clear audit trail, so that the integrity measure does not itself become a data-protection liability.

Get those three right and security stops being a series of emergencies and becomes a setting you choose with confidence.

HOW UNIwise THINKS ABOUT THIS - FOUR LEVELS, ONE PLATFORM

WISEflow treats assessment closed book integrity as a layered choice, not a single policy. Institutions pick the level - or the combination - that fits each assessment, all within one platform and all under the same GDPR-aligned, EU-hosted data processing, with defined retention for any captured media.

Device Monitor (monitoring). Full desktop access with complete activity logging, periodic screen capture and reviewable evidence. Best for tool-heavy exams — Excel, programming, CAD — that can’t run inside a locked-down browser but still need an integrity trail.
Lockdown Browser (lockdown). A dedicated client that confines the candidate’s device to the exam: no other apps, no copy-out, no unauthorised navigation, with built-in tools such as drawing, webcam capture and a code/LaTeX editor. Best for standard closed-book essay or MCQ exams that need a controlled environment.
Facial Comparison + Audio Detection (soft invigilation). Automated invigilation layered onto the lockdown browser — image capture matched against the student’s reference photo, configurable flags on low match scores, and voice detection with live transcription. It confirms the right person sits the whole exam, without a live human watching every candidate. Best for remote or on-site closed-book exams that need identity and presence assurance.
Full Proctoring (proctoring). Fully proctored online exams inside WISEflow — automated ID verification, webcam, screen and audio recording, an optional mobile second-camera room view, and your choice of review: self, AI, live human or post-exam. Browser-based, nothing to install. Best for high-stakes remote exams that need real-time supervision and a defensible record.

Because these levels live in one platform, an institution can run a low-touch Device Monitor exam in the morning and a fully proctored licensing exam in the afternoon - same system, same data governance, same student login.

THE POINT

The closed-book comeback is a sensible answer to AI. But “we’ve gone back to closed-book” is the start of the conversation, not the end of it. The institutions that will hold up to scrutiny — from students, data protection authorities and the press — are the ones that can show why each assessment carries the security it does, that the approach scales across the whole institution, and that every captured image, recording and log sits on EU infrastructure under clear rules.

That is the brief WISEflow is built for.

Rethinking special arrangements in higher education assessment

Rasmus Blok — Wed, 03 Jun 2026 08:26:13 GMT

A recent Norwegian article highlighted a sharp rise in applications for specially arranged exams, including extra time and separate rooms. While the figures were reported in a Norwegian context, the underlying issue is not uniquely Norwegian.

Across Europe, higher education institutions are seeing growing demand for assessment arrangements that enable students to demonstrate what they know without being unfairly disadvantaged by the format, environment or timing of an exam. At the same time, universities are under pressure to uphold academic standards, improve the student experience and run assessment more efficiently.

Read the Khrono article here.

The EU AI Act and assessment: December 2027 is not a snooze button

Rasmus Blok — Wed, 27 May 2026 08:29:47 GMT

The deadline moved. Use the extra time well.

On 7 May 2026, the EU Council and Parliament reached a provisional political agreement on the Digital Omnibus on AI. This pushes the application date for high-risk AI systems under Annex III - including AI used in assessment - from 2 August 2026 to 2 December 2027. High-risk AI embedded in regulated products receives a further extension, to 2 August 2028.

For European universities, that is sixteen extra months of runway on what had already been an uncomfortably tight deadline. But the key point is this: the deferral is conditional, and the underlying obligations have not changed. Only the date on which non-compliance starts to bite has moved - not the requirements themselves.

At UNIwise, we follow this legislation closely and are happy to share our perspective on what this agreement means in practice - both for institutions and for our shared work to ensure responsible use of AI in assessment.

Oral exams are back – and the real challenge isn’t pedagogical. It’s operational.

Rasmus Blok — Wed, 20 May 2026 07:55:23 GMT

For a while, oral exams felt like a legacy format: important in some disciplines, occasionally used for defences, but rarely the default at scale. Then generative AI arrived. Across higher education, I’m seeing more and more programmes rebalancing how they assess learning. Not necessarily by trying to “out-detect” AI (a game nobody wins for long), but by redesigning assessment so that what matters most is harder to outsource: understanding, judgement, and the ability to explain.

That is one of the reasons oral assessment is making a comeback. Sometimes it returns as a classic oral exam. Often, it returns as a hybrid model: a take-home submission (with aids allowed) followed by an oral defence. The take-home work supports deeper learning over time; the defence gives examiners a robust way to validate understanding and authorship in real time.

UNIwise launches Marking Studio at MoodleMoot Estonia 2026

Rasmus Blok — Wed, 06 May 2026 08:27:54 GMT

A new maker-first way to strengthen marking, feedback and integrity - without replacing your LMS

MoodleMoot Estonia 2026 became the stage for our first public launch of Marking Studio - our new, marker-first product suite designed to extend Moodle’s (or any other LMS/VLE) assessment capabilities through a clean LTI 1.3 integration. We joined as a sponsor with an exhibition booth and used the conference as our official launch moment - meeting Moodle administrators, educators, and partners in exactly the ecosystem we built Marking Studio for.

Marking Studio is built on a simple premise: keep Moodle or your LMS as the backbone - add Marking Studio where assessment needs more, directly from the course workflow. It is designed for opt-in adoption: start with one module, one course, one team, and expand only if and when it makes sense.

Happy Academics: The Metric That's About to Matter

steffen.skovfoged@uniwise.co.uk (Steffen Skovfoged) — Thu, 30 Apr 2026 09:43:55 GMT

Most universities can tell you how many exams they processed last semester. They can tell about submission rates, turnaround times, platform uptime. What almost none of them say anything about, and this is after 15 years of the sector going digital, is whether any of it is actually making their academics' lives better.

Not whether the system works. Whether it helps the academics.

Whether it reduces the time an academic spends on assessment. Whether it builds their confidence in the process. Whether it makes the hard parts of their job, the repetitive marking, the opaque workflows, the constant second-guessing about fairness and consistency, a little less hard.

That's the conversations we need to have and that we decided to make one of the central themes at our annual Partner Event this April. Not a product roadmap. Not a feature showcase. A conversation about the people who use WISEflow every day and whether we're genuinely serving them well.