Civil Law And Artificial Intelligence Employment Replacement Compensation Claims In Europe

1. Introduction

Artificial intelligence is increasingly capable of performing functions previously carried out by employees—for example, customer service, document review, accounting, recruitment, scheduling, translation, coding, logistics, quality control and certain managerial functions.

This creates a distinctive European civil/employment-law problem:

When an employer replaces employees with AI and the employees suffer financial or non-financial loss, when can they claim compensation?

There is not yet a major European court judgment directly establishing a general right to compensation merely because an employee's job was replaced by AI. The legal position is instead constructed from existing rules on redundancy, collective consultation, discrimination, automated decision-making, privacy, employment status, wrongful dismissal, and damages.

The EU AI Act expressly treats AI used in employment and worker management—including systems affecting recruitment, promotion, evaluation, allocation of tasks and termination of work-related contractual relationships—as high-risk AI in the relevant circumstances. (EUR-Lex)

Therefore, the central legal question is generally not whether an employer is allowed to use AI, but whether the resulting employment termination complied with the applicable employment, equality, data-protection, consultation and civil-liability rules.

2. Meaning of AI Employment Replacement

AI employment replacement occurs when an employer introduces an AI system or AI-supported technology and consequently:

eliminates existing jobs;

reduces the number of employees;

terminates individual employment contracts;

refuses to renew contracts;

transfers employees to lower-paid positions;

replaces employees with AI-assisted machinery;

reduces working hours;

removes particular occupational functions;

selects particular employees for redundancy through an algorithm;

uses AI to evaluate employees before deciding whom to dismiss;

replaces human managers with algorithmic management; or

reorganises the workforce around AI systems.

A distinction must be made between:

A. Genuine technological redundancy

The employer genuinely restructures its business because AI makes particular jobs unnecessary.

B. AI-assisted selection

AI is used to determine which employees will be dismissed.

C. Automated dismissal

The AI system itself makes or materially determines the termination decision.

D. Discriminatory AI replacement

The AI disproportionately removes or selects workers because of protected characteristics such as age, sex, disability or racial/ethnic origin.

E. Unlawful automated processing

The employer uses employee data in an automated system contrary to GDPR or applicable employment legislation.

These categories can produce very different compensation consequences.

3. Basic European Legal Framework

Several legal regimes may operate simultaneously.

3.1 EU AI Act

The EU AI Act, Regulation (EU) 2024/1689, identifies AI systems used in employment and worker management—including systems used for recruitment, decisions affecting employment relationships, promotion, termination, task allocation and monitoring—as potentially high-risk systems under Annex III. (EUR-Lex)

The AI Act also requires appropriate human oversight for high-risk AI systems. Article 14 requires systems to be designed so that natural persons can effectively oversee them. (EUR-Lex)

Consequently, an employer cannot necessarily defend an unlawful employment decision simply by saying:

"The algorithm made the decision."

The legal responsibility remains connected to the human organisation deploying the system.

4. Collective Redundancy Protection

The most important legal framework for mass AI replacement is the Collective Redundancies Directive 98/59/EC.

Where the relevant national thresholds are satisfied, an employer contemplating collective redundancies must consult workers' representatives in good time. The consultation must consider:

avoiding redundancies;

reducing the number of affected employees;

mitigating consequences;

redeployment;

retraining; and

social measures. (EUR-Lex)

This is particularly important for AI replacement.

Example

Suppose a company announces:

"We are installing an AI customer-service system and therefore terminating 300 customer-service employees."

The legal issue is not simply whether AI is more efficient.

The employer may have to consider and discuss:

whether all 300 positions need to disappear;

redeployment;

retraining;

phased implementation;

alternative employment;

selection criteria;

timing;

redundancy payments; and

other measures reducing the consequences.

The Directive specifically requires employers to provide information concerning the reasons for projected redundancies, categories and numbers of workers, timing, selection criteria and the method of calculating certain redundancy payments. (EUR-Lex)

5. Compensation Is Not Automatically Payable Merely Because AI Replaces a Job

This is a crucial principle.

European law does not generally create an automatic civil damages claim merely because technology eliminates an employee's position.

An employer may lawfully restructure its business for technological reasons, subject to national employment law.

Compensation becomes much more legally significant where there is an additional wrong, such as:

unlawful dismissal;

failure to comply with redundancy procedures;

failure to consult;

discrimination;

discriminatory algorithmic selection;

unlawful processing of employee data;

unlawful automated decision-making;

breach of contractual obligations;

violation of collective bargaining rights;

retaliation for exercising statutory rights; or

other actionable civil/employment-law misconduct.

6. The AI Replacement Compensation Claim

A useful legal formula is:

AI deployment → employment restructuring → termination → legal violation → damage → causation → compensation

The employee generally has to establish the applicable legal basis.

Step 1 — Employment relationship

Was the claimant actually an employee or worker?

Step 2 — AI involvement

Did AI cause, influence or merely support the employment decision?

Step 3 — Legal duty

What duty did the employer owe?

For example:

fair dismissal;

consultation;

equality;

privacy;

data protection;

contractual good faith;

occupational protection.

Step 4 — Breach

Was the duty breached?

Step 5 — Causation

Did the breach cause the employee's loss?

Step 6 — Damage

Possible damage includes:

lost wages;

lost benefits;

loss of employment;

loss of pension contributions;

loss of career opportunity;

discrimination-related damage;

non-material harm;

unlawful-processing damage.

Step 7 — Remedy

Depending on national law:

reinstatement;

compensation;

back pay;

redundancy payment;

damages;

declaration of unlawfulness;

reversal of a decision;

correction/deletion of data;

procedural remedies.

7. Case Law

Case 1 — Fujitsu Siemens Computers Oy v Finland, C-188/03

This is one of the most important cases for AI-driven mass redundancies.

The CJEU interpreted the Collective Redundancies Directive and held that the employer's decision-making and consultation obligations cannot simply be postponed until individual dismissals are formally issued.

The Court treated the employer's decision to terminate employment relationships as legally significant for the consultation framework. (curia)

Relevance to AI replacement

Suppose management has already decided:

"AI will replace this entire department."

It cannot necessarily wait until termination letters are issued before engaging in legally required consultation.

Principle

The technological decision and the resulting redundancies cannot be artificially separated to defeat consultation rights.

This can support claims for remedies where an employer bypasses mandatory redundancy procedures.

8. Case 2 — Danfoss, Case 109/88

In Danfoss, the CJEU considered discriminatory pay practices and the problem of opaque employer decision-making.

The case established important principles concerning the burden of proof where an apparently neutral and insufficiently transparent employment system produces discriminatory outcomes. (Infocuria)

AI relevance

An AI system may rank employees according to:

performance scores;

productivity;

attendance;

promotion probability;

"retention value";

predicted future performance.

If the employer cannot explain the relevant criteria and the results show a discriminatory pattern, the evidential burden may become important.

Example

An AI redundancy system repeatedly selects older employees because it predicts that younger employees have a longer "future contribution period."

That could create an age-discrimination issue even though the employer describes the algorithm as "neutral."

Principle

Algorithmic neutrality in appearance does not necessarily eliminate indirect discrimination.

9. Case 3 — Enderby v Frenchay Health Authority, C-127/92

In Enderby, the CJEU addressed indirect sex discrimination and the evidential consequences of significant disparities between predominantly female and predominantly male occupational groups.

The Court recognised that apparently neutral employment structures can produce a prima facie discrimination problem when substantial disparities exist. (curia)

AI replacement relevance

AI restructuring could disproportionately affect occupational groups dominated by one sex.

For example:

administrative roles;

care-related roles;

clerical work;

technical occupations.

If an AI transformation eliminates predominantly female positions while preserving predominantly male positions, the employer may face an indirect discrimination claim depending on the applicable national and EU rules.

Compensation

Potential remedies can include compensation under the relevant national equality legislation.

The claim is not:

"AI is unlawful."

It is:

"The AI restructuring produced an unlawful discriminatory employment outcome."

10. Case 4 — Meister v Speech Design Carrier Systems GmbH, C-415/10

In Meister, the CJEU considered access to information in a discrimination dispute.

The Court held that EU equality law does not automatically give an unsuccessful applicant a right to receive all information about another applicant, but refusal to provide information can be one factor relevant to establishing a presumption of discrimination. (Infocuria)

AI relevance

This is particularly significant for opaque AI employment systems.

Imagine an employer says:

"The AI determined that you were unsuitable."

The employee may want to know:

what criteria were used;

what data were used;

whether comparable workers were treated differently;

whether protected characteristics were indirectly reflected;

who reviewed the AI result;

whether a human decision-maker independently assessed the case.

Meister supports the broader importance of evidence and transparency in discrimination litigation, although it does not establish a general right to obtain an AI algorithm's source code.

11. Case 5 — SCHUFA Holding (Scoring), C-634/21

This is an especially important modern AI/algorithmic-decision case.

In SCHUFA, the CJEU considered automated scoring under Article 22 GDPR. The Court addressed circumstances in which an automated probability calculation could itself fall within the prohibition concerning automated individual decision-making where it has a decisive effect on a person. (Infocuria)

Although the case concerned credit scoring rather than employment, its reasoning is highly relevant to algorithmic employment decisions.

Employment example

An employer develops an AI score:

Employee A: 93% retention value
Employee B: 42% replacement value

The employer automatically dismisses employees below 50%.

The legal question becomes whether the AI output merely assists a genuinely independent human decision or effectively determines the employment outcome.

Importance

SCHUFA demonstrates why the distinction between:

AI assistance

and

AI-determined decision

can be legally important.

12. Case 6 — Bărbulescu v Romania, ECtHR

The European Court of Human Rights examined workplace monitoring and employee privacy in Bărbulescu v Romania.

The Grand Chamber stressed the need for appropriate safeguards when employers monitor employee communications and the importance of balancing workplace interests against privacy rights. (HUDOC)

AI replacement relevance

Modern employers can use AI to analyse:

emails;

keystrokes;

productivity;

attendance;

communications;

facial expressions;

voice;

behavioural patterns.

If such monitoring is then used to select employees for redundancy, the employee may have a privacy/data-protection argument in addition to an employment claim.

Principle

AI-based workforce surveillance does not exist outside fundamental-rights and privacy law.

13. Case 7 — Uber Systems Spain, C-434/15

In Asociación Profesional Elite Taxi v Uber Systems Spain, the CJEU examined the relationship between a digital platform and the underlying service and recognised the substantial organisational role exercised by the platform. (Infocuria)

This was not an AI redundancy case.

Its importance is structural.

AI relevance

Modern AI employers may argue:

"The worker is not really managed by us; the algorithm merely operates the platform."

European law increasingly looks at the actual substance of the relationship, rather than simply accepting technological or contractual labels.

This approach is reinforced by the EU Platform Work Directive, which requires employment status to be assessed primarily according to facts concerning the actual performance of work, including automated monitoring and decision-making. (EUR-Lex)

14. Case 8 — B v Yodel Delivery Network, C-692/19

The CJEU considered the classification of platform couriers and the concept of "worker" under EU working-time legislation.

The case concerned a courier operating under a services agreement and issues including substitution and work for other businesses. (Infocuria)

AI replacement relevance

AI-driven businesses may attempt to avoid employment obligations by describing individuals as:

contractors;

freelancers;

independent service providers;

platform participants.

The legal status may nevertheless depend upon the factual relationship.

This becomes important where an AI business restructures its workforce and claims that the displaced individuals were never employees.

15. New EU Platform Work Rules

Directive (EU) 2024/2831 is particularly important for algorithmic management.

It expressly regulates automated monitoring and automated decision-making in platform work. The Directive requires safeguards concerning significant decisions affecting workers, including contractual status and termination. (EUR-Lex)

Workers can obtain explanations for significant automated decisions and request review.

Where an unlawful automated decision cannot be rectified, the Directive provides for adequate compensation for damage sustained in the circumstances covered by its Article 11 framework. (EUR-Lex)

This is significant because it represents a move from the traditional principle:

"Human employer makes employment decision."

toward a legal framework that expressly regulates:

"Algorithm materially participates in employment decision."

16. AI Replacement and Collective Redundancy

Consider a company employing 1,000 people.

It introduces an AI system capable of replacing 250 employees.

The employer announces:

"250 employees will be made redundant."

Under the Collective Redundancies Directive, where the applicable thresholds are met, the employer must undertake consultation concerning matters including avoiding or reducing redundancies and mitigating their consequences. (EUR-Lex)

Possible measures

The consultation could concern:

retraining;

redeployment;

reduced working hours;

voluntary separation;

early retirement where lawful;

alternative positions;

internal mobility;

phased AI implementation;

severance arrangements.

Therefore, AI replacement can become a redundancy-law issue rather than simply a technology-law issue.

17. AI Replacement and Retraining

An important emerging question is:

Does an employee have a right to retraining rather than dismissal?

There is no general EU rule saying that every employee whose job can be automated must be retrained instead of dismissed.

However, retraining becomes legally relevant through:

collective redundancy consultation;

national employment law;

collective agreements;

works-council rights;

contractual obligations;

social-partner agreements;

sector-specific legislation.

The Collective Redundancies Directive expressly identifies retraining and redeployment as measures relevant to mitigating redundancy consequences. (EUR-Lex)

Thus, retraining may be an important factor in determining whether an employer complied with applicable obligations, but it does not automatically create an individual entitlement in every European country.

18. AI Replacement and Discrimination

This is potentially one of the strongest compensation routes.

An AI system may unintentionally reproduce historical discrimination.

Example

The AI learns from historical company data.

Historical employees:

70% male in technical positions;

30% female.

The algorithm learns that previous high performers were predominantly male.

It consequently gives lower retention scores to female employees.

The employer then uses those scores to select workers for redundancy.

The employer might argue:

"The AI did not use gender."

That may not end the inquiry.

Gender-correlated variables can produce indirect discriminatory effects.

The EU AI Act itself recognises that workplace AI may perpetuate historical discrimination affecting groups including women, older persons, persons with disabilities and persons of particular racial or ethnic origins. (EUR-Lex)

19. Age Discrimination in AI Replacement

Age is particularly relevant.

AI adoption may disproportionately affect older workers if:

the employer assumes younger employees adapt better;

training opportunities are allocated selectively;

AI productivity scores disadvantage workers unfamiliar with new software;

redundancy algorithms use age-correlated variables;

"future potential" becomes a selection criterion.

A redundancy apparently based on productivity may therefore conceal an age-discrimination issue.

The compensation claim would be based on the discriminatory selection, not merely on the fact that AI caused the job reduction.

20. Disability Discrimination

AI can also create disability-related discrimination.

For example, an AI productivity-monitoring system may penalise:

employees requiring additional breaks;

employees with reduced working speed;

employees working modified hours;

employees using accessibility technologies.

If those employees are disproportionately selected for redundancy, disability discrimination and reasonable-accommodation principles may become relevant.

21. GDPR Compensation

Where employee data are processed unlawfully, the GDPR provides another potential civil route.

Article 82 GDPR provides a right to compensation for damage resulting from infringement of the GDPR.

The CJEU has clarified that Article 82 compensation is compensatory rather than punitive, meaning that it is directed toward compensating damage actually suffered rather than punishing the employer. (EUR-Lex)

Possible AI-related GDPR problems

Examples include:

excessive employee surveillance;

unlawful profiling;

processing prohibited categories of personal data;

unlawful automated decision-making;

inadequate transparency;

unlawful retention;

using employee data for a new AI purpose without proper legal basis.

Thus an employee could potentially have:

employment claim + discrimination claim + GDPR claim

arising from the same AI restructuring.

22. AI Act Does Not Replace Employment Law

A common mistake is to assume:

"If the AI complies with the AI Act, the dismissal is lawful."

That is incorrect.

The AI Act operates alongside other legal regimes.

The AI Act itself preserves other obligations concerning information and consultation of workers and their representatives. (EUR-Lex)

Therefore:

AI Act compliance ≠ automatic dismissal-law compliance.

An employer may comply with AI-specific requirements and still violate:

national dismissal law;

collective redundancy law;

equality law;

GDPR;

employment contract;

collective agreement;

works-council rights.

23. Human Oversight

The AI Act's human-oversight requirements are especially important for employment AI.

High-risk AI systems must be designed to permit effective human oversight, and the persons responsible for oversight must have appropriate competence, training and authority. (EUR-Lex)

Therefore, an employer should ideally be able to demonstrate:

who reviewed the AI output;

what information the human decision-maker received;

whether the human could reject the AI recommendation;

whether alternatives were considered;

whether discriminatory effects were tested;

whether the employee could challenge the decision.

A nominal human signature on an automatically generated dismissal may not necessarily amount to meaningful human review.

24. Civil Liability for Faulty AI Employment Decisions

Traditional civil-law principles can also become relevant.

Suppose an employer uses an AI system that contains a serious defect.

The system:

incorrectly identifies employees as low performers;

produces systematically inaccurate scores;

contains biased training data;

fails to account for disability accommodations;

generates false misconduct findings.

If the employer knew or should have known of the defect and nevertheless relied on it, national civil/employment law may provide a basis for liability.

The legal chain becomes:

AI defect → erroneous employment decision → unlawful dismissal/other wrong → financial loss → damages

25. Contractual Claims

Employment contracts and collective agreements can create additional rights.

For example, an employment contract may contain:

consultation obligations;

bonus rights;

notice periods;

redeployment provisions;

training commitments;

redundancy procedures.

If an employer replaces the employee with AI but ignores those contractual obligations, the employee may bring a contractual claim.

The exact remedy depends heavily on the national law governing the employment relationship.

26. Wrongful Dismissal vs AI Redundancy

These should be distinguished.

Lawful technological redundancy

"The business genuinely no longer needs this position because AI has automated the function."

This can potentially constitute a legitimate redundancy situation, subject to national requirements.

Unlawful dismissal

"We used AI as an excuse to remove this particular employee without following required legal procedures."

This is different.

Discriminatory dismissal

"The AI systematically selected older workers."

Again, different.

Automated unlawful dismissal

"The AI automatically terminated workers without required human/legal review."

Again, different.

The compensation analysis therefore depends on why and how the employee was replaced.

27. Evidentiary Problems

AI replacement cases will often involve a major information asymmetry.

The employer may possess:

source-code documentation;

model cards;

training data information;

employee performance datasets;

audit logs;

model outputs;

risk assessments;

impact assessments;

human-review records;

communications concerning implementation;

redundancy-selection spreadsheets;

vendor contracts.

The employee may have only:

"Your position has been eliminated."

This makes disclosure and evidentiary rules extremely important.

The Meister case illustrates why lack of information can be significant in discrimination litigation, even though it did not establish a general right to obtain all employer information. (curia)

28. Causation

An employee must generally connect the AI system to the legal damage claimed.

For example:

AI introduced → employee selected → employment terminated → €40,000 lost earnings

But suppose the employer can establish:

"The position would have disappeared even without the algorithm."

Then causation becomes more complicated.

Similarly:

AI discrimination → dismissal → lost salary

creates a different causation pathway from:

lawful AI restructuring → lawful redundancy → ordinary statutory redundancy payment.

29. Damages That May Be Claimed

Depending upon national law and the legal basis, possible compensation can include:

Economic damages

unpaid salary;

notice pay;

redundancy compensation;

lost benefits;

pension loss;

lost bonuses;

job-search costs;

other proven financial loss.

Equality-related damages

discrimination compensation;

non-material damage;

injury to dignity;

career-related loss.

GDPR damages

material damage;

non-material damage caused by unlawful processing.

Article 82 GDPR is compensatory and seeks full compensation for damage actually suffered. (EUR-Lex)

Procedural remedies

Depending on national law:

reinstatement;

declaration of unlawful dismissal;

repetition of consultation;

annulment or reversal of a decision;

correction of records;

deletion/restriction of unlawfully processed data.

30. Employer Defences

An employer may argue:

1. Genuine business restructuring

The employee's role genuinely disappeared.

2. AI was not the decision-maker

A human made the final decision.

3. Objective selection criteria

Selection was based on legitimate business criteria.

4. No discriminatory effect

Statistical differences resulted from legitimate factors.

5. Proper consultation

Workers' representatives were consulted in accordance with applicable law.

6. Contractual compliance

Notice and redundancy payments were properly provided.

7. Lawful data processing

The employee data were processed under an appropriate legal basis.

8. Lack of causation

The alleged AI error did not cause the dismissal.

9. No compensable damage

The claimant cannot establish actual legally recoverable loss.

31. Important Distinction: AI Replacement vs AI Selection

This distinction is central.

32. Collective Redundancy + AI: Practical Example

Assume:

Company: European manufacturing company
Employees: 2,000
AI project: Automated production system
Potential job loss: 400 employees

The employer should examine:

Stage 1 — Business decision

Why is AI being introduced?

Stage 2 — Workforce impact

Which positions disappear?

Stage 3 — Consultation

Are collective redundancy rules triggered?

Stage 4 — Selection

Are objective and non-discriminatory criteria used?

Stage 5 — AI governance

Is the employment AI classified as high-risk?

Stage 6 — Human oversight

Who reviews AI recommendations?

Stage 7 — Data protection

What employee data does the AI process?

Stage 8 — Alternatives

Can workers be retrained or redeployed?

Stage 9 — Termination

Are national dismissal requirements satisfied?

Stage 10 — Compensation

Calculate the employee's statutory, contractual and tort/data-protection claims.

33. European Case-Law Principles Applied to AI Replacement

These cases are analogical rather than direct AI-replacement precedents. The European case law specifically concerning AI-driven job elimination is still developing.

34. Emerging Compensation Model

A useful way of understanding future litigation is:

Model 1 — Lawful AI redundancy

AI → genuine job elimination → lawful procedure → statutory/contractual redundancy rights

The employee receives whatever compensation national law provides.

Model 2 — Procedurally unlawful redundancy

AI → job elimination → consultation/dismissal rules breached → compensation

Model 3 — Discriminatory AI

AI → discriminatory selection → dismissal → equality claim → compensation

Model 4 — Automated decision violation

AI → prohibited/insufficiently controlled automated decision → unlawful processing/decision → GDPR/employment remedies

Model 5 — Multiple violations

AI → discriminatory monitoring → discriminatory selection → unlawful dismissal → several overlapping claims

35. Six Key Legal Tests for an AI Replacement Claim

An employee can examine six questions:

Test 1 — Was the job genuinely eliminated?

If yes, the case may principally concern redundancy law.

Test 2 — Was the employee individually selected by AI?

If yes, examine discrimination and automated decision-making.

Test 3 — Was collective consultation required?

If yes, examine Directive 98/59 and national implementation.

Test 4 — Did AI process personal information unlawfully?

If yes, examine GDPR.

Test 5 — Did the AI produce discriminatory results?

If yes, examine equality legislation.

Test 6 — Was meaningful human review performed?

If not, the legal risk associated with the AI decision becomes substantially more significant.

36. Important Limitation

There is presently no general European principle that "AI replaced my job, therefore the employer must compensate me."

Instead, compensation normally requires an independent legal basis.

The strongest potential bases are:

unlawful redundancy;

wrongful dismissal;

failure of collective consultation;

age/sex/disability/racial discrimination;

unlawful automated decision-making;

GDPR infringement;

breach of employment contract or collective agreement;

other national civil/employment-law wrongs.

The EU's newer AI and platform-work legislation nevertheless makes AI-specific transparency, oversight and accountability increasingly important. The Platform Work Directive expressly regulates automated decisions affecting platform workers and provides review and, in specified circumstances, compensation mechanisms. (EUR-Lex)

37. Exam-Ready Legal Principles

For an examination, remember:

AI replacement itself is not automatically unlawful.

Technological redundancy must comply with applicable national employment law.

Collective AI-driven layoffs may trigger Directive 98/59 consultation requirements.

The employer cannot necessarily avoid consultation by claiming that an algorithm made the decision.

AI selection can create indirect discrimination.

Opaque algorithmic systems create important evidentiary problems.

GDPR Article 22 can become relevant to significant automated decisions.

GDPR Article 82 can provide compensation for qualifying damage caused by GDPR infringements.

Human oversight is a central principle for high-risk employment AI under the AI Act.

AI Act compliance does not automatically make a dismissal lawful.

Employment status depends upon the substantive relationship, not merely technological labels.

Compensation generally requires a legally actionable wrong and provable damage.

Conclusion

Artificial-intelligence employment replacement claims in Europe sit at the intersection of employment law, civil liability, equality law, data protection and emerging AI regulation.

The most important legal distinction is between lawful technological redundancy and unlawful AI-assisted employment action. An employer may generally restructure its workforce when technology changes the economic need for particular jobs, but it must still respect applicable redundancy procedures, consultation rights, equality principles, privacy/data-protection rules, contractual obligations and AI-governance requirements.

The cases of Fujitsu Siemens, Danfoss, Enderby, Meister, SCHUFA, Bărbulescu, Uber and Yodel provide the principal European judicial building blocks. Together with the AI Act, GDPR and Collective Redundancies Directive, they provide a framework for analysing future claims in which workers argue that AI did not merely eliminate their jobs but caused an unlawful employment decision producing compensable loss. (curia)