AI training in the workplace must introduce employees to the safe and appropriate use of the AI systems they work with, in a way that reflects their responsibilities. Since 27 July 2026, Article 4 of the EU AI Act has required measures to support AI literacy, but it does not require a certificate, a fixed training programme or a dedicated AI officer. Nor do companies have to guarantee a particular level of competence for every individual.
What exactly is required for AI training in the workplace?
The business must take appropriate measures to enable employees to use AI systems competently. The appropriate measures depend on their responsibilities, their existing knowledge and the possible consequences of incorrect use.
According to the legal act of the European Union, the revised Article 4 of the EU AI Act has applied since 27 July 2026. The amendment is intended to make the obligation clearer and less bureaucratic. The European Commission also explains when the revised rules entered into force.
Companies are affected regardless of their size or sector. This does not apply only to software developed in house. Using a widely available chatbot or AI assistant within the business may be enough to make appropriate measures necessary.
What matters is how the system is actually used. An employee who uses ChatGPT to draft quotations needs different knowledge from an employee who reviews AI supported staff scheduling. By contrast, people who do not use the system do not need to complete the same programme as a precaution.
What should an appropriate briefing cover?
The content should reflect the specific tool and the work involved. A short internal briefing may be sufficient for a simple writing assistant, while more detailed rules are advisable when decisions concern staff, prices or customers.
For a business that uses ChatGPT or a similar assistant for emails, minutes and drafts, relevant topics could include:
- Permitted tasks: Employees learn what the tool may be used for and where a human decision is still required.
- Handling data: Customer data, employee details, trade secrets and confidential documents must not be entered without first being checked.
- Checking results: AI systems can produce information that sounds convincing but is incorrect. This type of error is often called a hallucination.
- Responsibility: Before content is sent or processed further, a responsible person checks the content, figures and recipients.
- Dealing with problems: Employees must know whom to notify about errors, unclear results or a possible data incident.
A short exercise using genuine examples from day to day work is usually more helpful than a general presentation. A skilled trades business might review a quotation, a tax consultancy could check a draft email to a client, and a care provider could examine an anonymised summary. Real personal data should not be used in exercises.
Before training takes place, the business should therefore establish which tools are being used. When planning AI and automation for recurring tasks, usage, responsibilities and the knowledge required should be considered together.
Which costly measures can businesses avoid?
A purchased certificate, a standardised training programme and a dedicated AI officer are not required. Nor is a full day of training for the entire workforce automatically necessary.
For example, a business with twelve employees, three of whom use a writing assistant for internal drafts, can organise a short, documented briefing for those three people. It can be delivered internally if the person responsible has sufficient knowledge of the tool, its limitations and the company’s rules.
More expensive external training may still be appropriate if the application is complex or its results could have significant consequences. This may be the case with staffing decisions, automated assessments or extensive processing of sensitive data. In these circumstances, the measure is a business decision based on risk, not a general requirement for every company.
A business may also choose to appoint an AI officer voluntarily. In small businesses, it is often sufficient to have one clearly designated person who records the tools in use, maintains the rules, organises briefings and answers questions. There is no need to create a new full time position for this purpose.
How should AI training be documented?
The documentation should provide a clear record of which measure was carried out, when it took place and who it was for. It should record at least the type of measure, its content, its duration and the people who took part.
The Bundesnetzagentur (the German Federal Network Agency) recommends three steps: identify the need, design appropriate measures, then document, review and regularly refresh them. It does not require a specific form or learning system.
A simple table may be sufficient for an internal briefing. It can record the date, the AI system used, the topics covered, the duration, the name of the person providing the briefing and the names of the participants. The business can also retain the presentation used, a work instruction or the examples discussed.
The requirements referred to do not stipulate that training must be repeated annually. However, the business should review the briefing when a new tool is introduced, functions change, additional employees are given access or errors occur during day to day work. A regular short refresher is also advisable even if the way the system is used remains unchanged.
Where can small businesses find practical support?
Small and medium-sized businesses can begin by looking at free or provider-neutral services. According to the Bundesnetzagentur, around 100 AI trainers from the Mittelstand Digital Zentren (digital centres for small and medium-sized businesses) provide support, workshops and AI checks.
Demand is already high. According to a Bitkom survey of businesses, 57 per cent of companies with 20 employees or more use AI. 70 per cent provide AI training, while 25 per cent do not yet offer any professional development on this topic.
Before choosing a service, you should answer three questions:
- Which AI systems are actually used within the business?
- Which employees operate or monitor these systems?
- What harm could result from incorrect outputs, unsuitable inputs or inadequate oversight?
The answers can be used to determine an appropriate measure. This might be a 60-minute internal briefing, a combination of basic training and work instructions, or an external workshop for specific roles.
ibx company advises small and medium-sized businesses in Hannover and the surrounding region on selecting and introducing suitable digital solutions. If you would like to assess your training needs, you can arrange a no-obligation initial consultation about using AI in your business.