Algorithmic Management and Employment Law: The Limits of Performance Measurement and the Data Protection Risk

Artificial intelligence and machine learning are now embedded in HR functions, from monitoring employee performance and allocating work to automated termination decisions. Voice and sentiment analysis in call centres, step tracking for warehouse couriers, commit-level metrics for developers and CRM-based sales analytics have produced what is now described as algorithmic management.

We look below at how companies can deploy these technologies while remaining compliant with Turkish employment law and the Personal Data Protection Law (KVKK), and at the legal exposure created when the limits are exceeded.

Where Algorithmic Management Sits in Employment Law

An employer's capacity to give instructions, direct the work and expect performance improvement under the employment contract is known as the right of management. But where a company relies on an algorithm to make deductions from pay or to open the way to dismissal, that right has limits.

Proportionality and good faith: If HR software records an employee's bathroom breaks to the second, that engages the principles of good faith and proportionality under Article 2 of the Civil Code.

Protection of personality rights: Under Article 58 of the Code of Obligations, minute-by-minute GPS tracking that subjects an employee to excessive stress and pressure can give rise to a claim for non-pecuniary damages. The employer's specific duty to protect the employee's personality is separately governed by Article 417 of the same Code.

The Data Protection Dimension

Employers often assume that because the hardware and software belong to the company, monitoring is unrestricted. Electronic performance data is personal data.

The consent trap: The argument that employees signed a consent form is weak. In the Board's decisions and guidance, consent in an employment relationship is treated as unlikely to be freely given, because of the dependency between employee and employer. Consent obtained on that basis is generally invalid.

Legitimate interest: Monitoring should instead rest on the legitimate interest ground in Article 5/2(f). That requires a documented balancing test: does the company's interest in productivity outweigh the intrusion into the employee's privacy?

The Restriction on Automated Decision-Making

Article 11/1(g) of the Personal Data Protection Law is the most significant constraint on algorithmic management. It gives every individual the right to object to a result produced to their detriment solely through analysis by automated systems.

Example: Where software compares a salesperson's performance against colleagues and automatically issues a written warning, or where a deduction flows through to payroll on the basis of algorithmic data without human review, the process is unlawful.

The practical answer: The conclusion from the statute is straightforward. The algorithm's output should remain decision support presented to HR, not the decision itself. A disciplinary or dismissal decision must carry the signature of an actual manager, and it must be possible to show that the manager assessed the file independently. "The system calculated it that way" is not a defence under Article 11/1(g).

Evidence in Unfair Dismissal Proceedings

Poor performance is a valid ground for termination under Article 18 of the Labour Act; it is not a ground for summary dismissal for cause. A dismissal on that basis therefore leaves the employee's notice and severance entitlements intact. What the employer is trying to achieve is not avoidance of severance but a finding that the dismissal was valid and dismissal of the reinstatement claim. Missing this distinction leads to a defence built on the wrong footing from the start.

Where validity is in issue, the principal evidence will often be the log report generated by the software.

If that data was produced by monitoring software deployed in breach of data protection rules, it constitutes unlawfully obtained evidence under Article 189 of the Code of Civil Procedure and cannot be taken into account. Without it, the dismissal is likely to be held invalid and the employer exposed to reinstatement remedies.

Practical Guidance for Boards

1. Where algorithms assess personnel, the parameters and metrics should be set out in a written algorithmic management policy communicated to staff. Employees should know which scores carry consequences.

2. Screen monitoring software installed on work devices, whether in the office or at home, should be visibly indicated to the user during the session.

3. Never give automation the authority to terminate employment or impose financial penalties on its own. The algorithm belongs in a decision-support role.

This article reflects general risks arising from digitalised HR policies and does not constitute a legal roadmap for any specific organisation.

Last updated: 10 August 2026.