AI literacy and applied AI in learning
AI in education is useful when people can judge it. With partners, we can design practice in which adults use, question and correct AI in realistic work situations. This page shows what we have built, what we can design with partners, and what we do not build.
Why AI literacy is a practical skill
AI tools give answers that can sound right and still be wrong. The EU's digital competence framework, DigComp 3.0, puts it plainly: AI output can be inaccurate yet plausible, and the person who uses it is responsible for checking it.
Knowing that is one thing. Noticing it on a busy working day is another. It takes practice to decide when to trust, check, correct or not use an AI tool, in a situation that looks like your own work.
DigComp 3.0 builds AI into all 21 of its competences, from evaluating information to collaboration, privacy and problem solving. AI skills are part of everyday digital competence, for staff and for educators alike.
The European Commission also notes that relying only on a system's instructions for use may not be enough to build AI literacy.
What the AI Act asks of organisations
- Since 2 February 2025, the EU AI Act has required providers and deployers of AI systems to take AI literacy measures for their staff and for others who deal with AI systems on their behalf. An organisation that uses AI tools at work under its own authority is usually a deployer.
- In 2026, the Digital Omnibus on AI (Regulation (EU) 2026/1744) amended this duty. Providers and deployers still take measures that support AI literacy, fitted to people's knowledge, experience and training and to the context in which AI is used, but the law no longer requires a specific level of AI literacy.
- No certificate is needed. The European Commission says an internal record of training and other measures is enough.
- National market-surveillance authorities supervise and enforce the duty from 2 August 2026.
The Commission's own guidance is in European Commission: AI literacy questions and answers.
Insight AI Act Article 4 explained, with the official sources
Last checked: 12 November 2026. This is not legal advice.
AI literacy scenarios we can design with partners
When an organisation or a consortium needs AI literacy training for staff or adult educators, we can design the practice part, together with partners who bring the subject and teaching expertise.
We map the learning outcomes of each scenario to the AI-related competences in DigComp 3.0.
How such a scenario works:
- Staff use an AI assistant in a realistic task from their own work.
- The scenario contains an error, a biased answer or a privacy risk.
- They decide whether to send, edit, check or stop.
- They see the consequence, reflect and try again.
Three scenario ideas:
Spot the AI mistake
An AI assistant drafts a reply to a client with one subtle error. Send it, edit it or check first?
The meeting note-taker
An AI tool takes notes in a team meeting. Who needs to agree to that, and what happens when it attributes a decision to the wrong person?
Client data and a public chatbot
A colleague is about to paste a client's details into a public chatbot. What do you do instead?
The method is the one behind all our scenario-based learning: decide, see the consequence, reflect, try again. Organisations can also keep a record of who practised what. In our VR conversation training (in Dutch), results per participant (scores, time and progress) go to the organisation's own environment in our portal.
Applied AI in our work: spoken interaction and an R&D study
In our own learning tools, we use AI only where it serves the learning goal: transparently, with human oversight, and with the AI Act and GDPR in mind.
Spoken interaction in VR role-play
In a VR role-play for airport security staff, staff talk freely with a virtual passenger, without clicking through answer menus. The passenger reacts to their approach. After each scene there is a short reflection with feedback, and the scene can be played again with another approach.
Showcase Trigion: free spoken conversation with a virtual passenger
An R&D study for the Dutch police
ASA, the Augmented Situational Awareness assistant, is an augmented-reality assistant for police officers. Among other things, it warns officers about people approaching from outside their field of view. For this, the software detects people on the device itself; it does not identify them.
Crow Group B.V. developed ASA as a phase 1 feasibility study in theme 2 of the Innovation Impact Challenge IX for the Dutch police (CIIIC and RVO). It is applied AI and research, not a learning tool.
Public summaries of the challenge are on the RVO website: RVO public summaries of the Innovation Impact Challenge IX (in Dutch).
Other AI features we mention on this site, such as assistants that answer from your own documents, are offers, not delivered work. They are described on our technology partner page. Journalists and researchers are welcome to send press and research enquiries.
What we don't build
AI that grades learners
In education and vocational training institutions, AI that evaluates learning outcomes, including when those outcomes steer the learning process, is high-risk under the AI Act, with rules applying from 2 December 2027. Where AI supports trainers in our tools, a person makes the decision.
Emotion recognition
AI that infers the emotions of people at work or in education institutions has been prohibited since 2 February 2025, except for medical or safety reasons. We record choices, time and progress instead.
Our own foundation models
We integrate existing models where they help; we do not train foundation models.
Empty labels
We do not put an AI label on a product without saying what the AI does.
Last checked: 12 November 2026. This is not legal advice.
Responsible by design
The principles we design AI features by:
Clear roles
For each AI feature, the project states who is the provider and who is the deployer under the AI Act.
Transparency
People know when they are talking to an AI system. Synthetic voices and images that could pass as real are disclosed.
Data
An AI feature gets only the data it needs, and the project agreement says which data that is.
Human oversight
A person reviews generated content and makes the decisions that affect learners.
Game elements follow the same rule. DigComp 3.0 lists gamification next to nudging and clickbait as strategies that steer users, so we keep it open and in service of the learning goal. More in gamification in education: what works for adults.
Our own rules for using AI at work are set out in how we use AI.
A useful reference for educators is the European Commission's 2026 update of its ethical guidelines on AI and data in teaching and learning.
AI literacy in Erasmus+ proposals
The Erasmus+ Programme Guide 2026 supports the "purposeful use of emerging digital technologies and innovation, including AI" under its digital transformation priority.
DigComp 3.0 gives partners a shared language for AI-related learning outcomes: AI appears in all 21 competences, and the framework has over 500 learning outcomes. A recital of the 2026 amending regulation names DigComp as a framework the Commission and the Member States could take into account when they support AI literacy.
For educators, the JRC is updating DigCompEdu, with educators' AI literacy built in throughout. The final version is due in spring 2027.
As partner for AI in learning in a consortium, we can prototype AI-supported learning tools, pilot them with learners, document them, release the learning content under an open licence and plan hosting and maintenance. For your next proposal, partner with us.
Insight DigComp 3.0 and AI competences
Framework fit. AI literacy scenarios like these map to DigComp 3.0 competences such as 1.2 Evaluating information, 2.4 Collaborating through digital technologies and 4.2 Protecting personal data and privacy. See how our work maps to EU competence frameworks.
Frequently asked questions
Last checked: 12 November 2026. This is not legal advice.
Is AI literacy training mandatory under the AI Act?
Providers and deployers of AI systems must take measures to support the AI literacy of their staff. The law, as amended in 2026, does not prescribe a course, a certificate or a fixed level. Measures should fit people's knowledge and the context in which AI is used. National authorities supervise this from 2 August 2026.
Do our staff need a certificate?
No. The European Commission says there is no need for a certificate. An internal record of training and other measures is enough.
What does Crow Group build with AI?
Spoken interaction with a virtual character in a VR role-play for airport security staff, and on-device person detection in an augmented-reality feasibility study for the Dutch police. We integrate existing models; we do not train foundation models.
Can AI assess or grade our learners?
In education and vocational training institutions, AI that evaluates learning outcomes is high-risk under the AI Act from 2 December 2027. AI can support trainers, for example by preparing feedback, but in our tools a person makes the decision.
Can your VR training detect stress or emotions?
No. The AI Act prohibits AI that infers the emotions of people at work or in education institutions, except for medical or safety reasons. We record choices, time and progress instead.
Does it work without a VR headset?
Yes. Scenarios can run in a browser, on a tablet or in VR.
Can you join our Erasmus+ project?
Yes, as technology partner for AI-supported learning tools: prototyping, piloting with learners, documentation and hosting.
Talk to us about AI in your project
Tell us what you want to achieve with AI, or how your staff should learn to work with it. We will tell you plainly where AI helps, and where it does not.