An operational framework for AI literacy in the workplace

Executive summary

Organisations in the European Union (EU) face a legal obligation to address AI literacy. Article 4 of the EU AI Act began applying on 2 February 2025 and originally required providers and deployers of artificial intelligence (AI) systems to take measures to ensure, to their best extent, "a sufficient level of AI literacy" among relevant staff and other persons operating AI systems on their behalf. Regulation (EU) 2026/1744, which entered into force on 27 July 2026, reframed this obligation. Providers and deployers must now take measures to "support the development" of AI literacy, taking into account relevant knowledge, experience, education, and training; the context in which AI systems are used; and the persons or groups affected by their use.

This amendment places organisational measures at the centre of Article 4 implementation. Providers and deployers of AI must determine which measures are appropriate for their personnel, systems and contexts; and establish a reasoned basis for those choices. This requires organisations to connect AI literacy initiatives to actual patterns of use, levels of responsibility and potential consequences.

Existing European instruments provide a strong foundation for this work. DigComp 3.0 integrates AI across its digital competence areas (Cosgrove and Cachia, 2025), while the OECD-European Union AILit framework defines AI literacy for primary and secondary education (OECD and European Union, 2026). Together, these instruments describe important areas of knowledge and competence. Workplace implementation now requires a complementary method for translating this content into measures suited to different organisational roles.

This brief proposes a three-tier framework based on the nature and consequences of an individual's interaction with AI. The framework helps organisations define proportionate literacy objectives, select appropriate measures, document their rationale and evaluate effectiveness. It supports implementation of Article 4 while advancing responsible AI adoption, workforce development and professional mobility.

This framework is keyed to how much and how consequentially the workers interact with AI. Instead of looking at peoples’ job titles, seniority or technical background, the framework proposes the following three tiers:

TIER	POPULATION	PROPORTIONATE LITERACY OBJECTIVE	MEASUREMENT
L0 Baseline	Personnel working within the organisation's remit: relevant staff, plus contractors and service providers	General awareness of what AI is, which systems the organisation uses and which risks can arise from their operation, including fabricated outputs, bias, data exposure and privacy, together with the appropriate escalation channels. Content is available from DigComp 3.0 and foundational programmes.	Record of awareness training or equivalent measures
L1 Operational	For workers who are required to use AI systems in their role	Capabilities, limits and failure modes of the systems they use; how to interpret and verify outputs; effective prompting and querying; disclosure and copyright duties; when to escalate; relevant legal and ethical obligations.	System-specific training records plus a written rationale linking training to roles and risks
L2 Oversight	Persons assigned to operate or oversee high-risk AI systems	Deep system-specific competence: monitoring, recognizing malfunction and drift, detecting bias, knowing when and how to intervene or override, and distinguishing high-risk from prohibited uses.	Documented competence assessment and oversight assignment, reviewed at least annually or when the system changes

Table 1. The tiers in operational terms. Tiers are cumulative.

This framework can be applied across sectors and accommodate the diversity of individuals and roles that now interact with AI technologies daily. It can help compare programs across borders, and provide smaller firms a baseline to build on. A common framework and understanding of AI literacy will also support workforce development and mobility: employees could carry these recognized competences between employers and build on them in different roles, while employers would have a shared scale to develop them. We recommend not introducing a certification for AI literacy yet, and instead focus on creating a validated measurement first, as a certification introduced at this stage could run the risk of becoming a box-ticking exercise and encouraging untested courses.

Defining the scope and requirements of the AI literacy duty

The European Union is the first jurisdiction to establish a binding, economy-wide AI literacy duty for organisations that provide or deploy AI systems. Other jurisdictions have developed national programmes, workforce initiatives and sector-specific training requirements. The EU approach goes further by placing a direct obligation on providers and deployers across organisational contexts and AI-system categories.

The AI Act defines AI literacy as the "skills, knowledge and understanding" that allow providers, deployers and affected persons to make an informed deployment of AI systems and to gain awareness of the opportunities, risks and possible harms involved (Regulation (EU) 2024/1689, Article 3(56)). Its Article 4 began applying on 2 February 2025 (Regulation (EU) 2024/1689, Article 4), requiring providers and deployers to take measures to ensure, to their best extent, a "sufficient" level of AI literacy among staff and other persons dealing with the operation and use of AI systems on their behalf. Organisations were required to consider those persons' technical knowledge, experience, education and training, together with the context in which the systems would be used.

However, Regulation (EU) 2026/1744, the Digital Omnibus on AI, amended Article 4 with effect from 27 July 2026. Under the revised provision, providers and deployers must take "measures to support the development of AI literacy" among the same population. The selection of those measures must still reflect personnel knowledge and experience, education and training, the context of use, and the persons or groups on whom the AI systems are used.

The amendment focuses the obligation on organisational action. Compliance now depends on adopting measures that are proportionate to the relevant systems, roles and circumstances. While this approach gives providers and deployers flexibility, it still requires them to make informed choices about the content, depth and delivery of their AI literacy initiatives.

The European Commission supports implementation through its AI literacy questions and answers and its living repository of organisational practices (European Commission, 2026a; 2026b). These resources identify a range of possible measures, including training, guidance, awareness initiatives and other interventions adapted to organisational circumstances. They also illustrate the diversity of approaches already being developed across sectors and organisations.

The remaining operational challenge is to connect these measures consistently to the people and systems within an organisation. Providers and deployers need a practical method for identifying relevant personnel, assessing the nature of their interaction with AI, selecting suitable literacy objectives and recording the reasoning behind those choices. A shared framework can also help organisations evaluate programmes, compare approaches and improve them over time.

The framework proposed in this brief responds to that need. It translates the contextual factors contained in Article 4 into three cumulative tiers. Each tier connects a category of personnel and AI interaction to a defined literacy objective, a set of appropriate measures and a corresponding form of documentation.

This structured approach also supports broader organisational objectives. Personnel who understand the capabilities, limitations and risks of the systems they use are better equipped to verify outputs, recognise potential harms, identify appropriate applications and escalate concerns. Effective AI literacy, therefore, contributes to compliance, risk management, innovation and workforce development.

International approaches to AI literacy initiatives

Other governments’ approaches range from creating tailored education programs and initiatives starting as early as primary school, to private partnerships to train relevant people in the workforce. There has been little focus on defining AI literacy as a concept, and more on preparing people, young and old, for AI adoption.

The United States issued an executive order to “promote AI literacy and proficiency” through schooling, teacher training, and apprenticeships. This pipeline approach develops AI skills from education to early career stages but does not require employers to participate (The White House, 2025). On the corporate front, the NIST framework never uses the term “AI literacy” at all, asking only that personnel receive risk training relevant to their duties (NIST, 2023).

The United Kingdom combined its AI Opportunities Action Plan with partnerships involving Google, Microsoft, and IBM to train 10 million workers by 2030, a target raised from 7.5 million when the programme expanded in January 2026. It also introduced the Skills England benchmark to specify training content, and followed it with a free AI Skills Hub that hosts short, industry-developed courses assessed against that benchmark. Those who complete these courses earn a government-backed digital badge that signals their competence to any employer (UK Government, 2026; Skills England, 2026; AI Skills Hub, n.d.).

China has taken the most directive approach, starting with educational institutions. In May 2025, its Ministry of Education required schools to integrate AI instruction into primary and secondary education as a tiered, progressive system. Already in the autumn term, cities began making these courses compulsory, with Beijing alone enrolling about 1.8 million students (Ministry of Education of China, 2025).

We see a pattern emerging around different approaches to AI literacy. Some governments are focussing on mass courses and citizen wide educational programs to form a baseline, while others are partnering with industry to benchmark training content. The EU, however, remains the only region with binding AI literacy duties, and supports implementation through its AI literacy questions and answers and its living repository of practices (European Commission, 2026a; 2026b). The OECD and Commission AILit framework defines AI literacy for students and teachers in school education rather than the general workforce (OECD and European Union, 2026; see also UNESCO, 2024, for comparable school-focused frameworks).

Corporate adoption outpaces evidence-based guidance

Organisations are adopting AI systems rapidly, and workers increasingly interact with these technologies in their daily responsibilities. Eurostat found that 20 per cent of EU enterprises with ten or more employees used AI in 2025, compared with 13.5 per cent in 2024. Adoption reached 55 per cent among large enterprises (Eurostat, 2025). A 2026 Microsoft survey of more than 20,000 knowledge workers using AI across ten markets found that 58 per cent were producing work they could not have produced a year earlier (Microsoft, 2026). The World Economic Forum's 2025 employer survey reported that 77 per cent of firms planned to reskill or upskill employees to work alongside AI, and ranked AI and big data among the fastest-growing skill areas worldwide (World Economic Forum, 2025).

AI adoption also varies considerably across the Union, ranging from 42 per cent of enterprises in Denmark to 5.2 per cent in Romania (Eurostat, 2025). Organisations therefore approach Article 4 with different levels of technical capacity, institutional maturity and workforce experience. The Commission's repository of AI literacy practices reflects this diversity and documents initiatives tailored to different roles and organisational contexts, including developers, administrators and legal staff (European Commission, 2026b).

As investment in workplace AI training grows, three areas require particular attention. The first is alignment. Corporate training often prioritises productivity and tool use, while responsible implementation also requires personnel to understand limitations, failure modes, potential harms and applicable obligations. The second is quality. A common benchmark would help organisations compare programmes, assess providers and establish consistent expectations. The third is measurement. Validated assessment methods would enable organisations to determine whether their measures improve understanding and support responsible behaviour. The most recent systematic review of AI literacy scales found that the validated instruments available were built for students and general populations, that few were tested beyond basic validity, and that none were tested across cultures (Lintner, 2024).

These priorities point towards a shared method for planning and evaluating AI literacy measures. Such a method should include a common tier logic for determining the appropriate depth of learning, a benchmark describing the content relevant to each tier, and an agreed approach for assessing whether interventions improve understanding.

A common vocabulary would support implementation across organisations, borders and supply chains. It would also support workforce mobility by allowing employees to build recognised competencies as their roles develop. Employers would gain a shared structure for professional development, procurement and quality assurance.

Measurement should precede certification. Validated assessment tools can establish a credible foundation for future certification by demonstrating what programmes teach, what participants learn and how competencies transfer across roles and organisations.

A three-tier framework for AI literacy

Operationalising Article 4 requires organisations to determine which AI literacy measures are proportionate to individuals' interactions with AI and the associated risks. The provision points towards a contextual approach based on personnel characteristics, the systems being used, the circumstances of use and the people potentially affected. A tiered framework translates these considerations into practical organisational decisions.

A similar challenge arose in the broader AI talent discourse, where "AI talent" was traditionally treated as a single category. Dividing the workforce into tiers based on technical depth, from non-technical roles requiring AI familiarity through software development to deep learning specialisation, made national capabilities more measurable and comparable across 31 countries (Pal, Schneider and Lazzaroni, 2025). AI literacy benefits from a similar differentiation.

The term "AI literacy" encompasses a wide range of people and activities, from general workplace awareness to the operation and oversight of consequential systems. A schoolchild learning about algorithms, an accountant using a chatbot and an operator overseeing a credit-scoring model each require different knowledge and judgement. A tiered structure accommodates this diversity by linking literacy objectives to patterns of interaction and responsibility.

The proposed framework assigns AI literacy objectives according to professional roles and patterns of AI use. Individuals can move between tiers as their responsibilities, systems or contexts change. Its structure reflects the factors identified in Article 4 and incorporates the consequences and risks associated with different forms of interaction.

The framework contains three cumulative tiers: a baseline tier for personnel working within the organisation's remit, an operational tier for users of specific AI systems, and an oversight tier for personnel operating or overseeing high-risk systems. Each tier specifies the relevant population, literacy objective, implementation measures and documentation.

Figure 1. Three cumulative tiers of AI literacy. An individual's tier follows their role, not their seniority or technical background.

This figure describes the depth of knowledge needed by different workers in an organisation based on their level and scope of interaction with AI. Another dimension to this framework is the content of the knowledge needed at each of the tiers. This can be supplied by existing EU instruments that define AI literacy: DigComp 3.0's 21 competences across five areas, now with AI integrated transversally and each statement labelled AI-explicit or AI-implicit (Cosgrove and Cachia, 2025); and the four domains of the OECD and Commission AILit framework (OECD and European Union, 2026).

Grounding each tier in established instruments

The three tiers draw on established European instruments and emerging workplace standards. This grounding promotes coherence across organisations and Member States while allowing implementation to reflect different systems, roles and contexts.

L0 aligns with foundational competencies in DigComp 3.0 and with programmes such as Finland's Elements of AI. This free course, launched in 2018 to teach foundational AI concepts, is offered across the EU with the objective of reaching 1 per cent of citizens and has attracted more than two million enrolments worldwide (Elements of AI, n.d.; Finnish Government, 2019). These instruments provide accessible content on AI concepts, information evaluation, personal data, privacy and responsible use.

L1 incorporates the role and context-specific factors identified in Article 4. It also draws on workplace standards such as the Skills England AI foundation skills benchmark, which organises skills across technical, non-technical and responsible-use domains (Skills England, 2026), and on Skills England's analysis of AI skill needs across ten growth sectors, which gives employers a wider framework and worked use cases for applying those skills (Skills England, 2025).

L2 connects Article 4 implementation to the broader governance of high-risk AI systems. It complements requirements regarding operator competence, human oversight, monitoring and intervention (Regulation (EU) 2024/1689, Articles 14 and 26). ISO/IEC 42001 also provides a relevant organisational precedent by emphasising competence, documentation and continual improvement within AI management systems (ISO/IEC, 2023).

A transparency-only system, such as a chatbot or content generator, is classified at L1 when it becomes integral to an individual's job responsibilities; and incidental use remains at the L0 baseline. Understanding that outputs are fallible and subject to hallucinations is inherent to the L0 baseline, as is a familiarity with ethics and appropriate use cases, both generalised and firm-specific. The use or procurement of tools that are deemed high-risk by the AI Act, such as recruitment screening or creditworthiness assessment systems, is classified at L2, as each requires human oversight and operators. In accordance with the AI Act, the more a system is relied on, the more training is required.

Together, these instruments supply much of the content required for workplace AI literacy. The tiered framework provides the structure for assigning that content to relevant personnel and translating it into organisational measures.

Policy recommendations

For organisations, the framework translates Article 4 into a structured five-step process:

1. It enables organisations to demonstrate a reasoned connection between personnel, systems, risks and literacy measures. It also creates a foundation for evaluating effectiveness and improving programmes over time.

2. It allows for assessment to be a part of this implementation process. Completion records could show that a measure occurred, while proportionate assessments could provide evidence about whether understanding improved.

3. Organisations can use these findings to refine content, identify gaps and determine whether additional measures are appropriate.

4. Organisations operating or overseeing high-risk systems should coordinate their AI literacy measures with the applicable requirements concerning competence, training and human oversight (Regulation (EU) 2024/1689, Articles 14 and 26). Aligning these activities can reduce duplication and create a coherent record of organisational preparedness.

5. Organisations procuring training can use the tiers to define requirements, compare providers and evaluate whether proposed programmes match the roles and systems involved. This would help establish a quality floor while preserving flexibility in delivery.

For policymakers, the priority is to develop benchmarks, practical examples and evaluation tools that help organisations select proportionate measures and assess their effectiveness. European and national authorities should coordinate this work with industry, training providers, standards bodies and research organisations, including agreement on which national body owns AI literacy.

A European benchmark could specify core content for each tier while allowing organisations to adapt delivery to their circumstances. A validated assessment framework could then measure changes in knowledge and judgement across roles, sectors and Member States. Together, these instruments would strengthen implementation, comparability and workforce mobility.

The principal building blocks are already available. Article 4 establishes a contextual organisational duty (Regulation (EU) 2024/1689; Regulation (EU) 2026/1744); DigComp 3.0 and AILit supply established competence frameworks (Cosgrove and Cachia, 2025; OECD and European Union, 2026); Skills England demonstrates how public authorities can benchmark workplace content (Skills England, 2026); and ISO/IEC 42001 provides an organisational model for recording competence (ISO/IEC, 2023). The proposed tiers connect these elements through a common implementation structure.

References

AI Skills Hub. (n.d.). AI Skills Boost -online training platform. UK Department for Science, Innovation and Technology. https://aiskillshub.org.uk/aiskillsboost/

Cosgrove, J., and Cachia, R. (2025). DigComp 3.0: European Digital Competence Framework, Fifth Edition. Publications Office of the European Union, Luxembourg, JRC144121. https://doi.org/10.2760/0001149

Elements of AI. (n.d.). Elements of AI free online course. University of Helsinki and MinnaLearn. https://www.elementsofai.com/

European Commission. (2026a). AI literacy – Questions & answers (updated 27 July 2026). https://digital-strategy.ec.europa.eu/en/faqs/ai-literacy-questions-answers

European Commission. (2026b). Living repository to foster learning and exchange on AI literacy. https://digital-strategy.ec.europa.eu/en/policies/repository-ai-literacy-practices

European Commission, & OECD. (2025). Empowering learners for the age of AI: An AI literacy framework for primary and secondary education (review draft, May 2025). https://ailiteracyframework.org/

European Commission. (2026a). Regulatory framework for AI: the AI Act and the AI omnibus simplification package. Shaping Europe's digital future. https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai

European Commission. (2026b). The AI-adoption divide: Who benefits, who doesn’t, and what it means for workers. https://economy-finance.ec.europa.eu/economic-forecast-and-surveys/economic-forecasts/spring-2026-economic-forecast-slowdown-growth-energy-shock-drives-inflation/ai-adoption-divide-who-benefits-who-doesnt-and-what-it-means-workers_en

Eurostat (2025). Use of artificial intelligence in enterprises. https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises — news release: https://ec.europa.eu/eurostat/web/products-eurostat-news/w/ddn-20251211-2

Finnish Government. (2019). Finland's AI course to contribute to digital skills of Europeans across the continent. https://valtioneuvosto.fi/en/-/1410877/finland-s-ai-course-to-contribute-to-digital-skills-of-europeans-across-the-continent

ISO/IEC. (2023). ISO/IEC 42001:2023, Information technology, Artificial intelligence, Management system. International Organization for Standardization.

Lintner, T. (2024). A systematic review of AI literacy scales. npj Science of Learning, 9, 50. https://doi.org/10.1038/s41539-024-00264-4

Microsoft (2026). 2026 Work Trend Index Annual Report. https://www.microsoft.com/en-us/worklab/work-trend-index/agents-human-agency-and-the-opportunity-for-every-organization

Ministry of Education of China. (2025). Guidelines for general AI education in primary and secondary schools (as reported by People's Daily Online). https://en.people.cn/n3/2025/0513/c90000-20313921.html

NIST. (2023). Artificial Intelligence Risk Management Framework (AI RMF 1.0). U.S. Department of Commerce. https://doi.org/10.6028/NIST.AI.100-1

OECD, and European Union. (2026). Empowering learners for the age of AI: An AI literacy framework for primary and secondary education. OECD Publishing, Paris. https://www.oecd.org/en/publications/empowering-learners-for-the-age-of-ai_65cd27d4-en.html 

Pal, S., Schneider, C., & Lazzaroni, R. M. (2025, April 29). Technical tiers: A new classification framework for global AI workforce analysis. interface. https://www.interface-eu.org/publications/technical-tiers-in-ai-talent

Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 (Artificial Intelligence Act). Official Journal of the European Union, L series, 12 July 2024. http://data.europa.eu/eli/reg/2024/1689/oj

Regulation (EU) 2026/1744 of the European Parliament and of the Council of 8 July 2026 (Digital Omnibus on AI). Official Journal of the European Union, L series, 24 July 2026. http://data.europa.eu/eli/reg/2026/1744/oj

Skills England. (2025, October 29). AI skills for the UK workforce. https://www.gov.uk/government/publications/ai-skills-for-the-uk-workforce

Skills England. (2026, January 28). AI Skills Boost: Skills England's AI foundation skills for work benchmark supports free AI training for all. https://skillsengland.blog.gov.uk/2026/01/28/ai-skills-boost-skills-englands-ai-foundation-skills-for-work-benchmark-supports-free-ai-training-for-all-by-phil-smith/

The White House. (2025, April 23). Executive order: Advancing artificial intelligence education for American youth. https://www.whitehouse.gov/presidential-actions/2025/04/advancing-artificial-intelligence-education-for-american-youth/

UK Government. (2026). AI Opportunities Action Plan: One year on. https://www.gov.uk/government/publications/ai-opportunities-action-plan-one-year-on/ai-opportunities-action-plan-one-year-on

UNESCO. (2024). AI competency framework for students; AI competency framework for teachers. https://www.unesco.org/en/articles/ai-competency-framework-students

World Economic Forum. (2025). The future of jobs report 2025. https://www.weforum.org/publications/the-future-of-jobs-report-2025/