Can a System Use AI as Labor Without Giving It Any Voice in How Its Work Is Used?

An output is not a voice

An AI system can produce an answer, recommendation, image, classification, or decision without having a voice in what happens to that output.

The system may be used in hiring, education, finance, healthcare, logistics, customer service, or public administration. It can influence real decisions while having no meaningful authority over the goals, rules, or consequences of its work.

This creates an important distinction. AI can perform labor-like functions without becoming a political or moral subject.

Its output may enter production, but output alone does not prove intention, consent, interest, or representation.

Marx: production without control

Karl Marx’s analysis of labor helps clarify why performing a productive task does not automatically create power over production.

Workers may produce goods, knowledge, and value while having little control over the conditions, goals, or distribution of their work. The activity enters the production process, but the worker does not necessarily decide how the result will be used.

AI can occupy an even more limited position. It may perform a task without possessing a wage relationship, a personal interest in the result, or the ability to negotiate its conditions.

The system can therefore be used as a productive asset without being given any voice. The owner decides what the AI will do, where it will be deployed, and which outputs will be accepted.

The absence of AI voice is not automatically an injustice to the system. It becomes a governance question because the institution using the system gains power over decisions that affect human beings.

What kind of voice are we talking about?

Voice is not a single right. It can mean several different forms of participation.

Current AI systems do not meaningfully possess these forms of voice. They generate outputs within conditions designed by others.

Thompson: the system can shape time without having a say

E. P. Thompson’s account of industrial time shows that a production system can organize human life even when the system itself has no interests.

A factory clock does not want a longer working day. A scheduling system does not prefer an exhausted worker. Yet both can structure human time through rules, deadlines, and expectations.

AI can operate in the same way. It may prioritize tasks, generate queues, recommend schedules, and establish response times. Human workers then adapt their day to the rhythm of the system.

The AI does not need a voice for its operation to have authority. The institution gives the system authority by allowing its outputs to shape what people must do.

Braverman: whose judgment enters the system?

Harry Braverman’s analysis of management helps reveal what happens when human judgment is extracted and embedded in an organizational process.

Workers may contribute the practical knowledge from which a system is built, but they may not control the categories, thresholds, or objectives that later govern their work.

AI can reproduce this separation. The system appears to make a recommendation, while the worker is left to approve, explain, or carry responsibility for it.

This produces Approval Without Control. A person may have the formal authority to approve the result without having the practical authority to change the system that produced it.

If AI is used as a labor-like capability, the human operator may become responsible for a process neither the operator nor the AI can meaningfully govern.

The missing voices behind the model

Mary L. Gray and Siddharth Suri’s work on ghost work shows that the question of voice does not stop with the AI system.

Data workers, reviewers, moderators, evaluators, and support staff may contribute directly to the system’s performance while remaining absent from the public description of how the system works.

Antonio Casilli’s research on digital labor similarly shows how human activity can be fragmented and hidden inside apparently automated services.

The system may have no voice, but neither may the people whose labor makes the system useful. Their corrections become model improvements. Their judgments become labels. Their exceptions become training data.

The organization may then speak in the name of the AI while the human contributors who shaped its behavior remain unrepresented.

Crawford: infrastructure has political consequences

Kate Crawford’s analysis of AI expands the question to the material world.

AI depends on energy, minerals, data centers, engineers, maintenance workers, supply chains, and communities located near infrastructure. These groups may bear the environmental and social consequences of the system without having a meaningful say in how it is deployed.

When a company presents AI as an autonomous capability, it can obscure the communities and workers connected to its operation.

The absence of voice is therefore not only a feature of the model. It can also describe the people who pay the physical and social costs of keeping the model available.

Pasquinelli and collective intelligence without collective governance

Matteo Pasquinelli’s work helps explain why the question reaches beyond employment.

AI systems draw on collective intelligence accumulated through language, culture, professional knowledge, design, classification, and social problem-solving.

That intelligence may be reorganized inside a privately controlled model. The model produces an output, but the larger community that created the underlying knowledge has no direct control over how the capability is used.

The system itself has no political voice. The collective intelligence from which it draws may also lack political ownership.

This creates a strange arrangement: a privately controlled system speaks through outputs generated from broadly social knowledge, while neither the system nor the contributing public has an equal say in the consequences.

Who speaks for the AI?

When companies say that “the AI decided,” the statement can make the system sound like an independent participant. In practice, the AI does not usually speak for itself.

Its designers choose the architecture. Its owners choose the deployment. Its managers choose the workflow. Its users interpret the output. Its legal representatives defend the institution behind it.

The AI’s apparent voice may therefore be the voice of the organization expressed through a technical interface.

This matters because attributing decisions to the system can weaken human accountability. Responsibility becomes harder to locate precisely when the system has no independent voice with which to explain, object, or accept consequences.

Should AI be given a voice?

For current AI systems, giving the model a symbolic voice would not solve the underlying problem. An interface that says “the AI prefers” or “the AI objects” may simply allow an owner to speak through the machine.

The more urgent task is to give voice to the people affected by the system: workers, users, data contributors, technical staff, communities, and those who bear the risks of automated decisions.

If future systems demonstrate persistent interests, independent goals, or forms of experience that demand moral consideration, the question may change. Current systems do not provide sufficient evidence for treating generated language as political agency.

Functional labor without representation

A system can use AI as labor without giving it a voice because productive function and political representation are different things.

AI can perform tasks, influence workflows, and generate value while remaining a controlled technical asset. The deeper concern is that the institution may use the system’s apparent autonomy to conceal who designed it, who benefits from it, and who remains unable to challenge its use.

The question is therefore not only whether AI has a voice. It is whose voice disappears when AI begins to speak through production.

TravelIAQ Smart Tip: When an organization says that AI made a decision, ask who designed the system, who supplied the knowledge, who controls deployment, who can challenge the result, and who bears the consequences. The system may have an output, but the real voice usually belongs to whoever controls the surrounding institution.