From Hiring to Firing, AI Is Rewriting Workplace Power and Worker Rights

Artificial intelligence is moving into the workplace faster than labour institutions are adapting to it. The disruption is no longer confined to machines replacing particular tasks. Algorithms are increasingly influencing who gets hired, what work people receive, how closely they are monitored, how their performance is scored and, in some cases, whether they continue to have access to work at all.

A study titled "From Algorithm to Law: How Artificial Intelligence Redefines Work and Labour Rights in Peru," published in the journal Laws by María Gracia Valdivia Corzo of Universidad Católica de Santa María, examines this changing relationship between technology, managerial authority and worker protection. Although Peru provides the legal and institutional case study, the questions raised by the research reach far beyond one country: what happens when managerial decisions become automated, opaque and difficult for workers to challenge?

The study argues that the most urgent regulatory challenge is not automation by itself. It is the growing delegation of managerial power to algorithmic systems without equally strong mechanisms for transparency, human review, accountability and worker participation. For governments trying to capture AI-driven productivity gains while protecting decent work, that distinction could shape the next generation of labour regulation.

The AI debate is shifting from disappearing jobs to invisible power

Public discussion about AI and employment has largely revolved around job displacement. The study points to a more immediate transformation already under way inside workplaces. AI systems can participate in recruitment, task allocation, scheduling, performance evaluation, workplace supervision and decisions affecting continued employment, allowing activities once performed visibly by managers to migrate into data-driven systems.

Algorithmic management changes the nature of workplace control because supervision can become continuous, highly scalable and embedded in software. Systems can collect large volumes of data, establish performance metrics, allocate work and influence organizational decisions at speeds and scales that conventional human supervision could rarely match. Efficiency may increase, but so can the employer's capacity to monitor and direct workers.

It has major policy consequences. A worker does not need to lose a job to experience the disruptive effects of AI. Employment can remain intact while autonomy declines, monitoring intensifies or opportunities become dependent on scores and classifications that are difficult to understand. As a result, the regulatory problem moves beyond protecting people from technological redundancy to protecting them within increasingly algorithmic workplaces.

Risks also extend across several established labour rights. Automated systems can reproduce discrimination from historical data, continuous monitoring can affect privacy and dignity, and performance algorithms can influence promotion, discipline, remuneration or dismissal. The study also links intensive digital monitoring and productivity scoring with concerns over stress, anxiety, job insecurity and reduced worker autonomy.

For employers, this creates a governance problem as much as a legal one. Purchasing an AI system from an outside technology provider does not remove responsibility for decisions affecting workers. As AI becomes embedded in human resources and operational management, businesses will increasingly need to understand what their systems measure, how recommendations are produced and whether those outputs can be defended when challenged.

Rights can exist on paper and still fail inside an algorithmic workplace

The study differentiates between substantive rights and procedural capacity. A country may already protect equality, privacy, dignity and job security, yet workers can struggle to exercise those rights if they cannot discover that an algorithm influenced a decision or obtain the information needed to contest it.

Opacity becomes more than a technical problem. Consider a worker denied promotion after an automated performance assessment. Existing anti-discrimination and employment protections may technically apply, but enforcing them becomes difficult if the worker cannot determine what data were used, which factors shaped the result, whether inaccurate information entered the system or how much weight the employer placed on the algorithmic recommendation.

Transparency alone may not solve the problem. Dumping technical documentation on workers does little if the information cannot be understood or connected to the actual decision. The study draws an important distinction between technical disclosure and practical intelligibility: affected workers need enough understandable information to comprehend why a consequential decision occurred and to challenge it effectively.

Human oversight faces a similar test. A nominal supervisor who simply approves an algorithm's recommendation does not provide meaningful protection. Effective review requires a person with sufficient authority, competence and access to relevant information to question the system's output and intervene when necessary.

This principle is globally relevant. As governments develop broad AI frameworks, labour regulation may need a more specific procedural layer covering disclosure, evidentiary access, review, appeal and remedies. Without those mechanisms, familiar workplace rights risk becoming progressively harder to enforce as decision-making becomes more automated.

Peru shows why the Global South cannot simply import Europe's AI rulebook

Peru provides an instructive example because regulation is not starting from zero. The country has constitutional labour protections, personal-data rules, labour inspection powers and a national AI framework. Its AI regulations classify specified employment applications, including systems affecting worker selection, evaluation, hiring, termination and working conditions, as high-risk and introduce transparency obligations.

The problem identified by the study is fragmentation. Different legal regimes protect different parts of the employment relationship, yet Peru lacks an integrated labour-specific procedure covering algorithmic management, access to evidence, contestation, collective consultation and coordinated enforcement. The regulatory challenge is therefore less about inventing rights from scratch and more about making existing protections operational before, during and after AI-mediated employment decisions.

Peru also demonstrates why regulatory models developed in wealthier economies cannot simply be transplanted elsewhere. Official statistics cited in the paper show that 70.9% of Peru's employed population was in informal employment in 2024. High informality, limited collective representation, unequal digital skills and constrained enforcement capacity can leave significant groups beyond regulations designed primarily for standard formal employment.

Similar challenges confront many developing economies. Platform workers, subcontracted workers and economically dependent self-employed people may be heavily governed by algorithms without fitting neatly into traditional employment categories. Digital labour regulation has to consider the reality of labour markets in which formal contracts are not the only route through which economic dependence and managerial control operate.

European rules nevertheless offer useful regulatory building blocks. The study examines the GDPR, EU AI Act and Platform Work Directive alongside the ILO's 2026 Convention No. 193 on decent work in the platform economy. Its recommendation is selective adaptation: borrow mechanisms that solve identifiable problems, such as human oversight or disclosure, while adjusting them to domestic institutional capacity rather than reproducing an entire foreign model.

The next frontier is not more AI principles, but enforceable workplace safeguards

The paper proposes a practical regulatory agenda centred on consequential uses of AI rather than treating every digital tool alike. Systems affecting hiring, remuneration, performance evaluation, discipline, dismissal, biometric monitoring or continuous surveillance would justify stronger safeguards than low-impact administrative technologies. Such a risk-based approach could help governments concentrate limited regulatory capacity where worker consequences are greatest.

Several safeguards emerge repeatedly from the analysis. Significant employment decisions should receive meaningful human review. Workers should know when algorithms materially influence decisions about them and should be able to obtain understandable explanations. Higher-risk workplace systems should undergo impact assessments and independent audits covering discrimination, privacy, occupational health and data quality. Worker representatives should also be consulted when AI substantially changes monitoring, evaluation or working conditions.

Collective governance deserves particular attention. AI systems often affect entire groups of workers at once by changing productivity targets, scheduling rules or surveillance practices. Treating every dispute as an individual privacy complaint overlooks this structural dimension. Collective bargaining and worker participation can give employees a role before technologies become embedded in workplaces, rather than leaving them to challenge consequences after deployment.

The study also has limitations. It is a doctrinal and comparative legal analysis rather than an empirical investigation of workplaces. It cannot establish how widely Peruvian firms currently use algorithmic management, how workers experience those systems or how effective different enforcement mechanisms will prove in practice. Evidence on AI-specific employment disputes in Peru also remains limited.

Those gaps point toward the next research frontier. Governments and development institutions need better evidence on where workplace AI is being deployed, which groups face the greatest risks, how workers understand automated management and whether proposed safeguards actually change outcomes. Comparative research across Latin America, Africa and Asia could also reveal which regulatory mechanisms survive the transition from highly formal labour markets to economies where informality and weak enforcement are persistent realities.