/ THE SHORT ANSWER
- 01Do not infer qualifications or availability from unrelated records.
- 02Compare burden distribution as well as coverage.
- 03Make overrides and challenges part of the workflow.
/ dotSuper point of view
dotSuper analysis: a useful roster system exposes the reasons and consequences of allocation choices.
Define the planning problem before the algorithm
Some concern whether a person is qualified for a station.
Others concern availability, team composition or how unpopular shifts are shared.
Put these in separate fields before considering any optimisation objective.
Section 87 addresses co-determination in working-time arrangements.
[1] Section 95 concerns personnel selection guidelines and explicitly covers AI involvement in their preparation.
[2] These provisions raise different questions, so do not treat a single conversation about software procurement as resolving every employee-related decision.
Describe the intended use in a short statement.
For example, the tool suggests staffing options for an approved production plan, while a named planner releases the roster.
The statement should explain what the system may influence and what information it is forbidden to infer.
Separate hard constraints from negotiable preferences
Record availability through the agreed planning route.
Avoid deriving availability from badge records, email activity or assumptions about an employee's family circumstances.
Hard constraints should make an assignment impossible when they are not satisfied.
Preferences should influence the comparison between otherwise feasible options.
Mixing both into a weighted score can permit a serious constraint to be traded against a small improvement elsewhere.
Document conflicts rather than inventing a feasible answer.
If no qualified employee is available, the system should show the gap and the affected production commitment.
A planner may then change the production plan or seek an authorised staffing solution.
Quietly relaxing a qualification condition would conceal the actual operational decision.
Review the evidence behind each suggested assignment
The following is an original review structure, not a claim about mandatory fields in German law.
Keep the reason attached to the released roster version.
Recomputing an explanation later may describe different inputs or weights.
Staff should be able to ask about the actual decision that affected them, not a simulation generated after the complaint arrived.
Minimise sensitive details in the explanation.
A colleague need not learn another person's personal circumstances to understand why their own assignment was reviewed.
Design separate views for the employee, planner and authorised HR staff, with each view supporting a specific task.
| Question | Inspect | If unresolved |
|---|---|---|
| Is the person eligible? | Current qualification record | Exclude assignment |
| Is availability reliable? | Approved availability input | Ask the planner |
| Why this person? | Visible allocation reason | Review the preference rule |
| Who bears the burden? | Comparable undesirable shifts | Compare another roster |
| Can it be challenged? | Named review route | Do not release automatically |
A hypothetical North Rhine-Westphalia assembly plant
Two possible rosters provide identical coverage.
The first repeatedly assigns unpopular late shifts to the same four employees because their historical acceptance rate is higher.
The second roster distributes those shifts across the eligible, available group according to an agreed rule.
It requires one additional planner conversation to handle a preference conflict.
That extra effort should be visible in the comparison rather than dismissed as a failure of automation.
The example shows why historical behaviour is a questionable shortcut.
Employees may have accepted earlier shifts under circumstances that no longer apply.
Ask whether past acceptance is an approved and appropriate planning input before using it, and provide a way to correct information that the system treats as persistent.
Check the current AI schedule without flattening scope
[3] That reported schedule does not classify this hypothetical roster tool, nor suspend other applicable employment and data protection requirements.
Assess the actual functionality, influence and intended purpose with the responsible specialists.
A tool that organises available information can differ materially from one that ranks workers, predicts reliability or determines assignments.
Vendor labels such as assistant and optimisation do not answer that assessment.
Use the review to improve the design now.
Remove unnecessary employee scoring, require reasons for planner overrides and preserve a manual route when data is incomplete.
These operating choices remain useful even when the legal analysis concludes that different requirements apply to different parts of the system.
Measure the quality of released decisions
Track how often planners correct eligibility data, how many assignments are challenged and whether unresolved conflicts recur for particular work areas.
Interpret these measures with the people affected, rather than using them to rank individual employees.
Compare the proposed roster with an independently prepared alternative before relying on automatic release.
Look for hidden assumptions about overtime, training time and station interchangeability.
A plausible timetable can be operationally impossible when the model has treated two similar machines as requiring identical skills.
The first useful implementation is a transparent comparison screen.
Show the alternatives, the unfilled positions and the reasons behind each choice.
Release authority should sit with someone who understands the production consequences and can respond meaningfully when an employee identifies an error or an overlooked constraint.
What this page cannot conclude
- 01No employment-law, collective-agreement or AI risk classification is determined for a particular employer.
- 02The scenario uses fictional employee groups and simplified planning constraints.
- 03This article was researched and drafted with AI assistance. Sources and limitations are provided for scrutiny; it is not an independent professional review or a compliance certification.
Sources
- 01Betriebsverfassungsgesetz, section 87Federal Ministry of Justice and Federal Office of Justice · accessed Sep 15, 2026
- 02Betriebsverfassungsgesetz, section 95Federal Ministry of Justice and Federal Office of Justice · accessed Sep 15, 2026
- 03Timeline for the Implementation of the EU AI ActEuropean Commission AI Act Service Desk · accessed Sep 15, 2026
This article was researched and drafted with AI assistance. Sources and limitations are provided for scrutiny; it is not an independent professional review or a compliance certification.
Our editorial standard · Found an error? Send a correction with its source.
/ CITE OR SHARE THIS GUIDE
Make the evidence easy to verify.
When you reference this guide, link to its canonical URL. That gives readers one stable place for the evidence, limitations and future updates.
dotSuper Research Desk. (September 15, 2026). Keep AI Shift Planning Explainable to Employees. dotSuper. https://dotsuper.net/feeds/applied-systems/germany-ai-shift-planning-human-decisions