Predictive Rostering – Legality

Predictive rostering means using historical data, employee availability, workload forecasts, attendance patterns, customer demand, skills, leave history, or AI/algorithmic tools to predict staffing requirements and automatically prepare employee shifts or duty rosters.

In India, there is no single statute specifically prohibiting predictive rostering. Its legality depends on how the roster is designed and used, particularly with respect to working-hour limits, rest periods, overtime, discrimination, privacy, contractual rights, and procedural fairness.

1. Working-hour limits

An employer cannot use an algorithm merely because it predicts that additional labour is required. The resulting roster must comply with applicable employment legislation.

For factory workers, statutory provisions concerning maximum working hours, weekly hours, rest intervals and overtime are particularly important. The Supreme Court in Gujarat Mazdoor Sabha v. State of Gujarat, (2020) 10 SCC 459 held that statutory protections concerning working hours and overtime cannot simply be overridden through executive exemption when doing so defeats the protective purpose of labour legislation. The Court specifically considered provisions concerning daily and weekly working-hour limits and overtime.

Application to predictive rostering: An AI system cannot lawfully schedule an employee for excessive hours simply because its prediction shows a shortage of staff.

2. Changes to established working hours

In Associated Cement Staff Union v. Associated Cement Co. Ltd., AIR 1964 SC 914, the Supreme Court considered disputes concerning working hours and recognised that working conditions can change when circumstances justify the change, but industrial adjudication must consider the relevant circumstances and justification.

Therefore, predictive rostering may be used to reorganise shifts, but a substantial alteration of established working conditions may still have to comply with applicable employment contracts, standing orders, settlements and labour legislation.

3. Discrimination in rostering

An apparently neutral algorithm can create discriminatory outcomes. For example, if historical data causes the system to repeatedly allocate undesirable night shifts to women, a particular religious group, disabled employees or another protected category, the employer may face equality or discrimination issues.

In R. Vasantha v. Union of India, the Madras High Court examined restrictions on women's working hours and held that differential treatment based merely on sex could violate constitutional equality principles where there was no rational connection between the classification and the objective sought.

Similarly, Triveni K.S. v. Union of India considered restrictions concerning women's employment during particular hours and examined them through the constitutional principles of equality and non-discrimination.

Application: A predictive roster should therefore be tested for discriminatory patterns rather than relying only on the claim that the algorithm itself is "neutral."

4. Privacy and employee data

Predictive rostering commonly uses personal information such as:

• attendance records;
• leave history;
• availability;
• location;
• working patterns;
• overtime records;
• performance information;
• sometimes health or family-related information.

The Supreme Court's decision in Justice K.S. Puttaswamy (Retd.) v. Union of India, (2017) 10 SCC 1 recognised privacy as a fundamental right and explained that an interference with privacy must satisfy requirements including legality, legitimate purpose and proportionality.

The Digital Personal Data Protection Act, 2023 also establishes rules governing processing of digital personal data, subject to its applicable provisions and exemptions.

Consequently, an employer implementing predictive rostering should identify what employee data is being collected, why it is necessary, how long it is retained and who can access it.

5. Algorithmic decisions should not become completely unreviewable

A significant legal risk arises where the employer says:

"The computer generated the roster, so management cannot change it."

That approach is problematic. The employer remains responsible for complying with employment law.

For example, if an algorithm repeatedly schedules one employee for undesirable shifts because of historical attendance data, the employee should have a mechanism to raise the issue and obtain human review.

The principle is particularly important where the roster affects:

• wages;
• overtime;
• promotion opportunities;
• working conditions;
• disciplinary consequences;
• family or religious obligations; or
• health and safety.

6. Statutory records and evidence

Predictive rostering also creates evidentiary issues.

In R.M. Yellatti v. Assistant Executive Engineer, (2006) 1 SCC 106, the Supreme Court discussed the significance of employment records and the circumstances in which adverse inference may arise when relevant records in an employer's possession are not produced. This principle was recently reiterated in Kishan Sharma v. Management of MCD (2026) concerning muster-roll and employment records.

Therefore, employers using automated rostering should maintain reliable records showing:

• the roster generated;
• subsequent modifications;
• working hours;
• attendance;
• overtime;
• leave;
• reasons for significant changes.

7. Important case laws

Conclusion

Predictive rostering is not inherently illegal in India. Its legality depends primarily on how the technology is used and what consequences the roster produces.

A legally safer predictive-rostering system should:

1. comply with statutory working-hour and rest requirements;
2. calculate overtime correctly;
3. avoid discriminatory scheduling patterns;
4. respect applicable employment contracts and standing orders;
5. process employee data lawfully and proportionately;
6. maintain accurate roster and attendance records;
7. provide a mechanism for employees to challenge errors; and
8. retain human oversight over significant employment decisions.

Thus, the main legal issue is not simply "Can an employer use AI to make a roster?", but rather "Does the AI-generated roster comply with labour, equality, privacy and employment-law requirements?"