AI in Academic Teaching: Supporting the Lecturer, Not Replacing Them

Artificial intelligence in teaching earns its keep where it shortens the lecturer’s work on materials, assignments, and feedback. It loses its value where it impersonates a teacher or decorates a press release. The boundary is straightforward. AI should take the tedious task of assembling materials from ten sources off the lecturer’s plate and give back time for working with students.

This article shows where AI genuinely helps in academic teaching, where the traps lie, and why the way it is implemented matters more than the choice of model itself.

Students Are Already Doing This — Universities Are Catching Up

First, an honest assessment. Artificial intelligence entered higher education faster than regulations could keep pace, and the numbers leave no room for doubt. A February 2025 UK study by HEPI and Kortext (1,041 students) found that 92% of respondents use generative AI tools — up from 66% the year before — 88% use them for assessed work, and the share of people not using AI at all fell from 47% to 12% within a single year. An international study published in PLOS One covering 23,000 students from 109 countries found that ChatGPT is a ubiquitous learning tool. Prof. Artur Strzelecki of the University of Economics in Katowice commented that virtually all of his students use it. At the University of Warsaw, 41% of students report regular AI use; the university was among the first to adopt formal GenAI guidance in teaching (led by Prof. Katarzyna Śledziewska’s DELab UW team).

Hence the second half of the assessment. Many universities have formal AI policies, but their application tends to be inconsistent — varying by faculty, supervisor, or even individual class group. The Lewiatan Confederation framed the risk precisely: the question is not “whether AI” but how to make it a daily support tool that raises teaching quality rather than “jewellery” the university showcases publicly while gaining no real benefit.

The Problem Is Not AI — It Is the Way We Assess

This point is worth developing, because it shifts the perspective of decision-makers. A study described in Science estimated, using an indirect survey method, that around 9% of students who used AI submitted AI-generated work despite knowing it breached the rules; among daily users, the figure rose to 26%. The authors’ conclusion is uncomfortable for traditional teaching: since the tool is ubiquitous, fighting the tool itself is a losing battle, and the space for action lies in how we assess and how clearly we define the rules. A university that does not address this will not stop AI — it will simply lose control over how students use it.

What AI Actually Does Well in Teaching

The value is greatest where the work is repetitive and time-consuming, while the intellectual stakes remain on the human side.

• Lesson preparation — draft outlines, materials, and assignments based on the syllabus; the lecturer approves and refines instead of building from scratch.
• Assignment variation — alternative versions of test questions, making copying harder and reducing workload with large groups.
• Feedback support — assistance in reviewing work and drafting comments, faster and more consistent, with the assessment itself remaining the lecturer’s responsibility.
• Student assistant — explanations grounded in course materials, available immediately rather than once a week during office hours.
• Identifying difficult areas — signals showing which parts of the material a cohort is systematically struggling with.

The common denominator? AI prepares the material; the human makes the decision. AI-supported content should also carry clear labelling — a requirement we will return to.

Sovereignty and Trust: What Matters Is Where the AI Runs

An issue that keeps returning in academic debate — and that decision-makers should take seriously — is dependence on large vendors and control over data. The Polish academic community is not limiting itself to discussion; it is building alternatives. PLLuM, the first Polish large language model, was developed by a consortium led by Wrocław University of Technology and made publicly available in February 2025, partly with public-sector applications and technological sovereignty in mind. Prof. Śledziewska puts the point bluntly: the greater danger may not be the tool itself but the presence of large technology companies inside the education system — which is why she calls for academic sovereignty and the development of homegrown solutions.

The practical implication for universities? When choosing an AI tool, the question “where does this run and where do the data go?” carries the same weight as “what can it do?” Teaching materials, student work, and personal data should not leave a controlled environment. This shifts the centre of gravity of the decision from the model to the architecture of the whole solution.

The LUMEN module in the MenToR platform uses models that process data exclusively within the European Union — teaching materials and student work never leave the university’s environment.

Compliance Is Not an Add-On

Deploying AI in teaching touches two regulatory regimes: GDPR (student data and work) and the AI Act (transparency requirements, including labelling of AI-generated content, applicable from August 2026). A well-built solution has these requirements built in, not bolted on — clear AI content labelling, data processed exclusively in the EU, access controls, and an event log. We unpack this in our article on GDPR and the AI Act in higher education.

From Teaching Support to a Coherent Environment

AI in teaching delivers the most when it is not a separate plug-in but part of an environment that also covers student services and class scheduling. In the MenToR platform, this is handled by the LUMEN module — it supports the lecturer in preparing and delivering classes and assists students in learning, with clear AI content labelling and data processed only in the EU. Combined with the AI dean’s office (DEAN) and hybrid classes (FORUM), it forms a single, coherent working environment for the entire university.

If you want to implement AI in a way that genuinely relieves lecturers rather than decorating your website, let’s start with a short conversation about your courses and processes.

Frequently Asked Questions

How many students use artificial intelligence? According to the February 2025 HEPI and Kortext study (1,041 students), 92% of respondents use generative AI tools, and 88% use them for assessed work. In the PLOS One study of 23,000 students from 109 countries, ChatGPT emerged as a ubiquitous learning tool.

Will AI in teaching replace the lecturer? No. AI prepares outlines, materials, assignment variants, and feedback support, but the intellectual decision and the grade remain with the teacher. The value lies in reclaiming the lecturer’s time, not in replacing them.

Can a university ban students from using AI? In practice, setting clear rules and assessment methods is more effective than fighting the tool itself. The Science study showed that the tool is ubiquitous; a university that ignores this does not stop the phenomenon — it loses control over how its students use AI.

Is using AI in teaching legal? Yes, provided GDPR and the AI Act are observed. The key requirements are: clear labelling of AI-supported content (transparency required from August 2026), data processed exclusively in the EU, access controls, and an event log.