AI literacy in practice: Session notes

AI literacy is not simply a compliance exercise. It is a practical workforce capability that helps people use AI effectively, judge its outputs, manage risk and remain accountable for the work produced. In a DWG Special entitled AI literacy in practice: turning a legal duty into real workforce capability, on September 16, members and guests heard how the EU AI Act can act as a catalyst for action, while the wider business case reaches well beyond the legal baseline.

Elizabeth Marsh, DWG Director of Research, explored the people readiness dimension: what AI literacy means in practice, why it rests on digital and data literacy, and how organizations can map and measure competence by role. Mirsad Capric, DWG Lead Consultant, then widened the lens to organizational readiness, covering governance, data and content, employee experience, operating models, adoption and value.

The EU AI Act puts AI literacy on the agenda

The session provided a plain-language overview of Article 4 of the EU AI Act, while stressing that the discussion was not legal advice. The key points were:

• The duty is live. Article 4 took effect on 2 February 2025. The slides also noted amendments in the 2026 AI Omnibus, alongside later deadlines for other parts of the Act.
• Providers and deployers are in scope. The requirement is relevant not only to organizations creating AI systems, but also to those deploying AI tools under their authority for professional use.
• There is no mandated course or certificate. Measures should reflect the knowledge, experience, role and risk context of the people using AI. Pointing employees only to basic system instructions may be insufficient.
• Evidence matters. Organizations should be able to show what training, guidance and communications they have provided, even though no single record format is prescribed.
• Compliance is the baseline, not the ambition. The stronger case for AI literacy is that it supports quality, value, responsible use and sustainable adoption.

Related reading: Whose job is AI literacy? Why the EU AI Act puts it on your desk

AI literacy depends on digital and data foundations

A recurring message was that organizations should resist treating AI literacy as a standalone topic. Employees need a foundation of digital literacy, then data literacy, before they can use AI consistently and responsibly. Gaps in basic tool selection, information handling, source checking and data judgement can quickly become AI risks.

DWG’s integrated model places workforce capability across seven connected domains:

1. Handling information and data: finding, interpreting, evaluating and organizing information across systems.
2. Digital communication and collaboration: choosing suitable channels and working effectively with others.
3. Digital content creation: creating and preparing digital content that is fit for formal use.
4. Safety, security and data protection: applying cyber hygiene, careful data handling and policy in practice.
5. Problem-solving and digital improvement: resolving issues, adapting to tools and improving how work happens.
6. Digital mindset and behaviours: building confidence, learning agility, professionalism, peer support, wellbeing and reflective practice.
7. AI and emerging technologies: selecting tools safely, critically reviewing outputs, using automation appropriately and being transparent about AI use.

What AI literacy looks like in practice

DWG’s Digital Workplace AI Literacy Model describes five dimensions of knowledge, understanding and use. It is a continuous learning loop rather than a one-off training sequence:

The discussion placed particular emphasis on evaluation and accountability. Easy-to-use AI interfaces can disguise the judgement required to decide whether an output can be trusted, shared or used in a material decision. Accountability remains with people and the organization; it does not transfer to the tool.

Move from one-size-fits-all training to role-based competence

A generic module for every employee is unlikely to be enough. Capability expectations should vary by role, task and risk. The session illustrated three practical proficiency levels:

• Core: performs routine digital tasks correctly in familiar systems and seeks help when something is new.
• Intermediate: works independently and reliably, adapting when tools or rules change.
• Advanced: applies fluency and judgement, anticipates problems and supports colleagues.

The same principle can be applied to key processes. In the example of producing a monthly management pack, an AI agent might draft commentary or condense a report, but people still need to validate source data, challenge unsupported claims, record where AI was used and take responsibility for the final decision.

Measure readiness, not just adoption

The speakers distinguished between evidence that people have access to AI and evidence that they can use it well. Licence counts, interactions and course completion can indicate activity, but they do not show whether employees can exercise sound judgement.

A capability assessment can help organizations:

• Gain insight: establish a real picture of workforce readiness and identify where capability may help or hinder AI plans.
• Focus investment: direct learning and support towards the roles and gaps that need them, rather than sending everyone on the same course.
• Compare confidence with competence: spot both under-confidence, which can suppress useful adoption, and over-confidence, which can increase risk.
• Track progress: create a point-in-time baseline and test whether interventions are changing capability.

It is not too late to assess readiness after AI has been introduced. The assessment can still provide a useful baseline for the next phase, particularly as organizations move towards more agentic systems. The session also cautioned that assessment should be positioned as a development tool, not as a mechanism that increases anxiety or fear about job security.

People readiness needs organizational readiness

The second half of the session showed why individual capability cannot carry the full burden. Someone may know how to operate an AI tool but still be unclear about which information may be shared, who approves an output, where accountability sits or how to raise a concern.

DWG’s AI Maturity Diagnostic considers eight organizational dimensions:

• People and organizational readiness.
• Governance and operating model.
• Responsible AI, risk and compliance.
• Technology, security and integration.
• Employee experience and human–AI interaction.
• Data, content and knowledge readiness.
• Change management and AI fluency.
• Adoption, value and continuous improvement.

Mirsad shared three connected ways to progress: assess maturity to establish a shared baseline; build capability through applied, role-based workshops; and adopt in practice by testing priority workflows, defining measures and scaling what works.

Take our free AI Pulse Check

The discussion also highlighted two common barriers to scale. First, pilots can continue without clear success measures or a defined account of value. Second, governance can become an obstacle if every use case follows the same route. A proportionate model distinguishes higher-risk, enterprise-wide or sensitive-data use cases from lower-risk personal or small-team uses, while retaining appropriate controls.

Practical next steps:

1. Clarify ownership. Bring digital workplace, HR and learning, AI governance, risk, data and business leaders into the same conversation. AI literacy is shared work.
2. Define what good looks like. Describe observable capability in language that fits your organization. Include digital, data and AI competence.
3. Differentiate by role and risk. Set target levels for role families and pay particular attention to consequential decisions and high-risk processes.
4. Assess confidence and competence together. Use more than self-reported confidence. Add scenarios, judgement questions and, where feasible, practical observation.
5. Map critical human–AI handoffs. Identify where AI generates, recommends or acts, where a person reviews, and where accountability sits.
6. Connect learning to real work. Use workshops, playbooks, peer support and guided practice around actual workflows rather than feature-led training alone.
7. Define value before scaling. Agree outcomes and measures at the start of a pilot. Use evidence to decide what should scale, change or stop.
8. Keep evidence. Record the guidance, learning, communications and support provided, both for improvement and to demonstrate the measures taken.

Continue the conversation

The related article provides a concise starting point for conversations about ownership of AI literacy and the practical implications of the EU AI Act.

Read: Whose job is AI literacy? Why the EU AI Act puts it on your desk

Categorised in:   → Digital literacy, Artificial intelligence and automation