AI transformation in employment law

AI has long since become part of everyday working life in many companies. According to a representative Bitkom survey published in September 2026, 57 % of companies in Germany with at least 20 employees are already using AI; a further 38 % are planning on or discussing using it. Generative AI, for example, can assist with drafting texts, analysing large volumes of data or preparing presentations. At the same time, around nine out of ten companies that are already using AI believe they have so far made little or no use of its potential. The figures make it clear that, whilst the first step has been taken in many places, the comprehensive transformation of work processes and organisational structures is, for the most part, still to come.

This is where the next stage of development comes in. Whilst the focus has often been on individual applications and simplifying specific work steps, the aim now is to systematically align work processes, job profiles and organisational structures with the use of AI. AI is thus evolving from a mere tool into a factor shaping the organisation of work as a whole. 

AI transformation presents companies with significant challenges in the field of employment law that go beyond the requirements of the European AI Act. It affects codetermination, employee data protection and anti-discrimination legislation, alters qualification requirements and job profiles, and may ultimately have an impact on staffing requirements.

The key challenge under employment law in future will therefore lie less in examining individual AI systems in isolation and more in shaping the way work is organised in the wake of AI-driven change. Companies should therefore not view AI transformation solely as an IT or compliance project, but should also take its implications under employment law into account.

AI governance: How can companies manage the use of AI to ensure legal compliance?

An AI-driven transformation process usually begins with a seemingly simple question: What form of AI should actually be used within the company and for what purposes? As the number of applications grows, it becomes increasingly apparent that it is no longer sufficient to assess the legal permissibility of each individual tool in isolation. Companies therefore need structures that enable them to manage the use of AI on a long-term basis. This begins with an inventory of the systems already in use. Without such structures, there is a risk of uncontrolled "shadow AI", particularly where freely accessible generative AI applications are concerned. Employees may use such applications without the company knowing what data are being processed in them or for what purposes the results are being used.

Responsibilities and approval processes should be defined on this basis. In particular, it must be clarified which AI applications may be used without further review, when the data protection, IT security, HR or legal departments need to be involved and which applications require in-depth assessment due to their intended purpose or their impact on employees. A phased process is conceivable here: identification of the use case; legal and technical classification; assessment of the potential impact on employees and their working conditions, as well as any participation rights of the works council; approval; and subsequent monitoring of the application's use.

Employees' AI literacy is also of particular importance. Article 4 AI Act requires providers and deployers of AI systems to take measures to support the development of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf, taking particular account of their technical knowledge, experience, education and training and the context the AI systems are to be used in. Training should therefore not be limited to the technical operation of a tool. Employees must, in particular, understand the limitations of the system in question, how to critically assess results and what data may actually be processed.

AI governance is therefore not merely a compliance task. It creates the organisational foundation for integrating new AI applications into the organisation in a way that is legally compliant and scalable.

AI and the works council: What codetermination rights are there?

The use of AI is also a key issue in codetermination. Works councils are often sceptical about the introduction of AI systems, which is hardly surprising given their virtually limitless technical possibilities and potential applications. When assessing the works council's potential rights of participation, it is necessary to consider the purpose for which a particular system is being used and the impact it may have on work processes and employees, as depending on these factors, the works council may have varying degrees of opportunity to participate. 

For instance, under section 90 (1) no. 3 German Works Constitution Act (BetrVG), the works council must be informed in due time of any plans concerning working procedures and operations including the use of AI. This is followed by consultation rights under section 90 (2) German Works Constitution Act (BetrVG). 

However, "genuine" codetermination rights – which can be enforced (if necessary, via the conciliation board) – are also an option. Section 87 (1) no. 6 German Works Constitution Act (BetrVG) is of particular practical significance in this context when technical devices are used that are capable of monitoring the behaviour or performance of the employees. The vast majority of AI systems are likely to be objectively capable of such monitoring, which, according to the established case law of the German Federal Labour Court (BAG), is sufficient in itself to trigger the right of codetermination. It does not depend on the employer having a subjective "intention to monitor", which, upon closer consideration, is surprising given that the wording of no. 6 refers to technical devices "designed" to monitor behaviour or performance. Given that our working lives are likely to be permeated by countless AI tools in the future, the question of whether works council involvement is strictly necessary when deciding on their introduction should be assessed with a sense of proportion.

Further participation rights may also be relevant in relation to selection guidelines, staff questionnaires or individual staff measures. Section 95 (2a) German Works Constitution Act (BetrVG) expressly clarifies that participation rights in relation to selection guidelines also apply is AI is used in drafting them. Insofar as the works council has to assess the introduction or application of AI in order to carry out its tasks, section 80 (3) German Works Constitution Act (BetrVG) expressly provides for this: In this regard, the involvement of an expert is generally deemed necessary. 

In the case of more extensive AI-driven transformations, participation rights under sections 111 ff. German Works Constitution Act (BetrVG) may also apply. The mere provision of an AI system as an additional tool does not, in itself, generally constitute an alteration to the establishment. However, if its use is accompanied by fundamental changes to the organisation of the establishment or work methods, or a significant reduction in staff, negotiations on a reconciliation of interests and, where appropriate, a social plan may be necessary (for further details, see Münnich/Lambertz). When it comes to project planning, the interplay of participation rights is crucial. If restructuring depends on an AI system being introduced, the works council may, through its right of codetermination under section 87 (1) no. 6 German Works Constitution Act (BetrVG), indirectly influence the timeline of the entire transformation process.

For companies, this means above all else that the works council's involvement should not begin only immediately prior to the technical go-live, but should be initiated well in advance. The more significantly the AI application concerned is intended to alter work processes, the sooner it should be clarified what participation rights apply and how these can be integrated into the transformation process. Otherwise, there is a risk that unresolved issues relating to codetermination will delay the company-wide roll-out.

AI framework works agreement: standardised rules rather than individual negotiations

The rising number of AI applications used in businesses raises another practical question: Do companies have to completely renegotiate the framework conditions under employment law for every new system?

In the case of a long-term AI transformation, it may make sense, in terms of time and cost, to set out recurring fundamental issues in a framework works agreement on the use of AI.

This framework works agreement could, for example, set out general principles governing the introduction and use of AI, information processes to be followed for new applications, guidelines on testing and pilot phases, data protection and data use, training measures, human oversight mechanisms and limits on performance and behavioural monitoring.

The key factor here is striking the right balance when it comes to the level of detail. A framework works agreement intended to comprehensively regulate every conceivable future application is likely to reach its limits quickly given the pace of technological development. Conversely, a framework regulation is of little use if every new application requires all matters to be renegotiated from scratch. A phased approach is therefore worth considering. The framework works agreement sets out the basic rules and the process for introducing new AI applications. Specific use cases are then categorised according to defined criteria and, depending on their function and their impact on employees, supplemented by information sheets, schedules or separate agreements.

Codetermination thus becomes part of company-wide AI governance. A well-drafted framework works agreement can therefore resolve the perceived contradiction that codetermination and speed need not be mutually exclusive when introducing new AI applications. When properly integrated, they can instead jointly pave the way for an AI transformation that is both legally compliant and fast.

Skills development as a key issue in AI transformation

In many cases, AI will not completely replace jobs, but will initially take over individual tasks. Routine tasks can be automated, whilst monitoring, evaluation and management tasks are becoming increasingly important. This development may take place in stages. To begin with, an AI system may merely assist with individual tasks. Later, it may take over certain tasks entirely. As a result, the nature of the work changes, with some tasks being phased out and new ones emerging.

From an employment law perspective, the question then arises as to the extent to which such a change can be implemented within the existing employment relationship. The starting point is the employment contract and the employer's right to issue instructions under section 106 German Trade, Commerce and Industry Regulation Act (GewO). The more clearly the work to be performed is defined in the employment contract, the more limited the scope for unilaterally altering the range of duties by way of instruction may be. In the event of significant changes, the works council's participation rights may also need to be taken into account, particularly in connection with transfers (section 99 (1) and section 95 (3) German Works Constitution Act (BetrVG)). Depending on the specific arrangements, issues relating to grading and remuneration may also arise.

Meanwhile, qualification requirements are changing. A distinction must be made between two levels here. On the one hand, employees need the necessary AI skills to be able to use the systems safely and appropriately, and on the other hand, changing job profiles may necessitate further upskilling or reskilling. Skills development is not simply a matter for human resources policy. Works constitution law also sets out specific participation rights in this regard. Under section 96 (1) German Works Constitution Act (BetrVG), the employer, at the request of the works council, has to determine the need for vocational training and consult with the works council on matters relating to vocational training. Section 97 (2) German Works Constitution Act (BetrVG) may take on particular significance in the event of changes to job roles brought about by AI: If the employer plans or implements measures which result in a change to the work of the employees concerned and their existing knowledge and skills are no longer sufficient for them to carry out their duties, the works council has to participate in the decisions relating to the implementation of vocational training programmes in the establishment. 

Particularly in the case of large-scale AI transformation projects, technical and organisational changes can therefore trigger a direct need for upskilling. Companies should therefore not wait until after an AI system has been introduced to assess whether employees can keep pace with the changed requirements. Rather, the question of which roles will remain in the future, which will change, and what skills will be required for them should already form part of the transformation planning. Skills development is thus an essential component of a successful AI transformation.

AI in HR: To what extent should algorithms be allowed to make decisions?

The use of AI in HR matters is a particularly sensitive issue. The potential applications here are wide-ranging. An AI system can pre-screen applications and assess candidates. Other applications can match employees to suitable roles based on specific skills, analyse performance data or support staff planning. It is also conceivable to use AI in connection with organisational changes.

The AI Act attaches particular importance to such applications. In particular, AI systems intended for the recruitment or selection of candidates, certain HR decisions or monitoring and evaluating performance and behaviour are classified as high-risk AI under Annex III of the AI Act. In accordance with the amended version of the AI Act published in July 2026, the specific requirements of the high-risk regime set out for this purpose will apply to these systems from 2 December 2027.

Regardless of the regulatory classification under the AI Act, however, data protection, anti-discrimination and works constitution legislation must still be observed. Specific limitations may arise in particular from Article 22 GDPR if a decision is based solely on automated processing and produces legal effects concerning a person or otherwise significantly affects them in a similar way. In this context, it is not necessarily enough to formally involve a person in the decision-making process. If a decision-maker merely accepts a recommendation or score provided by an AI system without independently and critically reviewing the result, this raises the question of whether a genuine human decision has actually been made. Then there are risks of discriminatory decisions. Whilst AI can standardise HR decisions to a greater extent, if systems are trained using historical data or employ criteria indirectly linked to protected characteristics, existing biases may be reproduced or even exacerbated. In such a case, the employer cannot simply claim that the decision was technically prepared by an algorithm. The more significant an HR decision is, the more important it is to have transparent criteria, clear lines of responsibility and genuinely effective human oversight.

AI and employee data protection: What are the limits of monitoring?

A further challenge posed by AI systems is that they often generate substantial amounts of usage data. From a technical point of view, it may, for example, be possible to track which employees use a system, how often they access it, which functions they use or what results are produced. If such information is linked together, it can yield increasingly detailed insights into the working methods and performance of individual employees. This blurs the line between system administration and performance and behaviour monitoring. However, not every technically feasible analysis should actually be carried out. Companies should instead determine, when introducing a system, which data are required for operations, security and further development, how long they should be stored, and who is authorised to access them. A subsequent change in the purpose of use is particularly relevant. For example, an AI application may initially be introduced solely as an assistance system. If the company subsequently realises that the existing usage data can also be used to assess the productivity of individual employees, the assessment under employment law changes significantly. A system that was introduced as an assistance system can thus become a tool for monitoring performance and behaviour as a result of a change in data usage.

Practice should therefore be guided by this clear principle: It is not the technical feasibility, but the predefined purpose that should determine which employee data are analysed. It is precisely at this point that AI governance, employee data protection and workplace codetermination are directly intertwined.

AI transformation: What does this mean for the organisation and staffing requirements?

It is at this point, if not before, that the discussion about AI leads to the actual transformation of the company. Companies invest in AI specifically because they expect it to deliver gains in efficiency and productivity. If processes are accelerated or individual tasks are automated, the question of how this will affect the organisation and future staffing requirements is bound to arise sooner or later. This development can take various forms. Any capacity that is freed up can be used for new tasks. Employees can be trained and placed in new roles. However, companies may also come to the conclusion that certain tasks or roles will no longer be required to the same extent in future, resulting in a potential surplus of staff.

Productivity gains can thus gradually lead to a change in staffing requirements. However, between the automation of individual tasks and a possible reduction in staff, there are usually further organisational decisions to be made, in particular the question of whether the capacity freed up can be utilised elsewhere and whether employees can be retrained for different roles. It is only at this stage that technological transformation can lead to restructuring under employment law. For dismissals for operational reasons, the mere introduction of an AI system is not sufficient. Urgent operational needs within the meaning of section 1 (2) German Act against Unfair Dismissal (KSchG) can only be deemed to exist if the implementation of an organisational business decision leads to a foreseeable permanent reduction in the need for labour by the end of the notice period at the latest. The relevant forecast must be objectively justified at the time the notice of dismissal is received. This is in line with the general principles of the German Federal Labour Court's case law on dismissal for operational reasons.

It is not the AI that "dismisses" the employee. The decisive factor is which specific organisational decision the employer takes on the basis of the technical possibilities opened up by AI and how this affects the need for labour. Simply pointing out that tasks may in future be "carried out by AI" is not sufficient for this purpose. Rather, the company must be able to demonstrate transparently how work will be organised in future and to what extent tasks previously performed by employees will be permanently eliminated as a result. Particularly where the organisational decision and the dismissal decision occur in close succession, case law requires a specific explanation of the impact on employment opportunities. Before issuing dismissals for operational reasons, opportunities for alternative employment must be considered. Particularly in the case of an AI-driven transformation, the question of whether employees can continue to be employed in modified or other vacant posts and what role prior training plays in this may become increasingly significant.

Depending on the scope of the organisational changes, the participation rights of the works council in the event of alterations as per sections 111 ff. German Works Constitution Act (BetrVG) may also need to be taken into account.

The key question under employment law is therefore not: "Can AI replace this job?" The crucial question is: What should the future organisation look like and what employment opportunities are available within it?

Conclusion: AI transformation and employment law – what companies need to bear in mind

AI transformation is not only changing individual work processes, but is also increasingly altering job profiles, skill requirements and organisational structures. Companies should therefore incorporate employment law considerations into their transformation planning from the outset. The following aspects are particularly important in this regard:

• Establish AI governance: Companies should identify which AI systems are already in use, define clear responsibilities and establish approval and monitoring processes for new applications.
• Involve the works council at an early stage: Depending on the intended purpose and impact of an AI system, different participation rights may apply. Involving the works council at an early stage can help to prevent delays in implementation.
• Review the framework works agreement: A framework works agreement on the use of AI can provide uniform rules for recurring issues and facilitate the introduction of new applications.
• Train employees: In addition to basic AI literacy, changing job profiles may require further training. Here, too, any rights of codetermination must be taken into account.
• Implement AI in HR in a legally compliant manner: When making AI-supported HR decisions, particular attention must be paid to data protection, risks of discrimination and effective human oversight.
• Protect employee data: Companies should specify the purposes for which usage data may be collected and analysed. Subsequent use for the purposes of performance or behavioural monitoring may alter the assessment under employment law.
• Assess the impact on staffing requirements: The introduction of AI alone does not justify dismissals for operational reasons. What matters are specific organisational decisions and their demonstrable impact on staffing requirements. Opportunities for continued employment and skills development must also be taken into account.

The central task under employment law is therefore not to regulate the use of individual AI systems. Rather, companies must organise their work in such a way that it keeps pace with technological changes and the new reality of work.