Using AI for shift planning is generally permitted in Germany. The decisive question is how the technology is used: as a tool that provides suggestions, or as a system that makes staffing decisions without human oversight. Several sets of regulations define the framework – the section below sets out precisely which ones apply.
You can be legally compliant by following these three basic rules:
- No personal data was uploaded to a generative AI – neither to the company’s system nor to so-called “shadow AI.”
- The AI never makes decisions on its own – a human reviews the plan and approves it.
- The results are transparent and verifiable.
This article provides a cross-sector overview of the legal landscape, covering health care and social services, logistics and security, retail, and industrial sectors. It explains what AI is permitted to do in shift planning, where the limits lie, and how you can implement its use properly.
Is it legal to use AI for shift planning?
Yes. There is no law prohibiting the use of AI in workforce management. The legal system does not regulate the tool itself, but rather its impact on employees.
The most important distinction is that between supportive and decision-making AI. A system that calculates a shift schedule draft, which a shift scheduler then reviews and approves, poses no legal risk. It only becomes problematic when software independently decides on work schedules, assignments, or rejections, and no human intervenes anymore. The following regulations address precisely this dividing line.
An Overview of the Legal Framework
These sets of rules determine what is permitted when using AI for shift planning. Four of them form the core – we’ve highlighted them in bold – and others may apply depending on the business and sector.
| Regulatory framework | What It Covers | Relevance to the shift schedule |
|---|---|---|
| GDPR | Processing of Personal Data, Automated Individual Decision-Making | Limits on Fully Automated Decisions, Data Minimization |
| EU AI Act | Risk-Based Regulation of AI Systems | Classification of HR/Workforce AI, Transparency, and Oversight Requirements |
| ArbZG (German Working Hours Act) | Maximum working hours, rest times, breaks | Binding limits that no AI is allowed to cross |
| AGG | General Equal Treatment Act – Protection Against Discrimination | No indirect discrimination against groups by the AI (e.g., in the assignment of shifts and hours) |
| BetrVG (Works Constitution Act) | Co-determination by the works council, if one exists | Consent Regarding the Allocation of Working Hours and Technical Monitoring |
| Employee Representation Law | Employee Participation in the Public Sector (BPersVG / State Laws) | Staff Council Instead of a Works Council for Public-Sector Employers |
| Employee Data Protection | Data Processing in the Employment Relationship (Art. 88 GDPR, § 26 BDSG; in the future, BeschDG) | Proportionality; proposed right to information regarding how AI works |
GDPR: Data Protection and Automated Decision-Making
The shift planning system processes sensitive data: work hours, availability, qualifications, and, in some cases, health information related to absences. The principles of data minimization and purpose limitation apply here. Use only the data you actually need for scheduling.
In the context of employment, Article 88 of the GDPR and Section 26 of the German Federal Data Protection Act
(BDSG) specify these requirements; a proposed Employee Data Act is also intended to enshrine a right to information regarding the functioning of AI systems in use. Article 22 of the GDPR is central: It gives employees the right not to be subject to a decision based solely on automated processing if such a decision significantly affects them legally or in a similar manner. A shift schedule generated fully automatically and published without review may fall under this provision, for example, if it assigns unpopular night shifts or denies requested vacation time.
In practice, it’s often exactly the opposite: An automatically generated shift schedule is perceived as objective, fair, and transparent. This eliminates the need for discussions—providing emotional relief for the person responsible for the shift schedule.
Article 22 of the GDPR does not prohibit the use of AI in shift planning. It requires that a human be responsible for the decision. As soon as a shift scheduler reviews, adjusts, and actively approves the AI’s recommendation, the decision is no longer considered “fully automated.”
EU AI Act: Risk Assessment of AI in Human Resources
The EU AI Act, Regulation (EU) 2024/1689 —has been in effect as the EU’s AI Regulation since August 2024 and classifies AI systems by risk. AI used for human resources management and task assignment generally falls into the “high-risk” category under Annex III. Such systems are subject to stricter requirements regarding transparency, documentation, and human oversight.
The “Digital Omnibus on AI” – Amending Regulation (EU) 2026/1744 (adopted on July 8, 2026, and published in the Official Journal on July 24, 2026) – has been in effect since July 27, 2026. It postpones the application of the high-risk obligations: Standalone high-risk systems as defined in Annex III—which include AI in human resources—must not comply with the requirements until December 2, 2027. For high-risk AI embedded in products as defined in Annex I, the deadline is August 2, 2028.
Important to note: The deadlines have been postponed, not the obligations.
The transparency requirements regarding AI use under Article 50 remain unchanged and take effect on August 2, 2026. The AI competency requirement under Article 4 has been in effect since July 27, 2026—and this applies to operators as well, meaning businesses that use an AI scheduling solution, not just the providers. Companies must ensure that shift schedulers understand the automated system and act in accordance with it.
In practice, this means more lead time, but the same requirements in terms of content. Anyone selecting an AI shift planning solution today should prioritize transparency and traceability—regardless of the exact deadline. Systems like Planerio’s automated shift scheduling, which disclose all criteria and decisions, meet foreseeable requirements without the need for retrofitting.
A custom-created Claude artifact or the shift schedule generated by ChatGPT probably doesn’t.
ArbZG: Working-Hour Limits Remain Mandatory
No AI can override the Arbeitszeitgesetz, the German Working Hours Act. Maximum working hours, minimum rest periods, and breaks remain in effect—as does the requirement to track working hours, regardless of the scheduling tool used. An AI-generated proposal that falls short of an 11-hour rest time is just as invalid as a manually created schedule with the same error.
Industry-specific constraints also come into play: In logistics, the driving and rest times specified in the Driver Personnel Ordinance apply; in manufacturing, collective bargaining agreements often dictate the terms; and in safety and emergency response, there are special rules governing operational readiness and on-call schedules. A good planning AI, such as Planerio’s, is aware of these constraints, including company-specific regulations; and plans within them, rather than ignoring them.
BetrVG: Co-determination by the works council
This rule applies only if there is a works council. Many smaller companies do not have one. In such cases, the right to co-determination does not apply, but the GDPR and the ArbZG remain binding. If a works council exists, it must be involved in the shift planning process when AI is used. Several regulations apply here:
According to § 87(1) of the Works Constitution Act (BetrVG) , the works council has a say in determining the scheduling and distribution of working hours (No. 2) as well as in the introduction of technical devices designed to monitor employees’ behavior or performance (No. 6). Planning software that records and analyzes working hours generally meets this criterion.
In addition, Section 90 of the Works Constitution Act (BetrVG) stipulates that the employer must inform the works council in a timely manner about plans for technical equipment and work procedures and consult with it. And according to § 80(3) of the Works Constitution Act (BetrVG) Consulting an expert is considered necessary when the works council must evaluate the introduction of AI. In practice, this means: Ensure that the process is initiated early and properly documented, ideally through a works agreement. In the public sector, this role is assumed by the staff council in accordance with the Federal Staff Representation Act (BPersVG) or the respective state staff representation laws.
What AI Can and Cannot Do in Shift Planning
The line is drawn between providing support and making decisions independently. This overview categorizes typical use cases:
What AI can do | What AI cannot do |
|---|---|
| Calculate draft shift schedules, which a person reviews and approves | Publish plans fully automatically, without anyone being able to intervene |
| Take into account availability, qualifications, and target hours | Analyzing health data or personal characteristics without a legal basis for analytics |
| Propose a fair distribution based on rules | Implementing the policy without a works council where co-determination applies |
| Automatically comply with legal and collective bargaining limits | Exceeding the ArbZG limits “on paper” |
| Identify Bottlenecks Early | Secretly Creating Performance or Behavioral Profiles |
| Assign shifts in a transparent and non-discriminatory manner | Systematically Discriminating Against Certain Groups (Violation of the AGG) |
Here’s an example from the retail sector: An AI that makes suggestions for peak hours across all stores and leaves the decision up to store management is not critical. A system that independently reassigns cashiers and only informs those affected afterward would be critical. The same applies in the care sector to the allocation of living areas, and in logistics to the assignment of routes to dispatchers.
Checklist: Legally Compliant AI Shift Planning in Seven Steps
Practical Tip: These points ensure the AI solution is robust – address them before implementation.
- Ensure that humans make the final decision. An AI suggestion is reviewed and actively approved; it is not published automatically.
- Involve the works council, if one exists. If there is a works council, inform and consult with it early on and, if possible, conclude a company agreement. If there is no works council, this step is not necessary.
- Review the data protection impact assessment. This is usually required when processing large amounts of employee data.
- Implement data minimization. Process only data relevant to planning; clearly define the purpose.
- Regulate data processing. Enter into a Data Processing Agreement (DPA) with the provider.
- Promote transparency. Employees know that AI is used and how it is used.
- Ensure traceability. The software discloses the rules it uses for planning.
Rule-Based AI Instead of a Black Box: What Businesses Should Keep in Mind
Not all AI is the same. When it comes to shift planning, the difference between probabilistic – that is, estimative – and rule-based systems is crucial.
Generic language models produce plausible but inconsistent results—and do not provide transparent justifications for them. This poses a risk for workforce planning: A result that cannot be explained or reproduced is difficult to justify to the works council and regulatory authorities. Rule-based, deterministic systems, on the other hand, operate according to clearly defined parameters and deliver the same result given the same initial conditions.
This is exactly where Planerio’s AI-assisted shift planning comes in. The automatic planning algorithm operates on a rule-based system and strictly adheres to both legal and operational requirements: It schedules based on target hours, never exceeds the specified work hour limit, takes into account rest times before and after shifts, maximum limits for consecutive shifts, and the necessary qualifications. Employees’ individual availability is factored in, and the underlying rules are visible and configurable in the interface – not hidden away in a black box.
The final decision rests with the user: Schedulers edit the draft, check for absences, and actively publish the schedule. This fulfills the requirement of Article 22 of the GDPR without sacrificing the time savings provided by automation. In addition, the integrated digital time tracking ensures that actual hours and working time limits align, and login via SSO and MFA protects the employee data being processed.
Frequently Asked Questions About AI-Powered Shift Planning
Generally, no. If the plan significantly affects employees, Article 22 of the GDPR applies and requires a human decision. In practice, this means that a planner reviews the AI-generated draft and approves it.
Where a works council exists and the software can allocate or evaluate working hours, the right to co-determination under Section 87 of the Works Constitution Act (BetrVG) applies. The works council must be notified in a timely manner and involved in accordance with co-determination requirements. If there is no works council, this obligation does not apply—however, the requirements of the GDPR and the ArbZG still apply regardless.
Only data necessary for planning, such as working hours, availability, and qualifications. The principles of data minimization and purpose limitation apply. Health data or other sensitive data require their own legal basis.
When employee data is processed on a large scale and in a systematic manner, a data protection impact assessment is usually required. If in doubt, consult with your data protection officer.
Under the EU AI Act, AI used for human resources management is generally considered high-risk, with heightened requirements for transparency, documentation, and human oversight.
Under the Digital Omnibus on AI (Regulation (EU) 2026/1744, in effect since July 27, 2026), these obligations apply to standalone high-risk systems only as of December 2, 2027; the transparency requirements under Article 50 remain unaffected and apply as of August 2, 2026. Those who rely on traceable, rule-based systems such as Planerio meet the requirements regardless of the effective date.
Note: This article reflects the legal situation as of July 2026 and is not a substitute for legal advice. Please consult a legal professional for an assessment of your specific case.
Learn what rule-based AI shift planning looks like.
Proven: over 4,000 sites already plan with Planerio – we’ll show you how to make it work in your organisation too.