It is 11:30 PM on a Wednesday, and I am staring at the twenty-fourth reflective essay of the night for my module. The grammar is flawless. The paragraph structure is immaculate. The actual human insight? Completely, utterly lost. No real connection with the student voice. And no real sign of session engagement.
This is the new reality of higher education. Many people don’t actually realise this but Artificial intelligence has technically been part of the educational ecosystem for years, running spellchecks, driving search algorithms, support analysis and quietly flagging ‘at-risk’ students in the background through tools such as Turnitin. But generative AI has violently kicked the front door in. Tools like ChatGPT, Gemini, Copilot and Claude can now draft, summarise, and rephrase complex theories in seconds. The days of the 3000 word essay are done. (Or at least I’d hope they are…)
The conversation usually immediately pivots to a charged debate about whether AI “belongs” in education. And ‘we need to ban our students using AI’ which I’ve heard from countless academics. I don’t think that is a useful position for anyone. It is already here, and students are experimenting with it whether we write it into the syllabus or not. Regardless of this we should be preparing students for the real world. And let’s face it, AI is certainly out there in the real world. The real question is how we adapt our assessments, protect academic fairness, and use these tools to kill the administrative grind so we can actually get back to teaching. We need to model the correct use of AI rather than pretend it doesn’t exist as a way to kick the can down the street.
The Reality of the Admin Grind
When people talk about AI in universities, they jump straight to cheating. That matters, but for those of us actually running courses, another major impact of AI comes from the administrative benefits, it buys back time to focus on the actual teaching and session development.
Drafting compliance emails, converting rough module notes into structured slides, producing four different variations of the same project brief for different cohorts—this is the hidden administrative drag that swallows a lecturer’s week without adding a single drop of educational value.
Microsoft pitches Copilot as an “orchestration layer,” which is a highly sanitised way of saying it can read a miserable forty-message email thread and spit out the three bullet points you actually need to care about. Used intentionally, this class of tool clears the deck. It gives staff the mental bandwidth to do the things that absolutely cannot be automated: face-to-face teaching, professional judgement, and building genuine confidence in a struggling learner through live explanation.
The Death of the Comfort Blanket Assessment
Generative AI hasn’t just improved at writing; it has mastered the art of plausible writing. It isn’t always factually correct, but it is usually good enough to pass a casual glance, especially in reflective or theoretical assignments. Remember most Universities, certainly in the UK set the pass bar at 40%. If using generic essay based assessment it’s not hard to hit this threshold using AI, even with its known flaws. Even complex scenario based assessments aren’t really that hard to cheat through using AI.
Because of this, our old academic comfort blankets are dead. Take-home essays with generic prompts and formulaic case studies are now incredibly fragile. Even if you design a highly specific, niche case study, a student can simply dump your entire PDF brief into an LLM, provide a tiny scrap of human context, and command the machine to generate a perfectly acceptable pass-level response.
The UK higher education sector is slowly waking up to this. We cannot hide from the plagiarism arms race. Relying on AI-detection software is a losing battle of false positives. Instead of trying to make an assignment magically “AI-proof,” we have to stress-test the design itself. We have to build assessments that make the actual human learning process visible. What is the experience we’re actually trying to sell? The collection of the certificate at the end of study or the journey to that point?
How We Actually Fix Assessment
We cannot just revert to pen-and-paper exams in sports halls, even though many of the more traditional academics are screaming for this. Yes it removes the AI problem, but in my view it doesn’t fulfil what education should truly be about. Supporting learners for truly developing a wide array of skills such as critical analysis, problem solving, initiative etc. Not just fact retention and memory recall! We need practical adaptations that force author ownership by design.
- Make the process legible, not just the product: If you only grade the final polished submission, you are simply rewarding whoever engineered the cleanest output. We need to shift the emphasis to process evidence—annotated drafts, decision logs, and short, messy commentaries explaining why a student changed their mind halfway through a project.
- Bring the friction into the room: Vivas, live demonstrations, class discussions and in-class drafting aren’t about catching students out. They are about seeing the thinking happen in real time. Even if a student used AI to prep their notes, they still have to stand up, look me in the eye, and defend that research live. That is a ring-fenced human skill required in almost every professional industry. Yes it takes the class a little while to wake up, but if you start this from day one they soon catch on.
- Use authentic constraints: Real-world projects feature messy data, furious stakeholders, and changing parameters. Scenario-based tasks that pivot halfway through force a student to use their own judgement rather than relying on a generic, pre-generated answer pattern.
- Set the boundaries: Vagueness creates chaos. If you do not explicitly define what constitutes acceptable AI use for a specific module, students will guess. Set clear rules, demand simple disclosure, and lower the temperature in the room. One of the hardest challenges I’ve had is convincing students that in my modules AI is green lit and encouraged. When I ask who is using it on day 1 we only have a couple of hands nervously raised. It takes the full length of the module to convince them that the whole class should be raising their hands in positive acceptance. What’s key is teaching the correct ethical way to use AI rather than bury our heads and pretend it doesn’t exist.
Immersion, Feedback, and The Quiet Costs
AI holds massive promise as a low-stakes thinking partner. It can provide alternative phrasing for a complex Building Information Modelling (BIM) theory, generate rapid formative quizzes to tighten up a student’s revision loop or quickly write up mini formative class exercises around any topic within seconds. Each bespoke to your brief. Moving beyond text, we are seeing immersive leaps in interactive design. Ubisoft’s Assassin’s Creed Discovery Tour is a brilliant example of this – stripping the combat out of a massive digital asset to create a purely educational, historically accurate architectural simulation that flexes around the learner.
If we get the assessment architecture right, we can start looking at where technology genuinely improves the learning experience.
But there is an uncomfortable reality here that the tech evangelists ignore: equity and data.
We are rapidly sleepwalking into a two-tier educational system. Students who can afford £20-a-month subscriptions to premium AI models with recurring memory will simply get better, faster tutoring and structural support than those stuck on restricted, hallucination-prone free tiers. Furthermore, some academics are feeding massive amounts of institutional and student data into these training models, often without clear long-term privacy frameworks. Yes models such as copilot enterprise claim confidentiality, but can we rely on that agreement knowing how many AI gone rogue cases we have. Gemini at least admits upfront that data may go back into training the model and may be reviewed by a human.
If we don’t treat this shift as a structural, architectural change, we will end up with the worst version of both worlds: students working around easily bypassed rules, educators buried under even more workload, and the slow erosion of meaningful learning. AI is not here to replace education, but it is ruthlessly exposing exactly where we have relied on lazy metrics and hidden labour. We as educators need to focus on the journey, and not dragging students towards a certificate with the least complaint.
Further Reading: Inside the Retro Tech Tonic Vault
- The Blueprint for Automated Environments: If AI is reshaping the classroom, it is also restructuring the physical world. A look at how automated tools are shifting human roles and physical compliance in the built environment.
- The Digital Sovereignty Dilemma: The idea that we can simply pull the plug on AI is a comforting fantasy. A deep dive into why stopping automated systems is far more complicated than a simple off switch.
- The Gradual Erasure of Craft: Automation rarely happens overnight. A look at the everyday software updates and interface changes where human capability is being quietly replaced.
- When Algorithms Build Worlds: What happens when video game environments are generated on the fly by algorithms? A critical look at the quiet replacement of human craftsmanship in gaming.
- Reclaiming Digital Agency: Digital convenience is brilliant until the authentication servers go down. Why building a physical media collection is about digital agency, not just nostalgia.
External Citations: The Reality of AI in Higher Education
The Russell Group Academic AI Principles: The definitive joint statement from 24 leading UK universities detailing how higher education must adapt teaching and assessment to ethically incorporate generative AI.
Russell Group: New principles on use of AI in education
UK Government Educational Policy Framework: The official Department for Education policy paper outlining the foundational stance on generative AI within the education sector and assessment frameworks.
GOV.UK: Generative artificial intelligence (AI) in education
National Centre for AI in Tertiary Education: The primary guidance hub from Jisc (the UK higher education technology body) on integrating and managing AI tools within university curriculums.
About the Author
I am Danny McGough, the founder and editor of Retro Tech Tonic.
For my day job, I work in higher education as an Assistant Professor and Course Director, spending my time designing and managing complex academic structures and modules or exploring into any research that interests me . My latest specialisms are around Building Information Modelling, Construction Innovation and Artificial Intelligence.
Retro Tech Tonic is where I step away from that institutional rigidity. It is my dedicated space to pick up a soldering iron, argue the architectural value of physical game media, and push back against the frictionless convenience of modern tech. I believe technology should serve as an intentional, human-controlled tool, and I write to figure out exactly where those boundaries lie.