AI competence under article 4 of the EU AI Act: Why training alone is not enough

Article 4 of the EU AI Act therefore makes AI competence an organizational requirement. Providers and operators of AI systems must ensure that the individuals involved possess an appropriate level of knowledge and skills. This includes, among other things, prior knowledge, experience, training, the context of use, and the affected groups of people.

The key takeaway: A one-time AI training session is not enough. AI competence only becomes effective when learning, clear rules, application-specific knowledge, and accessible experts work together on an ongoing basis. Westernacher Solutions views this not merely as a regulatory requirement, but as a prerequisite for the responsible, safe, and meaningful use of AI.

AI competence—referred to as “AI literacy” in the EU AI Act—does not mean that all employees must be able to program language models or explain their mathematics.

It is crucial that people have a basic understanding of how the systems they use work and what their limitations are. They should be able to identify risks and misuse, critically evaluate results, distinguish between permissible data, and understand their responsibilities in their respective work contexts.

This is particularly important when sensitive information is being processed or when AI results can have significant consequences—for example, in the judicial system, public administration, or religious organizations.

Example: An AI assistant can generate a compelling summary but may still misrepresent a deadline, an exception, or a source. AI literacy, then, involves not accepting the result at face value, verifying critical information against the original source, and refraining from entering inappropriate confidential data.

The EU AI Act does not require uniform training for everyone. The required level of competence depends on the AI system used, the context of use, the risks involved, and a person’s actual role.

In practice, a distinction can be made between foundational and role-specific AI competencies:

• Users need a basic understanding of the opportunities, limitations, data protection, permissible tools, and how to verify results.
• Project managers and executives must also assess risks, define responsibilities, and establish appropriate controls.
• Development and operations teams need in-depth knowledge of data quality, model behavior, security, documentation, monitoring, and human oversight.

Therefore, someone who uses an AI assistant for design work needs different skills than someone who develops an AI-based product, manages a client project, or oversees day-to-day operations.

What matters is not the job title, but the specific role one plays in working with AI.

Westernacher Solutions does not view AI expertise as a one-time mandatory training session. Knowledge must remain accessible in day-to-day work and continue to evolve alongside tools, use cases, and regulatory requirements.

Our approach combines four key elements:

1. Teaching the basics: Regular training sessions cover topics such as the practical use of Microsoft Copilot Chat and the responsible use of AI in development.
2. Keeping knowledge accessible: Records, information, and documentation make it possible to retrieve content later.
3. Role-Based Deep Dive: Teams gain the knowledge they need for their specific tasks, systems, and risks.
4. Ensure technical support: For complex issues, consolidated expertise and clear points of contact are available.

In this way, AI expertise is not only taught during training sessions, but is also integrated into processes, projects, and day-to-day decisions.

Training is only one part of a sustainable skills strategy. Common standards, documented knowledge, and access to experts are just as important.

Westernacher Solutions brings together AI expertise in its AI Competence Center. This knowledge is applied to the professional development of our employees, as well as to projects, products, and the consulting services we provide to our clients.

This approach combines two perspectives: Employees should be able to use approved AI tools safely and critically. Teams that develop, operate, or deploy AI solutions in client projects also need in-depth technical, organizational, and regulatory knowledge.

Shared AI expertise does not mean that everyone knows everything. It arises when people have the knowledge needed for their task, recognize their limitations, and know where to find additional expertise.

The requirement for AI competence under Article 4 has been in effect since February 2, 2025. Organizations should therefore clearly define which AI systems are used, which roles are involved, what knowledge is required, and how that knowledge is developed and kept up to date.

However, treating AI expertise merely as a compliance task is not enough. When implemented correctly, it enhances the quality of decisions, reduces misuse, and builds trust in new tools.

Employees can experiment with AI with greater confidence, scrutinize results more critically, and identify risks earlier. At the same time, organizations gain a solid foundation for integrating AI into their work not just on an ad hoc basis, but in a long-term and responsible manner.

Government agencies, the judiciary, and other organizations with a special responsibility toward citizens stand to benefit particularly from a structured approach. You can find additional insights in our articles on the EU AI Act for businesses, on sovereignty and artificial intelligence, and on trustworthy AI in public administration.

The EU AI Act makes AI expertise mandatory. However, the real added value only emerges when organizations turn it into a lasting capability: through role-based learning, clear rules, accessible knowledge, and expert support. In this way, a legal requirement becomes the foundation for the safe and responsible use of AI.

AI Competence—Training Alone Is Not Enough