কর্মক্ষেত্রে অ্যালগরিদমিক ব্যবস্থাপনা: এআই কি আপনার কর্মক্ষমতা মূল্যায়ন করতে পারে?

Algorithmic management, the use of software and AI to monitor, score and steer employees, is allowed in the Netherlands, but an algorithm may not be the sole basis for a decision that significantly affects you, such as dismissal, demotion or refusing a bonus. Article 22 of the GDPR reserves such decisions for a human who genuinely reviews them, the EU AI Act treats AI systems for evaluating workers as high-risk and prohibits AI that infers employees’ emotions, and under Article 7:669 of the Dutch Civil Code (BW) a dashboard score alone is never a dismissal ground.

This page explains how these systems work, what the GDPR, the AI Act, the Works Councils Act and Dutch dismissal law require, what the courts have decided, and how employees and employers should act.

What is algorithmic management?

Algorithmic management means software, rather than a supervisor, observes, measures and often judges work. It is common in warehouses, contact centres and platform work, and increasingly in office roles through productivity tools; a dashboard that scores staff on automatically collected data is enough to raise the legal questions.

Such systems count well: tasks completed, handling times, response times, script adherence, hours logged. They cannot see why a number moved, for example because an employee spent three weeks training a colleague or handled the most complex client. Measuring performance is lawful; the GDPR and Dutch employment law set conditions on how it is measured and, above all, how the output is used.

How do these systems score performance?

Three stages matter: data is collected, a model turns it into a score, and someone or something acts on it. Your rights attach to each stage.

Every data point is personal data and must be necessary and proportionate for a specified purpose under the GDPR. Keystroke logging or periodic screenshots to see who works hard are very difficult to justify, and the Dutch Data Protection Authority (Autoriteit Persoonsgegevens, AP) has been critical of such blanket monitoring.

A rule-based system (for example: flag an employee whose handling time exceeds a threshold three times a week) is rigid but readable. A machine-learning model trained on historical data is more powerful but opaque. An employer that cannot explain how a score came about cannot give meaningful information about the logic, cannot defend the assessment and cannot show it is free of discrimination; in practice, if you cannot explain it, you cannot rely on it.

What does the GDPR require?

The GDPR is the most practical instrument, because it applies to every employer processing employee data. Four rules do most of the work: the ban on solely automated significant decisions, transparency, a lawful basis with a data protection impact assessment (DPIA), and limits on monitoring.

No significant decision by machine alone

Article 22 of the GDPR gives you the right not to be subject to a decision based solely on automated processing that produces legal effects or similarly significantly affects you. Dismissal, non-renewal, demotion, withholding a bonus or refusing a promotion all qualify, and so can a score that decides which shifts you are offered.

A human who merely confirms the system’s proposal does not take the decision outside Article 22. Meaningful involvement requires a reviewer with authority to decide differently, the competence to judge and the time to do it. Where Article 22 applies, you have the right to human intervention, to express your view and to contest the decision.

Transparency and access

Employees must be told in advance that the system exists and what it does. Where automated decision-making takes place, the employer must give meaningful information about the logic involved and the consequences, and you can request it, with a copy of your data, scores and categories, through an access request under Article 15 of the GDPR.

Meaningful means substance: which factors were used and how they were weighed, in terms you can understand and act on. The employer must respond within one month, which can be extended in complex cases.

Lawful basis and DPIA

Employee consent is rarely valid because of the dependency in the employment relationship. Employers therefore rely on necessity for the contract, a legal obligation or a legitimate interest, the latter requiring a documented balancing test.

Systematic and extensive evaluation of people through automated processing requires a DPIA before the system goes live. It must describe the processing, justify necessity and proportionality, identify the risks to employees and set out the measures that address them.

How far may monitoring go?

In Bărbulescu v Romania the European Court of Human Rights held that monitoring an employee’s communications engages the right to private life and correspondence in Article 8 ECHR. The employer must inform employees in advance of the nature and extent of monitoring, pursue a legitimate aim and use the least intrusive means; covert monitoring requires a concrete suspicion, a limited duration and documented reasons.

For each data stream, the employer must be able to explain why it is needed, why a lighter measure will not do, and whether employees were informed. Continuous screen recording, keystroke logging and location tracking outside working hours usually fail that test.

Health data and protected leave

Health data is special category data. An employer may record that an employee is absent and for how long, but the diagnosis and limitations belong with the company doctor (bedrijfsarts). A system that infers a health condition from absence patterns, typing speed or tone of voice processes health data by inference.

A model that treats pregnancy, parental leave or care leave as underperformance turns a statutory right into a disadvantage. Breaches of the core GDPR principles and of data subjects’ rights fall in the highest fine tier the GDPR provides, and an assessment produced in breach of these rules is hard to rely on in an employment dispute.

What does the AI Act add?

The EU AI Act applies in stages and has two layers that matter at work. The ban on AI systems that infer the emotions of people in the workplace already applies, unless the system is intended for medical or safety reasons.

Software that scores the mood, stress, engagement or sincerity of employees from voice, face or writing is therefore prohibited; productivity management does not fall within the medical or safety exception. Sentiment analysis of what a customer says is a different question from inferring the employee’s emotions.

The second layer is the high-risk regime. AI systems used for decisions on promotion or termination, for allocating tasks based on behaviour or personal traits, or for monitoring and evaluating workers’ performance and behaviour are classified as high-risk. Under the AI omnibus, Regulation (EU) 2026/1744, in force since 27 July 2026, the application dates for these high-risk obligations have been postponed; the prohibitions already apply. Employers that deploy a high-risk system at work must inform workers’ representatives and the affected workers before putting it into service. See our article on কৃত্রিম বুদ্ধিমত্তার আইনি দিক এবং ইইউ এআই আইন.

Who is responsible: the supplier or the employer?

The employer. Under the GDPR the employer decides why and how employee data is processed and is the controller; the supplier is usually a processor bound by a processing agreement.

Under the AI Act an employer using a purchased tool is normally a deployer: it must use the system according to the instructions, ensure human oversight by competent staff, monitor its operation and inform workers. An employer that rebrands the system, modifies it substantially or changes its purpose can become a provider, with much heavier duties. Read procurement contracts with these roles in mind and be sceptical of any product that cannot explain its own output.

What applies to platform work?

The EU Platform Work Directive (Directive (EU) 2024/2831) contains a chapter on algorithmic management. It requires platforms to inform workers about automated monitoring and decision-making, restricts processing of data such as emotional state and private conversations, requires human oversight, and gives a right to an explanation and human review of significant decisions such as account suspension, partly regardless of employee status.

Until Dutch implementing legislation is in force, the GDPR and the AI Act remain the operative rules for platform workers.

Can an algorithm end your employment contract?

No. Under Article 7:669 BW dismissal requires a reasonable ground from a closed list and prior permission from UWV or a decision of the subdistrict court (kantonrechter), depending on the ground.

For unsuitability (Article 7:669(3)(d) BW) the employer must show that it informed the employee of the shortcomings in good time and gave sufficient opportunity to improve, and that the unsuitability is not due to insufficient care for training or working conditions. A dashboard reading is not an improvement process; the court will ask what was discussed, when and with what support. See our guide to ডাচ কর্মসংস্থান আইন.

Discrimination law looks at the outcome, not the intention. A system that disadvantages employees on grounds such as sex, race, age, religion, disability or chronic illness is unlawful under the General Equal Treatment Act (Algemene wet gelijke behandeling) and the specific acts on age and disability, whatever was intended. Historic training data and raw availability metrics are the usual causes, especially for part-timers and employees who took leave or were ill.

পূর্ত পরিষদকে কি সম্মতি দিতে হবে?

Yes. Under Article 27(1) of the Works Councils Act (Wet op de ondernemingsraden) the works council must consent to arrangements on staff appraisal, on processing and protection of employees’ personal data, and on facilities aimed at or suitable for monitoring attendance, conduct or performance; an AI performance system usually touches all three.

A decision taken without the required consent can be declared void by the works council. For employees who feel a system was imposed on them, the works council is often the quickest route to change.

What have the courts decided?

On 4 April 2023 the Amsterdam Court of Appeal ruled on access requests by drivers of Uber and Ola. In the Uber case (ECLI:NL:GHAMS:2023:793) the court found that the deactivation of three drivers’ accounts was based solely on automated processing, because no meaningful human involvement was shown, and ordered Uber to provide information about the decision-making; a trade-secret objection did not justify a full refusal.

In the Ola case (ECLI:NL:GHAMS:2023:804) the court held that a fraud probability score significantly affects drivers, so the right to information about automated decision-making applies. Employers should assume the same standard applies to a performance algorithm: “the model decided” is not an explanation.

In the SyRI case (ECLI:NL:RBDHA:2020:865, 5 February 2020) the District Court of The Hague held that the legislation behind a government fraud risk-indication system breached Article 8 ECHR because it was insufficiently transparent and verifiable. A system whose workings cannot be examined cannot be challenged. Our article on এআই এবং ফৌজদারি আইন takes these questions further.

How do you challenge an automated assessment?

Document your position first: the work the assessment overlooks, unusual periods, and any absence, leave or reassignment that depresses the figures. Notes made at the time carry far more weight than a reconstruction months later.

Then write to your employer that you exercise your GDPR rights: ask for the data used, an explanation of the logic and a human review by someone with authority to decide differently, and ask for correction of inaccurate data. The employer must respond within one month, extendable in complex cases. If the response is inadequate, you can complain to the AP or go to court; where the assessment supports dismissal or a change in terms, the employment law route is usually more effective and has short deadlines. See our article on এআই এবং বিগ ডেটার যুগে জিডিপিআর-এর অধীনে ডেটা গোপনীয়তা.

একজন কর্মচারী হিসেবে আপনার কী কী যাচাই করা উচিত?

• Were you told in advance that an automated system assesses your work and what it measures?
• Was any decision with a significant effect on you reviewed by a person who could decide differently?
• Have you requested your data, scores and an explanation of the logic under the GDPR?
• Do the figures reflect leave, illness, part-time work or unusual tasks that depress your output?
• Did the works council consent to the system?

নিয়োগকর্তা হিসেবে আপনার কী কী যাচাই করা উচিত?

• Is the purpose of the system written down, including which decisions it will influence?
• Was a DPIA carried out and the lawful basis documented before the system went live?
• Does the system infer emotions (prohibited) or fall within the AI Act’s high-risk category?
• Has the works council consented under Article 27 of the Works Councils Act, and were workers informed?
• Can reviewers overrule the output, with the context and time to do so?
• Do you audit outcomes for part-timers and employees who took leave or were ill?

What can we do for you with algorithmic management?

আমাদের কর্মসংস্থান আইনজীবি advise employers deploying AI in performance management and employees facing an automated assessment:

• We review the system against the GDPR, the AI Act and Dutch employment law before it goes live.
• We prepare the request for consent to the works council and the information to workers.
• We draft the GDPR access request and the request for human review of an assessment.
• We assess whether a performance file based on system data supports dismissal under Article 7:669 BW.
• We review supplier and processing agreements for the allocation of GDPR and AI Act duties.

সারাংশ

• Algorithmic management is lawful, but significant decisions may not be based solely on automated processing (Article 22 GDPR).
• AI that infers employees’ emotions is prohibited; AI used to evaluate workers is high-risk under the AI Act.
• Employees can obtain their data, scores and an explanation of the logic, and can contest decisions.
• The works council must consent to appraisal and monitoring systems under Article 27 of the Works Councils Act.
• A dashboard score alone does not establish a dismissal ground under Article 7:669 BW.

সচরাচর জিজ্ঞাস্য