Key Takeaways from the AI in TVET expert meeting in Magdeburg 2025

On February 18–19, 2025, an expert meeting on AI and TVET was held at the inspirational location of the Elbfabrik Factory of the future.

Experts from industry, academia, and international organizations gathered to discuss how artificial intelligence is reshaping the landscape of vocational training, the ethical and practical implications of these technologies, and what the future holds for educators and learners alike.

Conference overview

The event was co-organized by UNESCO-UNEVOC, Fraunhofer IFF, Otto von Guericke University, and the GiZ TVET Academy. The format was very well balanced between keynotes, workshops, panel discussions, networking events, tours of the Elbfabrik and the 360 degrees mixed reality lab Elbedome.

In the article below we summarize the key events and our takeaways from the conference.

Opening remarks

The day was moderated by Christian Blobner – Fraunhofer IFF, and Prof. Dr. Frank Bünning – Otto-von-Guericke University and Dr. Günter Podlacha – GiZ TVET Academy.

It started with some opening remarks:

Prof. Dr. Tina Haase: Head of Department Human-Centered Work Systems, Fraunhofer IFF.

Prof. Haase made an introduction to Fraunhofer IFF which aims to do applied research in „the factory of the future“ taking both a technical and a social point of view in making factories more efficient. Her personal focus is on design-optmimization of Human-Machine-Interfaces (HMIs).

Dr. Friedrich Huebler, Head of UNESCO- UNEVOC

Unesco-Unevoc is supporting UN member states in implementing TVET. To preserve human agency in a world which is becoming more automated, Dr. Huebler advocates for the ethical guidelines for the use of AI which have been recently published by UNESCO, and suggests that we keep these guidelines in mind during this expert meeting.

Keynotes

Two keynotes were given, with lots of thoughtful points.

Prof. Dr. Dr. hc. Georg Spöttl, University of Bremen

Prof Spöttl who is the Director of the Centre of Technology, Work and TVET (CTAB) at University Bremen and Director of Steinbeis Center InnoVET, made a keynote presentation.

Prof Spöttls’ discussion centers on whether AI in TVET bridges gaps or creates barriers. While rapid advancements in AI demand societal acceptance, there is a clear need to implement safeguards—such as preventing AI from autonomously advancing its own research capabilities and ensuring that content is rigorously verified.

Prof Spöttl argues that beyond technological literacy (as defined by ITEA), there’s a crucial need for vocational competence, ensuring that individuals can effectively shape work and technology to assess, manage, and utilize AI.

Questions arise about how much technology is appropriate, and what role humans should continue to play in work processes. Three potential scenarios are discussed: a technology approach, a human-centered approach (emphasized by UNESCO- UNEVOC), and a hybrid model combining both.

Notably, the current expectations on AI in the workplace is highlighted by a survey in 2024 which is asking 2,778 AI researchers. If science continues undisrupted, the chance of unaided machines outperforming humans in every possible task was estimated at 10% by 2027, and 50% by 2047. Are all human occupations becoming fully automatable?

This begs the question, how does this impact learning processes and curriculum development? There are concerns about “skill-skipping” where an overemphasis on outcomes might neglect the essential learning process.

Prof Spöttl recommends a hands-on driven approach in which to:

organize training in cyber-physical production environments (such as learning factories etc)
provide work-integrated training in schools and companies (both real and simulated)
to train directly on digital instruments / machines of production (learning HMIs)

Further, to address skill skipping, a framework for AI literacy in technology education is proposed which includes a “shadow curriculum”. Prof Spöttl also calls for work-process-related AI frameworks, and (UNEVOC) alliances for promoting AI in education, training and leaning – all aimed at developing TVET curricula that keep human agency at the forefront.

Prof. Dr. Uta Wilkens, Ruhr-Universität Bochum

Prof Wilkens who is also the director of HumAIne, a transfer hub for human-centeredwork with AI, made a keynote on the perspectives of AI in Vocational Education.

Prof Wilkens suggests that the main challenges of generative AI is that it shifts the nature of critical thinking from problem-solving to information verification and response integration. That over-reliance on AI may reduce deep and critical thinking, leading to an illusion of expertise.

Therefore, Prof Wilkens main message is that TVET must Teach AI literacy and effective AI usage. Foster human-AI role development while maintaining critical thinking. Incorporate knowledge of job design and ergonomics to address long-term ethical implications.

Prof Wilkens further argues that current approaches focusing on AI for automation, and AI as an idea-generation tool is not enough, that we need to develop integrative human-AI user concepts.

Approach	Organizational Benefit	Individual Benefit
No or unsystematic AI use	Limited efficiency, manual work remains	No AI support for learning & automation
AI for automation	Increased efficiency, but no long-term reflection on expertise needs	Routine tasks automated, but reduced demand for critical thinking
AI as an idea-generation tool	Boosts innovation	Helps users develop ideas and skills, but risks cognitive outsourcing
Integrative Human-AI User Concept	AI supports critical thinking and long-term job design	Users retain and enhance cognitive skills

For Industry 5.0 AI skills should be bundled with social, cognitive, and entrepreneurial skills. Companies need to balance AI skills with complementary human skills (e.g., leadership, creative thinking, project management).

Prof Wilkens discusses how AI’s Impact on Technical Vocational Education & Training (TVET) can be looked at from a quantitative and a qualitative lens. The key is that AI automation and rationalization will replace tasks, while innovation and structural change will be generating new tasks. The essentials are how we train and re-train, for the new tasks which will be created, and how we use job-design concepts to make sure we keep the human in the center.

For responsible AI integration, AI literacy alone is not enough—it must be combined with role development and job design. Organizations should: Develop AI-augmented roles rather than simply automating tasks. Implement AI in ways that enhance human skills, rather than diminishing them. Ensure workers understand AI limitations and ethical use.

HumAIne has created a toolbox for how to consider AI in strategy & business model development, instruments for work and work system design, and instruments for human-centered technology development.

Workshops

Three workshops were held in parallel, and attendees could either jump in between these workshops or stay the full session in one workshop. I personally stayed in the workshop 1 for the full session, so the summary of workshop 2 and 3 are very light.

Workshop 1: “Perspectives on AI (em)powered Training and Practice”

This workshop was moderated by Olga Harms and Ulrike Schmidt from GiZ, and we were asked to consider the following questions while going through the workshop:

How can AI empower trainers and learners rather than replace them?
What are the opportunities and ethical considerations of AI in training?
How can we ensure that AI is integrated in ways that truly benefit users, practitioners and learners?

Dr. Oliver Nahm from Bundesinstitut für Berufsbildung (BIBB) gave a high-level presentation on Empowering Training with AI: From Concepts to Implementation, main points below:

How to look at AI companion: should be seen as a well-read intern, who is a bit overly confident, even when he’s wrong. So needs guidance (specified prompts), and clarification of context.
Limitations: hallucinations / Prejudices (Political manipulation is completely forbidden in EU) / limited knowledge / Token limit = Short term Memory / Math, orientation, logic…
Use cases: lesson planning, creation of tasks, correction of tasks, roleplaying, chatbots for learning support etc.
Prompting: the best practice for prompting is to always include the goal, explain who the audience is, and what type of learning you would like to achieve.
Examination and Al: some institutions say that students can incorporate Al, but need to also include their prompts, and will need to be able to explain orally when asked. Teachers start focusing much more on oral exams.
Deskilling: there is a risk that we start taking AI too much at face value, and it is important to further strengthen critical thinking.
Disinformation: AI can be used for disinformation / incorrect information, which includes both text, photos, audio and video.
Al competences: should be taught, and skills on how to critically review information.
Promise of AI: potential re-humanization, pushing boring tasks to Al and keep people focusing on the fun tasks.
How to deal with Hallucinations? The problem with all AI models is that even if you use a RAG system (attached documents) and say that he should stick only to this. AI is still only 80-85% correct, would you trust such outcomes?
Flipped classroom: give factual materials and assignments to students to deal with at home, and let them come to class to have discussions and debates. Could lead to boom in critical thinking.

Rune Gråbæk and Lars Reng from Zealand Business College in Denmark gave a presentation on their project “AI Agents for Education in Healthcare”.

This team has made static mannequins more realistic by making them talk back to a health care trainee, which is achieved through some very simple microphones, speakers and a phone which has activated the voice command functionality of ChatGPT.
They have through prompting ChatGPT created very realistic patient personas with different healthcare problems, personalities, language skills, temperament etc. The results are that their AI agents lead to realistic roleplay between the trainee and the mannequin.
They currently only do this in Healthcare but point out that their approach can be used in any career focused roleplaying, such as when training sales people / customer support / coaches / craft-teachers (TVET) / secretary / knowledge experts / tutors / mentor etc.

Wassila Ben Aissa from GiZ in Tunisia presented a project within the topic of Thamm-Plus, “Comparison of training programs with an AI platform”.

This team has created an AI driven tool to support the work of administrational staff.
It aims to reduce processing time, improves accuracy, and ensures a standardized and scalable approach to diploma recognition.
Functionalities: Automated extraction of educational program structures. AI based translations (German – French). AI powered comparisons – identifying similarities and gaps. Expert review & continuous model improvement. Intelligent assistant.
It is still too early to clarify the accuracy of the comparison tools, but the team is working on improving the model.

Workshop 2: “AI competencies on the job: Co-creating a blueprint for workplace AI literacy”

Prof Sebastian Lang and Johannes Schleiss, from Otto-von-Guericke-University in Magdeburg ran a workshop in this they identified different personas with needs for learning. Then they created courses for these personas using an “AI course framework” (an open-source model), by looking at questions such as how can AI be used in this domain? Which type of data exists in this domain? Which qualifications do the persona need?

Workshop 3: AI in education and training (curriculums) and competency development

This workshop included a few presentations as per below:

“Potentials, Challenges and selected examples for addressing the implementation of AI into TVET” Dr. Harry Stolte
“AI in higher Education teacher education – current benefits and challenges” Hannes Tegelbeckers, Otto von Guericke University Magdeburg
Integration and Utilization of AI Tools in Teacher Training at BBPPMPV BMTI, Moh Sanni Mufti Alamsyah, BMTI Indonesia
“Leveraging AI for Skills Matching: Empowering India’s Workforce for Climate Action across Sectors” Gunjan Jhunjhunwala, CEEW, India

Podium Discussions

What started out with a few opening remarks from a podium discussion quickly led to an open conversation between everyone in the room. This was a highlight of the day, with so many valuable comments, below they are grouped under four different topics, coming from both panelists and the audience.

On how AI impacts the teacher role

In the general economy there is fear that AI will impact some roles, which will entirely disappear, while others will arise. But how will AI affect teachers, lecturers and trainers?
With more supporting technology such as AI, the role of a teacher is developing more into a “learning coach”, for social interaction, and the content in domain specific knowledge, and how to use these tools.
Learning embodies more than knowledge acquisition, there is ethics, attitudes, critical thinking project work, collaboration etc.
Teachers’ role is not only a knowledge transmitter, it also about forming the learners, so for sure the role will continue changing, is this not something good?
In some parts of training, in some tasks, AI is already replacing teachers, it’s seen already in some schools in UK, as well as the International University; if teacher is ill, then the replacement lecture is done by Al.
Self-guided learning does not work in TVET because of the need for hands-on training and immediate structured guidance, motivation, self-discipline etc.
Yes, that might be trued on lower level TVET, where students associate with the teacher and you cannot remove that teacher, but at higher level TVET the content is key and the teacher is not always important.
It’s not about replacing teachers, it’s about how do we change education.
Al or edtech will not replace teachers, only the teachers who use technology will replace those who don’t.

On implementation of AI

AI fits well with experiential learning and by putting the user in the center, as discussed earlier in the day.
Introduction of new tech needs an interdisciplinary approach, so it is important to create acceptance early in the process.
This is social sciences and people are wrong all the time, we don’t know what will happen. AI processes languages very well, so summarizes topics in a very good way. It can be a starting point on any topic but requires further deep diving without AI into harder topics.

On trusting AI

Hallucination is a problem though, 80-85% correct is not good enough. Does AI need oversight? Is it even possible?
AI hallucinates all the time, it’s only when it doesn’t fit within our expectations, that we realize that it’s hallucinating, remember, it’s just statistics.
The use of AI itself by all users reinforces the model, and this can both improve and impair the actual model, we don’t know fully how these models will adapt yet.

On the need for Educational reform

We have entrusted agriculture, healthcare, finance, transport etc. to technology over the last decades, but we have not done a lot in education,
We have been talking for a long time about reforming TVET (as well as other parts of education), let’s look at AI as the tool for actually doing that.
Education system is built for industry 2.0 and “employer dictates employee what to do” which does not match how the world is changing and the societal needs. Hence education in general and TVET must take on more open-ended Montessori styles of learning.

City Tour & Joint Dinner

The day was closed with a city tour of the center of Magdeburg, including history of the Kloster Unser Lieben Frauen, the Hundertwasser’s Green Citadel, the Magdeburg Cathedral, and the old defense walls of the city. Followed by a networking dinner.

Next day: Industry Perspective & Lab visits

The following day started with a presentation from industry, discussing employment skills and competencies, followed by visits to Elbdome and Elbfabrik, and a live coding exercise.

Defining 21^st Century employment Skills and Competencies

Daniel Büchner from IBM is responsible for a large software development & asset based consulting team.

Their biggest cost is personnel and the work they do is very time-consuming, so they look at how to include AI in their different workflows.

IBM has its own AI tool for business which is called Watson X and they have clear indications that their consultants which use this tool are much faster than those who don’t.

A challenge which they have identified is that many of their younger talents manage a lot of work which is run by AI, and thus might lose out on a lot of the contextual learning which is necessary later in their careers, such as to evaluate whether the output is good. There is a risk that AI is always giving the same standardized approach.

Top Employment skills for IBM: creativity, intuition, client relationship building, critical thinking, empathy, nuanced understanding of client needs, expertise in understanding client challenges and fostering collaboration, quickly analyzing data to make use of it.

Presentation of the 360°-Mixed-Reality-Lab Elbedome

Prof Haase from Fraunhofer IFF presented the Elbedome, a 360° projection system for VR without glasses. As you can see from the pictures below, the system can project a whole townsquare on these 360° walls, and we also got to see factories, swimming halls, and a futuristic city.

The Elbedome is used for education & training, interactive and creative workshops, construction planning & interior design, virtual prototyping, visualizing machines, systems and factories.

Tour of the Elbfabrik of the Fraunhofer IFF

Christian Blobner took us for a tour of the factory of the future, the Elbfabrik, where Fraunhofer IFF is doing applied reserarch, and helping industry to run test projects. They typically work on designing efficient, flexible, sustainable and secure value creation processes.

For an industrial automation and skills enthusiast, this was the highlight of the conference.

Live Coding exercise for presentation and teaching material creation

Dr. André Dietrich, TUBAF Freiberg held a very interesting coding exersice for the creation of Open Educational Resources (OERs) by leveraging LiaScript. LiaScript is a tool that enables educators to develop interactive and programmable content using Markdown, even without extensive technical skills. This approach aims to make digital content creation more accessible, allowing experts from various fields to contribute effectively to OERs.

Conclusion & Takeaways

The Expert Meeting on AI and TVET in Magdeburg 2025 underscored the transformative potential of artificial intelligence in technical vocational education and training. Key takeaways include:

Human-Centric Integration: While AI offers tools to enhance educational processes, it is imperative to maintain a human-centered approach, ensuring that technology serves to augment human capabilities rather than replace them.
Ethical Considerations: The rapid adoption of AI necessitates a robust ethical framework to guide its implementation, safeguarding against biases and ensuring equitable access to technological advancements.
Curriculum Evolution: Educational curricula must evolve to incorporate AI literacy, preparing learners to navigate and contribute to an AI-infused workforce effectively. Even better if it also includes leadership, creative thinking, project management.
Collaborative Efforts: The convergence of insights from academia, industry, and international organizations highlights the importance of collaborative efforts in shaping the future of AI in TVET.

As we move forward, embracing AI’s potential while addressing its challenges will be crucial in redefining vocational education for the modern era.

Organizers publication.

The organizers UNESCO-UNEVOC, Fraunhofer IFF, Otto von Guericke University, and the GiZ TVET Academy are expected to release an official publication in October 2025, make sure to follow their social media accounts and email newsletters since this will surely be an interesting read!

Share your comments!

Were you part of the meeting and think I forgot something? Let me know and we will include it in the article! Do you agree or would you want to share an alternative viewpoint? Comment below!

About the Author: Timo Wohlin-Elkovsky

I have been spending 13 years in the field of instructional materials for teaching STEM and TVET. I am currently the Co-director of TVET Journal and TVET Jobs, as well as the Founder & CEO of Edquip, a marketplace for technical training equipment.

Feel free to connect with me on LinkedIn and if you would like to collaborate with us at TVET Journal, we would be happy to discuss.