Developing Vocational Education and Training with Artificial Intelligence

Executive summary

Rapid labour market changes – driven by digitalisation, artificial intelligence (AI) and the green transition – are placing growing pressure on vocational education and training (VET) systems to develop and update curricula and qualifications more frequently and efficiently. While VET development processes are typically robust, collaborative and grounded in strong stakeholder engagement, they are often resource‑intensive and struggle to keep pace with fast‑evolving skill needs.

Why AI matters for VET development

Curricula and qualifications form the backbone of VET programmes, translating labour market needs into structured learning outcomes, programmes and certification. When they are outdated or misaligned with labour market changes, qualification and skills mismatches persist, leading to skills shortages, even where qualifications remain trusted signals of competence.

In this context, AI opens up new possibilities to make VET development processes more responsive, efficient and evidence‑driven. AI can support the timely detection of trends from large and complex data sources, facilitate mapping and comparison of competencies across standards, qualifications and curricula, accelerate drafting, revision and technical validation processes, and structure diverse stakeholder inputs more efficiently.

However, the added value of AI use in VET development depends less on the technology itself than on how effectively it is embedded within existing governance, quality assurance and multi-stakeholder, collaborative arrangements that underpin trust in VET systems.

Current and emerging AI use in VET development

AI use in VET development is uneven across stakeholders and countries, and exploratory. It is more common among public VET agencies and industry bodies, with limited and often informal uptake among VET providers and teachers. Current applications are predominantly supportive rather than transformative, complementing established VET development processes.

Al applications span the full development cycle: supporting input-oriented tasks such as labour market analysis, skills anticipation and mapping (e.g. Croatia, England [United Kingdom], Germany and Mexico); process-oriented activities such as consultation and synthesis of stakeholder inputs (e.g. Australia, Estonia, Nova Scotia [Canada] and Switzerland); and outputs such as drafting learning outcomes, checking consistency and aligning curriculum content with qualification frameworks and occupational standards (e.g. Korea, the Netherlands and New Zealand).

AI use is more visible in sectors characterised by rapid change – notably digital, green and AI-related fields – and in flexible forms of provision such as modular learning and micro-credentials. Countries with stronger data infrastructures and established skills intelligence systems are better placed to move beyond small-scale pilots and experiment with more advanced AI applications (e.g. Estonia and England). Most AI tools used in VET development are general‑purpose applications or customised versions of existing platforms. Notable exceptions include England’s SkillsCompass (a platform using natural language processing and generative AI to analyse occupational standards, real-time labour market data and foresight trends to identify skills needs); the Netherland’s SBB-Chat (a GPT‑4‑based platform for qualification drafting and process support by the Organisation for the Cooperation between VET and the Labour Market [SBB]); and Switzerland’s ICT sector, which has developed occupation-specific custom GPT models for modular curriculum design and other VET-related tasks. Fully integrated, system-wide solutions remain rare, although the Netherlands is moving towards a co‑ordinated, cross-institutional approach by providing a secure AI platform that integrates multiple AI tools, with outputs validated by VET teachers.

Addressing challenges in VET development

AI is not a panacea, but it can help alleviate several long-standing bottlenecks in VET development.

• Balancing stakeholder interests: VET development involves social partners, VET teachers and public authorities, supporting labour market relevance, learner employability and legal compliance. However, this collaboration also creates co‑ordination challenges and lengthy consultation processes, slowing decision making. AI can improve efficiency and inclusiveness by synthesising large volumes of input coherently and lowering participation barriers for less-resourced stakeholders through preparing evidence briefs.

• Managing time and resource constraints: VET development can take years, with lengthy procedures and limited resources as major constraints. AI can ease administrative and analytical burdens through drafting support, comparison tools and automated checks for validation and compliance – allowing experts to focus on high‑value judgement and consensus-building.

• Translating labour market change into timely updates: Labour market information is increasingly abundant and real-time but fragmented and difficult to interpret. AI can integrate heterogeneous data sources (e.g. job vacancies, skills forecasts and qualification registers) to identify gaps and emerging needs earlier, supporting timely and evidence‑informed updates, particularly in fast‑changing sectors.

Barriers and risks associated with AI use

Constraints limiting effective AI use in VET development include gaps in AI, digital and data literacy; uneven access to secure, fit‑for‑purpose tools; fragmented or low-quality data; limited VET‑specific guidance and institutional support; and resource constraints, particularly for smaller providers and industries.

Potential risks are equally important to consider. AI could weaken collaborative governance, especially when AI outputs substitute for stakeholder dialogue; blur accountability and transparency if responsibilities for AI‑supported decisions are unclear; or introduce quality concerns through over-reliance on unvalidated or biased AI outputs. Data protection and security risks are particularly important when handling sensitive, unpublished or commercially sensitive VET data. These challenges are not only technical but institutional, underscoring the need for careful design, governance and oversight.

Five principles

Building on emerging country experience and existing OECD and EU AI frameworks, five principles guide AI use in VET development:

• Human‑centred use: AI should support, not replace, expert judgement, validation and collective decision‑making in VET development.

• Diversity and inclusiveness: AI use should broaden participation and avoid advantaging only well‑resourced stakeholders, also respecting the choice not to use AI in VET development.

• Accountability: Responsibility for decisions and outcomes in VET development must remain with human actors and institutions.

• Transparency: VET stakeholders should understand when, how and for what purpose AI is used, and its limitations.

• Data quality, security and protection: AI use must rely on high‑quality, relevant data and robust safeguards, ensuring compliance with legal requirements and maintaining trust in VET systems.

Policy considerations

To move beyond fragmented experimentation and realise AI’s potential, countries should consider:

• Establish strategy with clear purpose and support: Establish coherent national and VET‑sector AI strategies that define where AI adds value in VET development, align AI use with system‑level objectives, and provide role‑ and task‑specific support.

• Manage risks through a human‑centred approach: Manage AI‑related risks by embedding human‑centred governance in VET development, complementing general AI frameworks with VET‑specific guidance that ensures expert validation and accountability for AI‑generated outputs.

• Balance diverse perspectives while ensuring equal access to AI use: Balance diverse stakeholder perspectives, institutional capacities and regulatory requirements by enabling innovation in AI use for VET development while ensuring equitable access.

• Co-create guidelines and build capacity for strategic AI use in VET development: Co‑create VET‑specific AI guidelines with social partners and practitioners, and strengthen continuous, targeted capacity‑building to support transparent and task‑relevant AI use across VET development processes.

• Strengthen data infrastructures and governance: Improve data availability, quality, interoperability and machine‑readability for VET curricula and qualifications, while reinforcing strong data governance, security and ethical safeguards for trustworthy and effective AI use.