EU AI Act: Companies Need to Step Up Their AI Expertise

The EU AI Act has required providers and operators of AI systems since February 2, 2025, to take measures to ensure that employees who work with AI systems have a sufficient level of AI competence. In practice, however, training in many places has not kept pace with the use of AI. Starting August 2, 2026, the relevant national market surveillance authorities will be able to monitor and enforce compliance with this requirement.

The EdTech company Pinktum points out that the EU AI Act does not require any specific training or certification. However, the training must be appropriate for the systems used, the tasks involved, and the users’ prior knowledge.

What companies should do now

The foundation is an overview of AI usage within the company: Which systems are used for what purposes and with what data? What risks exist? In many companies, it is also advisable to have an AI policy that specifies which applications are permitted, what data may be entered, how the quality of results is verified, and who is responsible. In doing so, companies should also consider which external parties—such as service providers or freelancers—work with AI systems on behalf of the company and to what extent these individuals must also possess the necessary AI expertise.

Joachim Pawlik, CEO of Pinktum, sums it up: «Companies should not view continuing education as a one-time formality, but rather build AI expertise systematically and in a way that is relevant to real-world applications. This is not only legally important, but also helps unlock the true benefits of AI.»

Alois Krtil, CTO of Pinktum, adds: «Those who understand the capabilities and limitations of AI and can critically evaluate results will avoid mistakes and use the technology more productively. To achieve this, the learning content must be tailored to the specific tasks of the employees.»

Checklist for Building AI Competence

Pinktum has defined six steps to guide companies in systematically building AI expertise:

1. Identify target audiences and training needs: Companies should track which individuals work with which AI systems, how they use them, and what prior knowledge they bring to the role. This should also include external individuals who work with AI systems on behalf of the company. For example, someone who uses AI to generate initial drafts needs different skills than a manager who incorporates AI-driven analyses into decision-making.

2. Establish a shared understanding of the fundamentals: All users should have a basic understanding of how AI systems work, what possibilities they offer, and where their limitations lie. This includes being aware that even convincingly presented results can be incorrect, incomplete, or biased.

3. Communicate internal rules and responsibilities: Employees need to know which systems they are authorized to use for which tasks, what data can be entered, and who reviews and approves the results. In practice, for example, it must be clear which customer data or internal documents may be entered into an AI system.

4. Provide role- and application-specific training: Basic training establishes a common understanding, but it is not sufficient for all tasks. The training content should be tailored to the specific tasks and systems used, as well as the associated risks.

5. Practice and develop critical thinking skills: Training should include typical work situations and practical exercises. For example, employees should practice checking AI-generated results for factual errors, inappropriate sources, and potential biases.

6. Document and update training records: Companies should maintain a transparent record of which target groups have been trained on which content. If new systems, functions, or use cases are introduced, the training content should also be reviewed and updated so that the development of the necessary AI expertise can be transparently demonstrated to regulatory authorities.

Source and further information: https://www.pinktum.com