This chapter explores how artificial intelligence (AI) is beginning to reshape the development and revision of curricula and qualifications in vocational education and training (VET). Drawing on OECD survey data and country case studies, it shows that while AI adoption remains uneven and largely experimental, its use is expanding across key tasks of VET development – from labour market analysis and stakeholder consultation to curriculum drafting, mapping and format validation. The chapter highlights how governments, industry partners and VET providers are using AI to enhance efficiency, responsiveness and evidence‑based decision‑making. It also examines the types of tasks, programmes and data where AI is most commonly applied, including green and digital skills, modular VET and micro‑credentials, and discusses the governance, data and capacity conditions shaping current practices.
Developing Vocational Education and Training with Artificial Intelligence
Abstract
How do countries use AI to support VET development?
Current and emerging use of AI to support VET curriculum and qualification development:
AI use in VET development is at an early stage – largely supportive, uneven and often informal, with significant variation across stakeholder groups and countries. VET policymakers report the highest awareness and adoption of AI in VET development, followed by industry stakeholders and VET providers.
The most relevant and common applications spans input (data analysis, forecasting and mapping), process (stakeholder consultation and evidence synthesis) and output (content generation and framework modernisation) tasks of VET development.
AI use is relatively visible in areas linked to digital, green and AI-related skills, and in the development of modular provision and micro-credentials.
AI-supported VET development draws on a broad mix of labour-market intelligence, qualification frameworks, administrative data and stakeholder feedback. Countries use a spectrum of tools ranging from general-purpose AI applications to customised institutional systems, with system-level adoption remaining rare and largely contingent on strong governance, infrastructure and co‑ordination.
Artificial intelligence (AI) is increasingly used to develop and revise curricula and qualifications in vocational education and training (VET). While the adoption of AI remains at an early stage in most countries with a varying degree of advancement, its potential to enhance efficiency, responsiveness and evidence‑based decision making is increasingly recognised (Delia, 2025[1]; OECD, 2026[2]; OECD, 2025[3]; Moein et al., 2024[4]). AI can be deployed across all stages of the VET development process – from initial information processing to support consultation and labour market analysis, to the drafting, revision and format validation of curricula, qualifications and occupational standards, provided that barriers and potential risks are well managed (see Chapter 4).
AI use in the development of VET curricula and qualifications (thereafter, VET development) remains limited, uneven and largely exploratory. The results from the OECD Surveys conducted as part of this study (OECD, 2025[3]) indicate that 41% of VET policymakers, 36% of industry representatives and 21% of providers use or are aware of AI use in this context (Figure 3.1, Panel A). This suggests that experimentation is concentrated among central government agencies and industry partners than VET institutions, partly reflecting the fact that VET institutions focus more on teaching, training and programme delivery than development and updates of curricula and qualifications. Current applications are mostly supportive, including preparing draft competencies and learning outcomes, analysing job postings or other evidence on labour market demand, and mapping curricula and qualifications to detect overlaps and inconsistencies.
These figures are indicative and likely underestimated partly due to disclosure concerns, but broadly align with wider benchmarks. The 2023 OECD Digital Government Index highlights while some governments have been deploying a wide range of initiatives to enhance their capacity to use AI, implementation is still a challenge for most – 70% of countries had used AI to improve internal governmental processes, while only 33% had used AI to enhance policy design and implementation (OECD, 2025[5]). AI uptake in EU firms is around 42% (European Commission, 2020[6]), global business adoption ranges from 42% to 72% (IBM Watson, 2022[7]; McKinsey, 2024[8]), while generative AI use surged to 65% in 2024, up from 33% in 2023 (McKinsey, 2024[8]). While 36% of teachers used AI in their teaching in general according to 2024 OECD Teaching and Learning International Survey (OECD, 2025[9]), within VET early signs of growth are more evident: 65% of VET educators have tried AI for lesson planning, content creation and personalised learning (VET2Sustain, 2025[10]).
The OECD Surveys (OECD, 2025[3]) conducted as part of this study also show a wide country variation in AI use for VET development, ranging from 67% of respondents in Estonia to 10% in Ireland, although sample sizes in several countries are small. In England (United Kingdom), half of VET stakeholders reported using or being aware of AI in this context, followed by 44% in Lithuania and 43% in Germany and 39% in Korea (Figure 3.1, Panel B).
Percentage of respondents who use AI or are aware of the use in VET curriculum and qualification development
Note: The number of respondents is in bracket. In Panel B, only countries with more than five respondents are presented.
Source: OECD (2025[3]), Developing vocational qualifications and curricula with AI: Surveys and stakeholder interviews.
AI use among VET policymakers
Governments are increasingly leveraging and exploring AI to automate and tailor public services, improve decision making and support civil servants – including in education and skills sector (OECD, 2025[5]). Moreover, government access to vast amount of data that can be used as inputs for AI systems (OECD, 2025[5]). In VET, AI is being piloted to assist with curriculum and qualification development, underpinned by national digital and AI governance frameworks that enables sectoral agencies (including those in education and VET) benefit from the national standards and infrastructures and ensure security, interoperability and trust. For example (OECD, 2025[3]):
In Croatia, the Ministry of Labour, Pension System, Family and Social Policy (MROSP) has used AI in developing a Catalogue of Green and Digital Skills that applies across sectors, while primarily aiming to identify and fund in-demand skills, with a particular emphasis on digital and green transitions. The catalogue draws on competences extracted and classified from the Croatian Qualifications Framework (CROQF).
In the Czech Republic (hereafter ‘Czechia’), the National Pedagogical Institute (NPI) uses AI on a pilot basis to support VET development, particularly for mapping competencies in the National Register of Qualifications (NSK) and improving consistency in methodological materials. AI is applied across all VET sectors (ISCED levels 3‑5) for initial content mapping, drafting and refining methodological materials, translation, and ensuring alignment between curricula and qualification frameworks, always under human expert supervision. Current use remains in development and testing phases, with plans to expand towards more systematic AI‑supported content development while safeguarding quality and coherence.
In England (United Kingdom), Skills England/IfATE developed the SkillsCompass, which uses natural language processing (NLP) and generative AI to analyse labour market data and generate suggested content for qualifications and occupational standards. The AI-powered tool is being trialled to create duty statements for occupational standards and to update standards based on AI-detected labour market trends.
In Estonia, government-led initiatives such as the X-tee platform have set high standards for secure data exchange and interoperability across government agencies. These standards enable VET-related agencies such as the Estonia Qualification Agency (EQA) to pilot AI tools within protected national infrastructures and under strict human oversight (OECD, 2025[5]).
In Finland, since 2024, the Finnish National Agency for Education (EDUFI) has used AI on a voluntary basis to support the development of qualification units (modules; micro-credentials) that cater to the green and digital skills. These units outline key work tasks and associated skill requirements. The initiative was led by EDUFI’s content creators, who serve as preparers and experts of national qualification requirements, including competences and assessment criteria across fields. AI has supported the initial drafting of qualification units (modules) initially in digital and green transition areas and later also in fields such as social and healthcare. In 2025, EDUFI launched a project to scale up and systematise the use of AI tools in qualification development across VET sectors. The project helps to use AI more coherently in the preparation of qualification requirements, initial consultation to combine different data sources and to define the competences needed in different fields.
In Germany, the Federal Institute for Vocational Education and Training (BIBB) mirrors the Ministry of Justice’s cautious, security-focussed approach by using AI only in closed, controlled environments, prioritising data protection, human oversight, legal compliance and incremental adoption to ensure draft VET regulations meet national standards for legal compliance and public trust (OECD, 2026[11]).
AI use among industry partners
AI adoption in the world of work is outpacing education and training (Borgonovi et al., 2025[12]), creating both motivation and pressure for VET systems not only to adapt curricula and qualifications in line with evolving occupational demands, but also to reconsider how these curricula and qualifications are developed. Evidence from the OECD employer survey on the impact of AI in the workplace in 2022 (Lane, Williams and Broecke, 2023[13]) shows substantial AI uptake in industry: 42% of employers surveyed in the financial sector and 29% in the manufacturing sector report AI use, primarily in data analytics, fraud detection, production processes and maintenance. Larger employers and younger, male and more educated workers show higher uptake (Lane, Williams and Broecke, 2023[13]). Overall, almost 35% of workers in AI-using companies interact directly with AI, and 60% of employers report positive impacts: on productivity (80%), job performance (63%) and working conditions (Lane, Williams and Broecke, 2023[13]). These trends are echoed in EU-wide and global surveys, indicating that 42‑57% of companies use at least one AI technology (European Commission, 2020[6]; McKinsey, 2024[8]).
As AI transforms business processes to the emergence of new occupational profiles, integrating AI-driven labour market insights and employer feedback into the development and updates of qualification frameworks and curricula can help ensure that VET programmes remain relevant, responsive and future‑ready. Recent OECD work on firm-level AI adoption highlights the importance of the development and updates of qualification frameworks and curricula in the context of AI-related skills (OECD/BCG/INSEAD, 2025[14]).
In Estonia, an industry partner reports that LLMs have been used since 2024 in VET development, following improved model capability, but without systematic integration. While AI models (ML, LLM) are used mainly for background analysis, skills forecasting and efficiency support, they remain cautious about AI use beyond data gathering, analysis and process automation (OECD, 2025[3]).
Emerging cases include the Netherlands, where some industry partners have used AI-driven labour market data since 2022 to support market and skills analysis; and countries where AI use began more recently (2024‑2025) for topic exploration, presentation refinement, case‑study development for training exercises, and checking learner submissions for certification (Ireland), initial consultation and data‑finding support (Portugal), and language translation and formulating topics for training programme development (Lithuania) (OECD, 2025[3]).
In Switzerland, an industry partner reports that AI was first used to support the development of a five‑day training course, initiated by an industry association, and is applied in analysis phases for revising job profiles, such as comparing skills across professions. Other partners report institutional AI adoption from 2023‑2024, including licensed GPT and Copilot Chat, mainly for initial consultation and drafting (OECD, 2025[3]).
In the United Kingdom, an industry partner reports AI adoption since 2024, but with use largely limited to background policy research. Efforts are emerging to formalise usage through training, documentation and internal AI strategy, although constrained by time and resource limitations. For example, the company’s Head of Digital record individual uses, promote possible uses and offer training to those not yet using AI (OECD, 2025[3]).
AI use among VET providers
While overall AI use among VET providers in curriculum and qualification development remains limited and largely exploratory, VET teachers and curriculum developers are increasingly experimenting with AI to support tasks such as drafting curriculum documents, mapping competencies and creating learning activities. Some early‑moving providers report more advanced adoption. For example:
In England, interviewed VET providers have used AI since 2022 to support teaching planning and content development across a wide range of subject areas, driven by the availability of AI tools and led institutionally by Directors of Blended and Online Learning within established teaching and learning policies. AI use spans further and higher education provision, supporting technical, academic, employability, and personal development skills, mainly in drafting, planning, and learning‑resource development, with limited application so far in assessment. Other VET providers report more recent, organic adoption over the past 12 months to support curriculum design and resource creation in skills‑based programmes (including ESOL, English, maths, and digital skills), alongside the development of staff policies, learner guidance, and targeted staff training. Use remains voluntary, with uneven uptake due to varying staff confidence and limited capacity – particularly among part‑time staff.
In Ireland, private or community‑based providers use AI for drafting programme components or conducting initial labour market scans, while public Education and Training Boards (ETB) remain cautious and rely more heavily on traditional stakeholder consultation and formal QA processes. Some private VET providers began using AI in early 2024 as part of their digital transformation strategies to improve efficiency, personalisation, and labour‑market responsiveness. AI has been tested for labour market analysis, content creation, curriculum mapping, and learner resource development for QQI Level 5 and 6 (ISCED 3‑4) programmes (e.g. draft lesson plans, quizzes and assessment templates aligned with award specifications). Use has evolved from experimentation to a supportive tool embedded in curriculum planning and instructional design, with staff training and internal policy development alongside existing QQI quality-assurance agreements under which all AI-assisted outputs are reviewed and validated by qualified staff.
In the Netherlands, ROC Mondrian (a regional VET centre) is piloting multi‑agent AI systems to support curriculum building, including drafting competencies, structuring learning plans and aligning content with qualification dossiers (OECD, 2025[3]).
In Switzerland, several VET providers report AI use to support drafting exam materials, lesson plans or module descriptions, with some institutions collaborating across VET schools to share AI‑supported competency matrices and teaching resources during VET reforms, especially in ICT professions where technological change is rapid (OECD, 2025[3]). In the Canton of Zurich (a region in Switzerland), VET teachers organically began forming informal AI user networks as early as 2021, partly in response to the late publication of revised VET regulation for IT professions and associated module descriptions, which were essential for curriculum development. AI was used to accelerate the production and quality of curricula, competency mapping, lesson plans, assessments, and quality checks, mainly for school‑based components of full VET programmes and some modular content, especially in ICT and other fast‑changing technical fields. Only recently have some VET teachers been permitted to reclaim the costs for professional-grade AI accounts. However, this practice remains unstructured and lacks formal regulatory clarity from educational authorities or supportive institutional frameworks or clear cantonal policy direction.
VET providers in Croatia, Korea, Lithuania and Slovenia generally report limited, informal use. In Korea, VET teachers note that AI is used mainly for simple applications – such as drafting or summarising course materials – and often on an individual, ad‑hoc basis rather than as part of the institutional-level workflow. In Croatia and Lithuania, VET teachers occasionally use AI for idea generation, background searches or lesson preparation on an individual basis (OECD, 2025[3]).
Overall, AI enhances efficiency in data analysis, supports stakeholder co‑ordination and content generation, but does not replace expert-led decision making and high-stake validation. AI applications in VET development can be broadly categorised into three tasks or phases: input, process and output, as noted in Chapter 2. While these applications remain largely supportive and exploratory, they are increasingly shaping how VET systems analyse labour market needs, co‑ordinate stakeholder inputs and produce relevant content.
Input processing: Data and information gathering and analysis
AI is increasingly used to collect, structure and interpret large volumes of data and information to inform VET development. AI can reduce manual workload, accelerate the updating of qualifications in response to labour market change, and improve coherence within qualification frameworks and responsiveness to emerging skills demands. AI is also increasingly used to manage the complexity of regulatory frameworks that govern VET systems and to improve efficiency in regulatory monitoring to ensure accuracy and legal conformity.
Labour market analysis and forecasting with big data integration: AI processes and analyses large volumes of job advertisements, competency data, occupational standards and foresight studies to identify emerging skills and anticipate future skills needs, and inform curriculum and qualification updates and skills catalogues. The below examples illustrate AI’s potential to strengthen evidence‑based decision making and accelerate updates, provided expert oversight ensures quality and contextual relevance (OECD, 2025[3]).
In Croatia, a research lab at Algebra University has developed custom AI systems that can analyse large datasets, such as national education and employment records, including education pathway data that can be linked to curriculum data and employment outcomes, to identify mismatches between training provision and labour market outcomes, for creating relevant training programmes.
England’s SkillsCompass pilot uses natural language processing (NLP) and generative AI to define knowledge, skills and behaviours across more than 700 occupations based on real-time vacancy data, provided by Textkernel that also applies AI. England’s AI-driven analysis underpins initiatives such as the AI Skills Insight Report and Digital Skills Framework refresh, integrating vacancy data, foresight studies and labour market intelligence to provide a holistic view of current and emerging skills needs over a two‑ to five‑year horizon.
Estonia’s OSKA system (run by the Estonian Qualifications Authority [EQA], providing key labour market insights, since 2016) employs machine‑learning (AI) dashboards to provide interactive forecasts of workforce and competence needs, supporting curriculum designers to align VET programmes with projected demand. The EQA collaborates with Headai, an AI company, to apply Cognitive AI, Big Data, and Natural Language Processing to analyse job advertisements and curriculum data, producing real-time skills maps and predictive analytics to identify skills gaps and better align curricula with labour market needs (Headai, 2021[15]).
Germany’s Federal Institute for Vocational Education and Training (BIBB) applies big data analytics for skills forecasting and skills demand analysis to monitor changes in labour market demand and inform VET policy and curriculum development. In collaboration with research partners such as the Institute for Employment Research (IAB) and the Institute of Economic Structures Research (GWS), BIBB combines administrative data, labour market statistics and large‑scale analyses of job advertisements to identify emerging occupational trends, shifts in skill requirements and potential skills shortages. These analyses feed into long‑term qualification and occupational projections (e.g. the BIBB‑IAB “QuBe” projections) and are complemented by more granular tools, such as the Occupations and Skills Radar, which synthesises multiple data sources to provide timely, relevant insights on skills demand for policymakers, social partners and sector bodies.
In Mexico, AI is applied throughout the design and updating of upper secondary VET programmes. In particular, AI supports the development of “relevance and prospective studies (i.e. identifying emerging skills trends)” for technical degrees. AI-generated analysis is used as a reference to explore tendencies, identify emerging competencies, and double‑check information provided by the faculty, with all information validated by the experts (OECD, 2025[3]). In addition, data-driven platforms such as “shapingskills.mx” use NLP to analyse job descriptions and align qualifications with labour market demand (Caratozzolo et al., 2025[16]). Curriculum design co‑ordinators in General Directorate of Industrial and Technological Education (DEGETI) experimented with AI tools (e.g. Perplexity) during the creation of the AI technical career (AI and Embedded Software Systems) and cross-reference competencies and job descriptions where Mexican references were lacking.
Mapping competencies, curricula and qualifications: AI compares and clusters competencies across qualifications and standards to detect overlaps, inconsistences and gaps, supporting curriculum alignment with labour market needs and defined competencies. AI effectively supports large‑scale text classification and mapping tasks, though human oversight remains essential for quality assurance and contextual interpretation. For example (OECD, 2025[3]):
Croatia used AI-based text classifiers, trained on over 30 000 short texts from European and international databases, using machine learning models such as fine‑tuned Bidirectional Encoder Representations from Transformers (OECD, 2026[11]).. These models were further trained on national skills and competency data to classify over 15 000 competencies as “green” or “digital”, accelerating updates to skills catalogues. Croatia is also testing AI to harmonise overlapping competencies in its national qualifications registry and to improve alignment with employer feedback (OECD, 2026[11]).
England’s SkillsCompass uses NLP to support comparison of VET qualifications by automating extraction and semantic analysis of learning outcomes, standardising terminology and mapping curricula to occupational profiles.
In Estonia, the EQA piloted AI tools to automate the comparison of occupational standards and VET school curricula, including semantic matching of learning outcomes with industry skill requirements and national qualification levels.
In Korea, national agencies related to VET have begun exploring AI-driven mapping of competencies within the National Competency Standards. AI is used to analyse large datasets of occupational profiles and training modules, identifying redundancies and gaps across sectors, to support updates.
In Mexico, AI is used to compare competencies across qualifications and double‑check information and input provided by faculty.
In the Netherlands, AI is being tested to map learning outcomes across VET qualifications and detect inconsistencies within sectoral frameworks. Pilot projects focus on integrating AI with existing databases of occupational standards and job vacancy data to support the development of forward-looking curricula.
Regulatory and legislative monitoring: AI assists in monitoring complex and evolving legislative frameworks, summarising regulatory requirements and tracking updates to ensure compliance of VET curricula and qualifications that address skills needs, with national laws and sector-specific regulations. AI tools can automate the extraction, structuring and summarisation of legal texts, reducing manual workload and improving accuracy. For example, (OECD, 2025[3]):
In Estonia, the IT sector monitors 30‑40 legal acts and their updates, including the Information Society Services Act, Personal Data Protection Act, Cybersecurity Act, and the Occupational Qualifications Act as well as EU AI Act. AI-assisted tools are being piloted to monitor these acts, identify amendments and assess implications for occupational standards and training requirements. This approach helps maintain alignment between VET provision and a rapidly evolving legislative environment.
Germany uses AI to compare regulatory documents and draft training regulations, ensuring compliance with federal standards.
In Switzerland, discussions are underway on using AI to review examination regulations under the Federal Act on Vocational and Professional Education and Training, aiming to streamline approval processes while safeguarding legal compliance.
Streamlining process of stakeholder consultation and decision making
AI can support collaborative processes and co‑ordination among VET stakeholders, and streamline consultation processes, reducing time and cost in consultation for VET development and improving the quality of input from employers and other stakeholders in VET development and strengthening evidence used in decision making.
Facilitating stakeholder engagement: AI can analyse transcripts from employer and expert focus groups and generate synthetic data to simulate VET programme designs (OECD, 2025[3]).
In Croatia, Algebra University lab applies custom AI systems to process and analyse consultation data and generate synthetic scenarios for simulating and testing new training programmes, helping experts refine decisions.
In Estonia, co‑ordinators of occupational qualification councils have begun using AI to support working groups by drafting initial descriptions of competencies and standards, which are then refined through stakeholder consultation. This approach improves clarity and speeds up consensus-building.
In Nova Scotia (Canada), AI tools such as ChatGPT and Copilot are used during stakeholder workshops to prompt and spark discussion and fill gaps during the discussion in real-time, as a “back pocket” tool while facilitating a workshop for developing occupational standards with industry experts. These tools act as research assistants, helping experts focus on high-stake validation and decision making rather than administrative or repetitive tasks.
Accelerating VET development cycles: AI tools assist co‑ordinators in drafting curricula, qualifications and occupational standards and speeding up discussion during workshops by supporting more informed and prepared discussion, as described above (e.g. Estonia, Nova Scotia and Switzerland), shortening workshop duration and reducing costs (OECD, 2025[3]).
In Australia, Jobs and Skills Councils (JSCs) develop and maintain nationally accredited VET qualifications. JSCs are working to reflect the impacts of AI in their workforce planning analysis, and incorporate AI learning outcomes into the design and review of qualifications and training products. The Future Skills Organisation is the JSC for the finance, technology and business sectors and is undertaking exploratory research to create potential efficiencies in using AI in training package development processes enabling faster time to market and stronger industry alignment.
In Switzerland, professional organisations delegate technical and pedagogical tasks to expert bodies such as Swiss Federal University for Vocational Education and Training (SFUVET) and specialised consultancies in VET design, or cantonal education departments with technical capacity, to reduce technical and administrative burdens and ensure methodological rigour. These bodies are exploring AI to accelerate documentation and increase efficiency in stakeholder consultation process. Discussions are underway on using AI to reduce the duration and cost of workshops organised for employers to identify skills needs and learning outcomes in the development of occupational profiles and training plans.
Supporting evidence‑based decision making: AI is increasingly used to strengthen evidence‑based decision making in VET development. Semantic matching – enabling comparing and aligning concepts, terms or text based on their meaning, even when skills are described differently – and predictive analytics help integrate diverse data sources such as occupational standards, job vacancy data and foresight studies, into actionable insights. These tools do not replace expert judgement but provide an analytical foundation for strategic planning. For example (OECD, 2025[3]):
In England, the SkillsCompass supports decisions on updating qualifications and signalling new competencies to awarding bodies and training providers.
In Estonia, labour market data with machine learning-based forecasting dashboards enable policymakers and curriculum designers to align VET programmes with projected demand, ensuring responsiveness to technological and structural changes in the economy.
In Germany, AI-driven analytics are integrated into skills demand studies conducted by BIBB. These analyses inform the revision of training regulations and help prioritise occupations for modernisation, particularly in sectors undergoing rapid digital transformation.
Output tasks: Content generation and systemic improvement
AI supports producing and refining the outputs of VET curricula, qualifications and occupational standards.
Drafting and revising qualification and curriculum descriptions and content (OECD, 2025[3]):
Croatia’s Academic and Research Network (CARNET) pilot used generative AI (e.g. ChatGPT, DeepSeek) to create non-obligatory AI curriculum (specific to VET) in two months instead of a year, combining expert input with AI suggesting competencies, activities and benchmarks.
In Estonia, AI was used to draft descriptions for occupational qualification standards and VET curricula. For instance, VET expert working groups drafted 13 new upper secondary VET curricula, improving clarity and speed.
In Germany, the BIBB-led KINO project (KI in Neuordnungsverfahren) tests AI’s role in drafting training regulations using a secure, closed-source adaptation of ChatGPT. While AI performed relatively well in administrative tasks such as document comparison, its performance in generating regulatory content was limited, reaffirming the central role of human experts and VET stakeholders in curriculum and qualification design (OECD, 2025[3]).
AI also plays a role in assessment design within qualification and curriculum development in New Zealand. The AI-Generated Assessments Project, led by ConCOVE Tūhura (Centre of Vocational Excellence for Construction & Infrastructure), uses AI for VET assessment design to produce high-quality, personalised assessments that address diverse learner and industry needs while meeting required standards. The project focusses on developing AI-generated assessments for the Trades Essentials micro-credential and on establishing an ethical framework to ensure safety, fairness and transparency in AI-assisted assessment generation (ConCOVE, 2025[17]).
In Nova Scotia (Canada), AI is being experimentally integrated into the development of occupational standards and VET curricula, particularly within the Red Seal and provincial trades systems. AI tools such as ChatGPT and Copilot are used to generate performance criteria, and review existing standards for updates. These tools accelerate the drafting process, before the final content is validated by subject matter experts.
Modernising frameworks and classification systems: AI-driven tools help modernise occupational classifications, merge redundant profiles and align with other classifications and standards such as ESCO (OECD, 2025[3]).
In Croatia, AI is applied in the modernisation of the National Classification of Occupations (NKZ10). In 2024‑2025, the working group for the Classification integrated advanced AI methods to improve the accuracy, coherence and interoperability of occupational classifications. Using the LangChain ecosystem, OpenAI models and the Chroma vector database, the group applied large language models (LLMs) to generate occupational descriptions, perform semantic grouping and map Croatian occupations to European standards (European Skills, Competences, Qualifications and Occupations, ESCO). This process enabled the merging of redundant or overlapping occupations, streamlined the classification system, and facilitated international comparability. The successful implementation of these AI methods resulted in significant time and resource savings, while reinforcing the need for human validation and expert review at each stage (OECD, 2026[11]).
Improving relevance, efficiency, accuracy, consistency, coherence and interoperability (OECD, 2025[3]):
In Czechia, VET providers have begun leveraging AI primarily as a support tool for linguistic and content adaptation tasks such as translating professional content from source languages into Czech, ensuring technical terminology is accurately rendered for local contexts.
In England, AI pilots such as the duties (tasks) generator and linguistic matching of technical qualifications (TQ) submissions to knowledge, skills and behaviours (KSBs), as well as Standard Occupational Classification (SOC) code assignment, support alignment with national frameworks and language accessibility and inclusivity.
Estonia plans to use AI to automate the comparison of occupational standards and school curricula, particularly for micro-credentials, as well as learning outcomes with skills needs and qualifications, to reduce manual workload and accelerate approval.
Korea is experimenting with AI for format and structural quality checks of NCS documents.
In the Netherlands, the Organisation for the Co‑operation between VET and the Labour Market (SBB) has developed “SBB Chat” to detect overlaps and automate operational tasks within secure environments, despite broader government restrictions on AI use.
AI is used to integrate green, digital and AI skills into VET
AI and digitalisation are reshaping the skills needed in the labour market, driving changes in occupations, qualifications and VET curricula. To stay relevant, new qualifications and curricula must include AI and other digital skills, with greater flexibility to adapt to evolving competence needs. Nearly all of the 29 OECD countries that participated in an OECD study on this topic have national rules and guidelines on developing students’ digital competences, including in VET (Figure 3.2). Countries vary in how they integrate digital competences – whether through national or local curricula, as transversal skills or distinct subjects (OECD, 2023[18]).
Examples of integration of digital technologies and sustainability in VET and education-wide
A snapshot of the rules and guidelines on integrating digital competences in the curriculum
Source: OECD (2023[18]), OECD Digital Education Outlook 2023: Towards an Effective Digital Education Ecosystem, https://doi.org/10.1787/c74f03de-en.
In this context, countries are advancing AI use in VET development in programmes targeting green, digital and AI‑related skills, reflecting growing demand for new competencies linked to digitalisation and the green transition. Across countries, AI is being explored to support skills anticipation and curriculum development and updates. Several examples illustrate these developments (OECD, 2025[3]):
Croatia has used AI to support the development of a green and digital skills catalogue, while Algebra University applies AI systematically in curriculum design, labour market analysis, stakeholder consultation and synthetic data generation for digital and green skills training including micro-credentials and modular adult education, supported by in‑house cloud infrastructure and custom‑trained models.
Estonia places strong emphasis on digital and green skills, with AI piloted for VET curriculum drafting, comparison and the development of non-regulatory competence profiles in the ICT sector (e.g. data analyst, data steward, data engineer).
Finland is advancing the integration of green and digital skills into VET, with AI being explored for curriculum development and skills forecasting. While systematic AI use is still limited, pilot projects are underway to use AI for analysing labour market trends and aligning VET offerings with emerging green and digital occupations, particularly through micro-credentials (OECD, 2025[19]).
In England, AI-driven insights have already led to the revision of several occupational standards, for example Digital Device Repair Technician (Level 3) and Mechatronics Maintenance Technician (Level 3).
AI is used to develop micro-credentials and modular VET
Many countries are in transition from rigid qualification standards to more flexible, competence‑based frameworks (OECD, 2025[24]), reflected in the growing use of micro-credentials and modular VET. In parallel, countries are increasingly exploring AI use to support the development of micro‑credentials and modular VET to enable more flexible and responsive learning pathways aligned with evolving labour market demand, including digitalisation and the green transition. Although integration remains uneven and largely experimental, several countries are piloting AI to structure modular curricula and support the design of micro‑credentials (OECD, 2025[3]):
Estonia is advancing to competence‑based and modular VET systems, and piloting AI for developing, drafting and revising modules and micro‑credential proposals, and comparing proposed modules or micro‑credentials against existing occupational standards and qualification‑framework level descriptors. Estonia’s VET curricula are structured around modules derived from occupational standards and VET schools can adapt up to 30% of national curricula, allowing for local flexibility. Hundreds of micro-credentials have been submitted, triggering the need for more efficient approval processes.
In Switzerland, the ICT sector is developing modular and competence‑based qualifications and examinations (mainly higher VET), using custom GPT tools tailored for each profession. AI‑generated content is systematically validated against the modular building block system (Modulbaukasten platform).
In England, some further education colleges use AI tools such as TeachMatic, ChatGPT and Copilot for curriculum co-design, teaching resource creation, synthesising stakeholder feedback and enhancing modular content relevance, for bespoke packages and modular sessions.
In Finland, the integration of emerging skills through modularisation and micro‑credentials is a strategic priority, and VET providers are experimenting with AI to support modular curriculum design and the development of micro‑credentials, particularly in technology and sustainability‑related fields.
Overall, these cases show that AI is beginning to function as an enabling tool for modularisation and micro‑credentials in VET, but they also highlight persistent barriers to scaling, including data quality, privacy concerns, limited technical capacity and the absence of clear governance frameworks.
The availability, scope and quality of data strongly shape how much insight could be gathered and generated to support VET curriculum and qualification development using AI. Countries with more granular information are better positioned to apply AI and advanced analytics to generate useful insights for VET development; where data are limited or fragmented, AI use tends to focus on narrower tasks or likely be more biased. Increasingly, private companies and platforms such as Lightcast or OpenSyllabus offer big‑data analytics or AI techniques to fill these gaps by analysing online job vacancy data and course/training descriptions, which could inform and support evidence‑based curriculum planning and provision.
Before using AI in VET development, stakeholders mainly used labour market data (e.g. job postings, skills demand and sectoral trends), national and regional qualification frameworks and standards, administrative and performance data, existing curricula and competence matrices, as well as qualitative feedback from employers, teachers and learners (Table 3.2). International benchmarks and research were also used to contextualise national developments.
Since the emergence and use of AI, VET stakeholders increasingly draw on a wide range of data inputs to underpin AI‑supported VET development. While the combination of these data sources enables richer analysis, their availability, interoperability and legal accessibility vary considerably across systems, influencing both the ambition and the maturity of AI applications.
Source: OECD (2025[3]), Developing vocational qualifications and curricula with AI: Surveys and stakeholder interviews.
In practice, AI tools used or piloted range from widely available general‑purpose applications to more advanced institutional solutions. Several countries including Mexico report using tools such as ChatGPT, Copilot, Gemini or similar systems to synthesise large volumes of publicly available information – such as job descriptions, qualification frameworks, employment databases or academic literature – and to support tasks like drafting, classification or document review. In Finland, for example, AI supports when EDUFI qualification experts analyse, classify and structure relevant administrative and sectoral data in collaboration with relevant stakeholders (e.g. eRequirements, EDUFI publications and data of the new competence needs in different occupations), although access to large, machine‑readable datasets still remains limited. Estonia is moving towards making learning outcomes machine‑readable, with the aim of enabling more systematic AI‑supported alignment between VET curricula and labour market needs.
More broadly, the type of AI tools adopted reflects different levels of organisational maturity, capacity and governance. Basic use often relies on free or open‑source tools applied at the individual stakeholder or classroom level (Figure 3.3, A), while more advanced stages involve customised or licensed solutions embedded at institutional and sectoral level (Figure 3.3, B and C). Several examples show the development of bespoke systems tailored to specific organisational or sectoral needs and requirements (Figure 3.3, B), requiring greater investment of resources, infrastructure and governance capacity. Examples include: England’s SkillsCompass, ICT sector in Switzerland, SBB Chat in the Netherlands by the Organisation for the Co‑operation between VET and the Labour Market (SBB), and Algebra University in Croatia.
Faster and more cost-effective than bespoke could be customising the existing AI platforms (Figure 3.3, C) to meet specific needs and add modifications. For example, Korea’s HRDK customised and trained OpenAI GPT models exclusively on NCS and HRDK data with unique architecture and features designed only for a specific task.
Cross‑institutional or system‑level adoption of AI remains rare (Figure 3.3, D), but emerging initiatives illustrate its potential where data availability, legal frameworks and stakeholder co‑ordination are sufficiently developed. For example, in the Netherlands, recent initiatives such as EduGenAI and CompetentAI point towards a more systemic approach, aiming to provide a network of VET institutions with a shared, secure environment for using generative AI responsibly in curriculum design, skills analysis and educational development (OECD, 2026[11]). Rather than relying on individual or ad hoc uses of general‑purpose AI tools, these initiatives focus on creating common infrastructures and governance frameworks, enabling institutions to access AI functionalities while addressing concerns related to data protection, transparency and consistency. By embedding AI within a controlled, cross‑institutional setting, the Dutch approach illustrates how AI adoption can move beyond isolated pilots towards more co‑ordinated use across the VET system, although these initiatives remain at an early stage and their long‑term impact is yet to be assessed.
Examples, by degree of control and boundary of AI systems used or piloted for VET development
Note: “Customised” refers to the use of a standard solution such as OpenAI GPT or similar, therefore based on an existing product or framework, then adapted to meet specific needs and add modifications (fine‑tuning, additional features or integration with internal systems). For detailed examples, see OECD (2026[11]), Ten Country Case Studies on VET Development and AI Use, www.oecd.org/content/dam/oecd/en/publications/support-materials/2026/06/developing-vocational-education-and-training-with-artificial-intelligence_e9f76b4e/Ten-country-case-studies-on-VET-development-and-AI-use.
Source: Author’s elaboration based on OECD (2025[3]), Developing vocational qualifications and curricula with AI: Surveys and stakeholder interviews.
References
[12] Borgonovi, F. et al. (2025), “AI adoption in the education system: International insights and policy considerations for Italy”, OECD Artificial Intelligence Papers, No. 52, OECD Publishing, Paris/Fondazione Agnelli, Turin, https://doi.org/10.1787/69bd0a4a-en.
[23] Borgonovi, F. et al. (2023), “Emerging trends in AI skill demand across 14 OECD countries”, OECD Artificial Intelligence Papers, No. 2, OECD Publishing, Paris, https://doi.org/10.1787/7c691b9a-en.
[16] Caratozzolo, P. et al. (2025), Discover tomorrow, shape today: shaping skills report, https://hdl.handle.net/11285/703615.
[17] ConCOVE (2025), AI-Generated assessments for VET, https://www.manukau.ac.nz/wp-content/uploads/2025/10/AI-Final-Report.pdf.
[1] Delia, C. (2025), European insights: adoption of AI in TVET institutions – challenges, opportunities and recommendations, UNEVOC Research Briefs, https://atlas.unevoc.unesco.org/research-briefs/european-insights-adoption-of-ai-in-tvet-institutions-challenges-opportunities-and-recommendations.
[6] European Commission (2020), European enterprise survey on the use of technologies based on artificial intelligence, https://ec.europa.eu/newsroom/dae/document.cfm?doc_id=68488.
[15] Headai (2021), Forecasting Estonian future skill needs, https://wp.headai.com/forecasting-estonian-future-skill-needs/.
[21] Huismann, A. and U. Hippach-Schneider (2024), Implementing European priorities in VET: making national VET agile, flexible, innovative, attractive, inclusive and quality-assured: Germany., Cedefop Refernet thematic perspectives, https://www.cedefop.europa.eu/files/implementing_eu_priorities_germany_2024_2.pdf.
[7] IBM Watson (2022), IBM’s Global AI Adoption Index 2022, https://www.oecd.org/content/dam/oecd/en/publications/reports/2023/03/the-impact-of-ai-on-the-workplace-main-findings-from-the-oecd-ai-surveys-of-employers-and-workers_ad686e91/ea0a0fe1-en.pdf.
[20] Korkala, S. and M. Vasenius (2024), Implementing European priorities in VET: making national VET agile, flexible, innovative, attractive, inclusive and quality-assured: Finland. Cedefop Refernet thematic perspectives, https://www.cedefop.europa.eu/en/country-reports/implementing-european-priorities-vet-finland-2024.
[13] Lane, M., M. Williams and S. Broecke (2023), “The impact of AI on the workplace: Main findings from the OECD AI surveys of employers and workers”, OECD Social, Employment and Migration Working Papers, No. 288, OECD Publishing, Paris, https://doi.org/10.1787/ea0a0fe1-en.
[8] McKinsey (2024), The state of AI in early 2024: Gen AI adoption spikes and starts to generate value, https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-2024.
[4] Moein, M. et al. (2024), “Beyond Search Engines: Can Large Language Models Improve Curriculum Development?”, in Lecture Notes in Computer Science, Technology Enhanced Learning for Inclusive and Equitable Quality Education, Springer Nature Switzerland, Cham, https://doi.org/10.1007/978-3-031-72312-4_17.
[2] OECD (2026), OECD Digital Education Outlook 2026: Exploring Effective Uses of Generative AI in Education, OECD Publishing, Paris, https://doi.org/10.1787/062a7394-en.
[11] OECD (2026), Ten Country Case Studies on VET Development and AI Use, OECD, Paris, https://www.oecd.org/content/dam/oecd/en/publications/support-materials/2026/06/developing-vocational-education-and-training-with-artificial-intelligence_e9f76b4e/Ten-country-case-studies-on-VET-development-and-AI-use.
[3] OECD (2025), Developing vocational qualifications and curricula with AI: Surveys and stakeholder interviews, Unpublished.
[24] OECD (2025), Empowering the Workforce in the Context of a Skills-First Approach, OECD Skills Studies, OECD Publishing, Paris, https://doi.org/10.1787/345b6528-en.
[5] OECD (2025), Governing with Artificial Intelligence: The State of Play and Way Forward in Core Government Functions, OECD Publishing, Paris, https://doi.org/10.1787/795de142-en.
[9] OECD (2025), Results from TALIS 2024: The State of Teaching, TALIS, OECD Publishing, Paris, https://doi.org/10.1787/90df6235-en.
[19] OECD (2025), Vocational Education and Training and the Green Transition in Finland, OECD Reviews of Vocational Education and Training, OECD Publishing, Paris, https://doi.org/10.1787/4d29a34a-en.
[18] OECD (2023), OECD Digital Education Outlook 2023: Towards an Effective Digital Education Ecosystem, OECD Publishing, Paris, https://doi.org/10.1787/c74f03de-en.
[22] OECD (2023), OECD Employment Outlook 2023: Artificial Intelligence and the Labour Market, OECD Publishing, Paris, https://doi.org/10.1787/08785bba-en.
[14] OECD/BCG/INSEAD (2025), The Adoption of Artificial Intelligence in Firms: New Evidence for Policymaking, OECD Publishing, Paris, https://doi.org/10.1787/f9ef33c3-en.
[10] VET2Sustain (2025), VET2Sustain survey current and future use of AI and XR in VET, https://vet2sustain.eu/wp-content/uploads/2025/05/Article-ENG-VET2Sustain-Survey-Findings.pdf.