Perspectives · Work and Technology

If Machines Do the Work, What Is Human Work Worth?

As machines take on more tasks once performed by people, the central question is not only what work remains, but how society distributes the value that technology creates.

If Machines Do the Work, What Is Human Work Worth?

For most of human history, the limits of work were also the limits of the body. A field could be ploughed only as quickly as an animal or a person could pull the tool. A message travelled no faster than the horse, ship, or wire carrying it. A calculation took the time required for someone to perform it. Machines changed those limits, first by amplifying muscle, then by extending reach, memory, calculation, and now some forms of language and pattern recognition.

Every major shift in technology has raised a version of the same question: if a machine can do what a person once did, what becomes of the person? The answer has never been determined by the machine alone. It depends on who owns it, who can use it, which work society values, how quickly institutions adapt, and whether the gains are shared or concentrated.

Artificial intelligence makes the question feel immediate because it reaches into tasks that were often treated as distinctively human: writing, translation, design, analysis, customer support, programming, and decision assistance. But the deeper issue is not whether a machine can produce an output that resembles human work. It is what we mean when we say that work has value—and whether the economic price of a task is a reliable measure of the human being who performs it.

A job is a bundle of tasks, not a single action

We often talk about jobs as if each were one indivisible activity: accountant, teacher, engineer, translator, designer, technician. In reality, every job is a bundle of tasks, judgements, relationships, and responsibilities. Some are repetitive. Some depend on context. Some require physical presence. Others involve trust, negotiation, care, or taking responsibility when the answer is uncertain.

A machine may automate one part of a job without replacing the whole role. Software can draft a report, but someone may still need to decide which question is worth asking, check the evidence, explain the result, and accept responsibility for what happens next. A diagnostic system can highlight a medical image, but care also involves a patient’s history, consent, communication, and the difficult judgement of what to do when the evidence is incomplete.

This distinction is central to the International Labour Organization’s 2025 analysis of generative AI and jobs. Its global assessment found that about one in four workers is in an occupation with some degree of exposure to generative AI, while concluding that transformation of jobs is more likely than wholesale redundancy because most occupations still require human input (ILO, Generative AI and Jobs: A 2025 Update). Exposure is not a forecast that a quarter of workers will lose their jobs. It is a measure of potential contact between the technology and tasks within occupations.

That difference matters. The right question is not simply, “Can AI do this job?” It is, “Which tasks can it do, at what quality and cost, under what conditions—and what happens to the rest of the job?”

When productivity rises, who receives the benefit?

Suppose a tool lets a worker complete in one hour what previously took four. That is a genuine productivity gain. But the gain does not decide what happens next.

The employer might ask the worker to produce four times as much for the same pay. The company might reduce prices and reach more customers. The worker might finish earlier and have more time for family, study, or rest. The company might need fewer workers. Or the benefit might be divided among several of these outcomes.

Technology creates possibilities; institutions determine much of the distribution.

This is why the economic value of human work cannot be discussed only in terms of efficiency. If a machine makes a task cheaper, the price of that task may fall even when the task remains useful. If a worker uses AI to produce more, the extra value may flow to the worker, the employer, the owners of the technology, or customers paying lower prices. The distribution depends on bargaining power, competition, labour rules, ownership, and the availability of alternatives.

The World Economic Forum’s Future of Jobs Report 2025 makes a related distinction between automation and augmentation. Employers surveyed expect the share of work tasks performed mainly by humans alone to decline by 2030, while machine-only work and human-machine collaboration grow. The report also raises the question of whether workers will share in the income created when more output depends on advanced machines and proprietary algorithms (World Economic Forum, Jobs Outlook). These are employer expectations, not certainties about the future. Still, they underline that productivity and prosperity are separate questions: a society can become more productive without distributing the rewards evenly.

The market price of work is not the worth of a person

We routinely confuse what the market pays for an activity with the value of the person performing it. They are not the same.

A task may be poorly paid because many people are forced to compete for it, because the work is difficult to measure, or because the people who need it have little purchasing power. Another task may be highly paid because it is scarce, protected by credentials, connected to ownership, or positioned near a profitable bottleneck. These prices tell us something about economic arrangements. They do not provide a complete moral ranking of human contributions.

Care work makes this visible. Feeding a child, supporting an elderly relative, listening to someone in distress, or helping a person recover from illness can be essential to a functioning society. Much of this work is unpaid or underpaid. Its market price does not capture all the value it creates because the benefits spread across families, communities, and future generations.

The same is true of maintenance. Preventing a system from failing may look like doing nothing. A network engineer who avoids an outage, a technician who catches a dangerous fault, or an administrator who quietly keeps backups reliable may produce no dramatic visible event. Yet the absence of failure can be the result of skilled work.

If automation makes some outputs abundant, society may need to reconsider why it rewards certain tasks and neglects others. The answer should not be to pretend every activity has the same economic value. It should be to recognise that market price is one measure among several, and a poor substitute for human dignity.

When intelligence becomes cheaper, judgement matters more

It is tempting to imagine that the future belongs to people who can do what machines cannot. That formulation is unstable: the boundary moves whenever technology improves. A skill that once seemed uniquely human may become partially automatable; a new tool may make a previously difficult skill accessible to more people.

A more durable distinction is between producing an answer and being accountable for using it. A system can generate a plausible explanation, but it may not know whether the question is appropriate, whether the source is trustworthy, who could be harmed, or when uncertainty should stop an action. Human judgement is not infallible, and machines can support it. But responsibility does not disappear merely because a recommendation came from software.

In technical work, for example, generating code is only one part of engineering. Someone must understand the environment in which it will run, define the failure conditions, assess security, observe behaviour in production, and decide what level of risk is acceptable. A generated configuration can be syntactically correct and still be wrong for the system it enters. A convincing answer can be based on a false assumption. The scarce skill is often not producing more output; it is knowing what should be built, how to test it, and what consequences to anticipate.

This does not mean that judgement will always command a high salary. Markets do not automatically reward the most socially important abilities. It means that as production becomes easier, the ability to frame problems, verify outputs, coordinate people, and take responsibility may become more consequential to the quality of the result.

The transition is not the same for everyone

Technological change rarely arrives evenly. A worker with reliable internet, time to learn, and access to useful tools may benefit from AI. Another worker, doing similar tasks under close monitoring and with little bargaining power, may experience it as a demand to work faster or as a reason to cut staffing.

The effects also differ across countries and industries. A tool that complements a highly trained professional in one organisation may be used to standardise or reduce entry-level work in another. A company may create new roles, but the people who lose their existing jobs may not live in the same place or have the skills required for those roles. “New jobs will appear” is not a complete answer to someone whose income has disappeared today.

Training is important, but reskilling cannot solve every problem by itself. There must be real opportunities to move into, enough time and income to learn, and a way for workers to participate in decisions that change their jobs. If a company captures all the productivity gain while asking employees to absorb the risk of transition, the result may be technically efficient and socially fragile.

A responsible transition requires attention to job quality, worker autonomy, education, social protection, and the distribution of gains. It also requires honest measurement. A pilot that saves time for a small group does not automatically prove that an entire occupation can be replaced safely or profitably. Nor does a claim that AI creates jobs erase the costs borne by people whose work is displaced.

What should a more productive society make possible?

Imagine a society in which machines handle a growing share of repetitive and dangerous work. That could mean fewer injuries, less drudgery, faster medical research, better access to knowledge, and more time for care, learning, craft, and civic life. It could also mean concentrated ownership, insecure employment, constant performance monitoring, and a smaller share of income going to people whose labour has become easier to substitute.

Both outcomes are possible in principle. Technology does not choose between them.

If productivity increases, one meaningful question is whether people receive any of the time that has been saved. A four-hour task completed in one hour can become four times as much work, or it can become a shorter working day. That decision is not encoded in the software. It is negotiated through business choices, public policy, collective organisation, and cultural expectations about what productivity is for.

Another question is ownership. If a small number of firms own the machines, models, data, and infrastructure that produce an increasing share of economic output, then access to the resulting income may become more concentrated. Competition, public investment, worker ownership, taxation, and other institutional choices can affect how widely benefits spread. No single measure is a universal answer, but treating distribution as an afterthought is itself a choice.

The goal should not be to preserve every task in its old form. Some work is exhausting, dangerous, or needlessly repetitive; removing it can be a genuine improvement. The goal should be to ensure that the people affected by change have a meaningful stake in what replaces it—and that material security does not depend entirely on whether an employer can still sell their time.

Work is more than employment, but income still matters

People often find purpose in work. It can provide structure, companionship, mastery, recognition, and a sense of contributing to something beyond oneself. But employment is not the only source of meaning, and it is dangerous to romanticise work that is exhausting or exploitative. A society should not require people to prove their worth through constant labour.

At the same time, telling people that human value is independent of employment offers little comfort if losing a job also means losing housing, healthcare, food security, or the ability to support a family. The moral argument that people matter beyond their productivity must be matched by practical arrangements that protect them when the labour market changes.

That is the tension at the centre of the automation debate. We may need fewer hours of some kinds of labour while still needing fair ways to distribute income, opportunity, responsibility, and social recognition. The question is not only how to help people find another job. It is how to organise a society in which technological capability can grow without making human security increasingly conditional.

The question behind the machine

When a machine does a task, it does not follow that the human who once performed it has become less valuable. It means that the relationship between capability, scarcity, and payment has changed. Whether that change improves lives depends on decisions that extend far beyond the machine.

We should ask who owns the productivity gain, who bears the transition costs, what work remains invisible, and whether people have the power to shape the systems affecting them. We should also ask what we want the saved time and new capability to make possible. If the only measure of progress is that more output can be produced with fewer paid hours, then human beings will be judged by a metric designed for production alone.

A better measure would ask whether people are safer, more secure, more capable, and freer to spend their time on work and relationships that matter. Machines can change what is possible. They cannot settle what a fair society owes its people.

If machines do the work, the value of human work will not be decided by technology alone. It will be decided by the choices we make about ownership, dignity, responsibility, and the kind of life increased productivity is supposed to support.

Share this article
Facebook WhatsApp X LinkedIn Medium
Quick navigationEsc