AI at Swiss Vocational Schools: Between Practical Relevance and Academic Integrity

Vocational Education and AI: A Different Starting Point

Swiss vocational education is unique worldwide. Around two-thirds of all young people in Switzerland complete a vocational training program — a combination of apprenticeship in a company and classes at a vocational school. This dual system has particular strengths, but also particular challenges when it comes to generative AI.

At the gymnasium, the AI debate is primarily a question of academic integrity: did the student write the essay themselves? In vocational education, a second dimension emerges: how does AI use in the school portion relate to AI use in the training company? And how do you handle professions where AI use is explicitly part of professional competence?

This article shows how Swiss vocational schools are answering these questions — with concrete recommendations for the individual practical assessment (IPA), the Berufsmaturitaet thesis, and collaboration between school and company.

What Makes Vocational Education Different From Gymnasiums

Four factors make the situation at vocational schools fundamentally different from gymnasiums:

1. The company as a third actor: Beyond school and student, there's the training company with its own expectations and rules. A company that uses ChatGPT internally can hardly accept the school banning it entirely.
2. Practical relevance as core claim: Vocational education is meant to prepare for the working world. If AI becomes standard in the workplace, the school must reflect this — not ignore it.
3. Heterogeneous professions: AI relevance varies enormously across professions. A software developer uses AI daily. A carpenter rarely needs it for core work. An office clerk falls in between.
4. The IPA as practical exam: The individual practical assessment takes place in the company but is evaluated by external experts. School rules don't directly apply here.

The IPA (Individual Practical Assessment): Special Case and Acid Test

The IPA is the central final exam in basic vocational education. It takes place at the training company, lasts from a few hours to several days depending on the profession, and is evaluated by examination experts. For the AI question it's especially relevant because:

• The work takes place in a professional context where AI may be part of daily work.
• Evaluation often covers both process and product.
• The examination experts don't have the same history with the student as a teacher does.

AI Rules for the IPA by Professional Field

IT and Mediamatik

In IT training, AI is a particularly sensitive topic: GitHub Copilot, ChatGPT, and Claude are standard in many developers' daily work. The IPA must therefore differentiate:

• Allowed: AI assistance for code completion, debugging, research — provided the student can explain and modify the code.
• Not allowed: Complete code generation without independent adaptation. The IPA should test the student's competence, not the AI's.
• Documentation: All AI use is recorded in the IPA documentation. Experts ask specifically about AI-assisted passages during the professional discussion.

Commercial Professions (KV)

In the commercial sector, the situation is less clear. AI can create business letters, emails, reports, and presentations — all core competencies of commercial professions. IPA regulations must therefore be more restrictive than daily work practice:

• Written IPA components are typically declared as independent work.
• AI may be used for research and spell checking, not for creating business correspondence.
• The professional discussion tests whether the student understands the content and can independently explain the documents.

Trades and Technical Professions

In trade and technical professions (carpenter, polymechanic, electrician), AI plays a lesser role in practical work. The written IPA documentation is vulnerable, however — similar rules apply as for commercial professions.

Healthcare Professions (FaGe, FaBe)

Healthcare assistants and social care workers have IPAs that often include written reflection and documentation components. AI use for reflection texts undermines the pedagogical purpose of this component. Here the rule is: independent work without AI support for reflection texts.

The Berufsmaturitaet Thesis

The Berufsmaturitaet thesis is the academic counterpart to the IPA — a written paper produced during the vocational baccalaureate that corresponds in scope and rigor to a small matura thesis. For AI detection, the same principles apply as at gymnasiums:

1. Level 3 as standard: AI as writing partner with journal. All final formulations are original work.
2. AI journal: Ongoing documentation of AI use.
3. Regular check-ins: At least two meetings with the supervising teacher.
4. Technical verification: Detector use on the final version.
5. Oral defense: Presentation and professional discussion where the student defends their work.

A special feature of the Berufsmaturitaet thesis: it often directly relates to the training company. When a commercial apprentice writes about digitalization in their company, they have access to internal data and experiences no AI can deliver. Assignments emphasizing this practical connection are automatically more AI-resistant.

Collaboration Between School and Company

The biggest challenge in vocational education is coordination between learning locations. If the school bans AI but the company uses it daily, confusion arises. Conversely: if the school allows AI but the company expects students to "do everything themselves," there's friction too.

Three recommendations for learning-location coordination:

1. Formulate shared principles: School and training company agree on a simple framework covering both contexts. Core message: "AI is a tool, not a replacement. In exam situations, the examining institution's rules apply."
2. Involve training supervisors: The company's vocational trainer should know which AI rules apply at school — and vice versa. A brief exchange at the start of the training year often suffices.
3. Professional organizations as multipliers: Industry organizations can formulate sector-wide recommendations on AI use in training — giving individual companies and schools orientation.

Sector-Specific Perspectives

ICT Professions

In the ICT sector, AI competence is part of the job profile. The training plan for IT specialists is continuously updated. The question here isn't "whether AI" but "how to use AI responsibly." Vocational schools that blanket-ban AI act against the training plan.

Commercial Professions (KV Reform 2023)

The KV Reform 2023 fundamentally restructured commercial basic education — with stronger focus on action competencies. AI fits this picture: the competence to draft a business letter remains relevant even when AI helps with the draft. What matters is that the student can evaluate, adapt, and take responsibility for the result.

Health and Social Care

In health and social care professions, reflective competence is central. Care documentation, case discussions, learning reports — all must be authentic and personal. AI use in these areas is not just an integrity problem but a pedagogical one: delegating reflection to a machine means not reflecting.

Trades and Crafts

In many trades, the AI question is primarily relevant for written documentation: work journals, reports, IPA documentation. The practical exams themselves — installing a kitchen, laying cable, milling a workpiece — are AI-immune. The written components should be protected with the same principles as at gymnasiums: draft requirements, oral defense, technical verification on suspicion.

Data Protection in Vocational Education

Data protection questions in vocational education parallel those at gymnasiums but add a dimension: when IPA texts are uploaded to a detector, they may contain company-internal information — client names, process descriptions, confidential business data. This requires extra care:

• Anonymization: Remove company-internal details before text enters the detector.
• Swiss hosting: Even more important than at gymnasiums, as trade secrets may be involved.
• Inform the training company: The company should know that IPA texts may be technically verified — and which data protection measures apply.

A Concrete Scenario: Commercial Clerk, Second Year

Anna is in her second year training as a commercial clerk at a Zurich fiduciary firm. At vocational school, she must write a paper on "Digitalization in my training company" (Level 3). Here's how it goes:

1. Topic and outline: Anna creates an outline and discusses it with her teacher. She uses ChatGPT to brainstorm subtopics — she records this in her AI journal.
2. Research at the company: Anna interviews her training supervisor about the firm's digitalization strategy. This interview is original work no AI can deliver.
3. Draft: Anna writes a rough draft and uses ChatGPT for feedback on the structure. The AI suggests reordering two sections. Anna decides herself whether to follow the suggestion.
4. Clean copy: Anna writes the final version herself. She uses DeepL Write for grammar correction. The AI journal documents all steps.
5. Submission: The paper is submitted with the independence declaration and AI journal. The teacher spot-checks the text with AIDetector.ch.
6. Assessment: The work is good — and traceably self-written. The transparent AI use is acknowledged positively.

Recommendations for Vocational Schools: The Action Plan

1. Adopt an AI policy: Simple, clear, adapted to the professions offered. The five-level model can be adopted directly and adjusted to the needs of each profession.
2. Inform training companies: A brief letter or information session at the start of the school year. Core message: "This is how we handle AI at school. This is what we expect from students. You can supplement this in the company."
3. Train teachers: Half a day per year on AI competence, focusing on detector use, assignment design, and handling suspected cases.
4. Procure a compliant detector: Centrally for the school, with DPA and Swiss hosting. AIDetector.ch offers vocational schools the decisive advantage: support for all national languages and classroom mode for Berufsmaturitaet thesis batches.
5. Coordinate IPA guidelines with cantonal examination commissions: AI rules for IPAs shouldn't be set by each school alone but in coordination with cantonal examination authorities and professional organizations.

Conclusion: Vocational Education Needs Its Own Path

The AI debate in Swiss vocational education can't simply be copied from gymnasiums or universities. The dual structure, the diversity of professions, and the close connection to the working world demand their own approach — one that combines practical relevance and integrity instead of playing them against each other.

The good news: vocational education has a built-in advantage no other level has. The IPA takes place under observation. The professional discussion is part of every exam. Practical competence can't be faked with ChatGPT. What remains is the challenge of the written components — and for that, there are now tools, models, and experience that vocational education can use too.

Sources

• SERI (State Secretariat for Education, Research and Innovation), framework curricula for basic vocational education.
• ICT Vocational Education Switzerland, training plan for IT Specialist (updated edition).
• SKKAB (Swiss Conference for Commercial Training and Examination Branches), implementation aids KV Reform 2023.
• OdASante, training plan for Healthcare Assistant.
• Federal Act on Vocational and Professional Education and Training (VPETA), SR 412.10.
• Federal Act on Data Protection (revFADP), SR 235.1.