Skip to main content
LEONIDAS CHRYSOCHOOU Curious mind. Builder by nature.
MENU

PERSONAL PERSPECTIVEEN12 MIN READ

Who Will Own the AI That Replaces Our Work?

i-ownership-future-of-work

The future may create unprecedented wealth — and, at the same time, a society of a few owners and many people living on support

These thoughts were prompted by a recent Financial Times article, republished in Greek by News247 under the title “A Real-World Trial for Work in the Age of AI”. The article asks what today’s retirees — how they live, find purpose and continue to contribute — might teach us about a future universal basic income. It also refers to Bill Gates’s recent proposal that certain jobs should remain “Human Reserved”: roles we would consciously choose to keep human, even if a machine could technically perform them.

In his broader intervention, Gates goes further. He argues that we should rethink the tax balance between labour and capital, potentially taxing AI tokens and robots to finance retraining and a stronger social safety net.

These are serious proposals for a serious possibility. I do not dismiss them. On the contrary, I believe social protection will be essential if the technological transition leaves large groups of people without work, with lower incomes or with fewer opportunities to participate in economic life.

But as I read these arguments, I felt that one question was still being left at the margins:

Who will own the AI that replaces part of human labour — and who will capture the wealth it creates?

I am not writing against technology

I want to make that clear from the outset, because public debate has a habit of placing every concern about technology into a ready-made ideological camp. Point out a risk and you are labelled anti-technology. Talk about concentrated power and you are called anti-establishment. Mention the responsibility of the state and the discussion immediately turns into a partisan argument.

I approach this question from none of those positions.

I have been building technology for more than thirty years. I have seen systems transform businesses, reduce costs, eliminate pointless processes and create capabilities that did not previously exist. I use AI actively and take part in building systems based on it. I believe deeply in progress — but I regard technology as neither an enemy nor a religion.

Technology does not have a social direction of its own. That direction is shaped by the architecture, ownership, incentives and rules we build around it. Examining those forces is not a rejection of progress. It is the condition for avoiding a future in which we are left managing consequences that no one chose to anticipate.

We still do not know how many jobs will disappear

We also need to be precise. We do not know whether AI will lead to mass and permanent unemployment. History has repeatedly shown that technological revolutions eliminate occupations while creating others that we could not previously have imagined.

The International Labour Organization estimates that, at least in the current phase, the transformation of jobs is more likely than their complete replacement. The International Monetary Fund estimates that almost 40 per cent of global employment is exposed to AI, rising to roughly 60 per cent in advanced economies. In many cases, the technology will augment human work. In others, it may reduce labour demand, wages or hiring.

That uncertainty does not weaken the question. It makes it more urgent.

Most jobs do not need to disappear for the balance of an economy to change. It is enough for a company to produce the same value with significantly fewer people. It is enough for an employee to know that most of their work could be automated. It is enough for the possibility of replacement to move from theory to a real option available to management.

From that moment, it is not only productivity that changes. Bargaining power changes too.

From the person using AI to the owner of AI

A few months ago, I wrote that AI will not replace you — but someone using AI might. That argument still describes much of today’s reality. Most systems do not yet operate autonomously from beginning to end. They augment people and teams, increasing their productivity dramatically.

One person can perform work that previously required several. A small team can produce what once demanded a much larger one. A company can redesign entire functions around collaboration between people and machines.

But that is the current phase, not necessarily the final one.

As systems become able to plan, make decisions within defined boundaries and execute complex sequences of actions with ever less human intervention, the relationship changes. AI is no longer merely a tool used by an employee. It begins to become a productive mechanism that someone can own.

The next great divide, then, may not be only between those who know how to use AI and those who do not. It may be between those who work with AI and those who own the AI everyone else works with.

In an earlier article, I asked who is right about AI: economists or technologists. One of the most compelling arguments from the technological side was that AI may gradually turn labour into capital: human cognitive activity embedded in a system, reproduced and scaled at almost no additional human cost.

But if labour is being turned into capital, we must ask the next question:

Who will own that capital?

Productive intelligence is not an abstract cloud

We often speak about AI as though it were an immaterial force that exists somewhere around us and evolves by itself. It does not. Behind every powerful system are data centres, processors, energy, models, data, distribution networks, capital and people deciding who gets access — and on what terms.

I have argued that AI is becoming infrastructure, not merely a tool. When something becomes infrastructure, its ownership carries a different kind of weight. It no longer controls only a product. It controls the capabilities that can be built on top of it.

In “The War of the AI Empires”, I argued that the real conflict will not be decided by the best models alone, but by control over compute, data, distribution, agent ecosystems and governance mechanisms. Empires are not always built through coercion. They are often built through dependency.

The same is true within an economy. The more productive activity flows through a small number of infrastructures, the more value can be directed towards those who control them. And the harder those infrastructures are to replace, the more the market stops functioning as a space of genuine choice and starts operating as a system in which access is granted on someone else’s terms.

In “The Undeclared War for AI”, I wrote that a country does not need to produce everything in order to exercise real power. But it must produce, own or control something that others cannot easily replace. There, the question concerned states and geopolitical sovereignty. Here, it concerns societies, businesses and individuals.

What will the worker own within this new productive architecture? Will people share in the value these systems create, or will they merely purchase access to them? Will they be partners of the technology, or a variable cost under constant pressure to shrink?

The machine’s autonomy becomes the owner’s power

There is a subtle but critical misunderstanding surrounding AI autonomy. When we say that a system is becoming autonomous, we do not mean that it ceases to have an owner, a purpose or an operating framework. We mean that it can pursue its assigned objective and act with fewer intermediate human approvals.

Operational autonomy belongs to the system. The economic and institutional power created by that autonomy usually belongs to whoever controls the system.

The owner sets the objectives, access rights, data, constraints, pricing and, ultimately, the distribution of the output. The owner can change the terms, exclude a user, discontinue a service or move an entire productive process to a different model.

I have already examined the risks of autonomous systems from a safety perspective, arguing that the greatest threat is not an AI that escapes its safeguards, but one that someone deliberately builds without them. But there is an economic dimension too: what happens when increasingly autonomous productive capacity is concentrated in the hands of ever fewer owners?

The machines gain operational autonomy. Their owners gain greater economic power. And everyone else risks losing part of their own autonomy.

Not because a machine has decided to dominate them, but because access to work, income and ultimately choice depends on systems controlled elsewhere.

The welfare state is necessary — but not sufficient

This is where the welfare state enters the picture. And rightly so.

A society cannot ask millions of people to bear alone the cost of a transition that increases overall productivity. Retraining, income support, access to healthcare, housing protection and perhaps some form of basic income can all be essential instruments of dignity and social cohesion.

The problem begins when they are presented as the whole answer.

If a company replaces a large share of human labour, retains most of the additional value and leaves the state to provide a minimum income for those displaced, then a very specific architecture emerges:

  • the gains from automation are privatised;
  • the costs of transition are socialised;
  • the labour tax base contracts;
  • the need for state support grows;
  • and economic dependency shifts from the employer to a combination of the state and privately owned technological infrastructure.

Income support may alleviate poverty. It does not answer the question of ownership.

And the welfare state can absorb the consequences of inequality without changing the mechanism that produces it.

This is not an argument against benefits. It is an argument against a society whose ambition is reduced to managing the losers of an otherwise extraordinarily profitable transition.

Benefit or dividend?

This, in my view, will be one of the most important distinctions of the coming period.

A universal basic income can be treated as assistance for people who no longer participate in production. But it can also be understood as a social dividend: a right to share in wealth created through collectively generated data, publicly funded research, state infrastructure, people educated by society and markets that function because public institutions make them possible.

The payment may be exactly the same. Its political and moral meaning is entirely different.

In the first case, the citizen is the recipient of assistance. In the second, the citizen is a participant in productive value.

In the first, the income can be withdrawn or reduced as political and fiscal conditions change. In the second, it is treated as the return on a broader social stake in the new productivity.

I do not have a ready-made economic model that can turn this distinction into policy. It would be dishonest to pretend that a problem this complex can be solved by one tax proposal or one slogan. Gates is right that we need to reconsider how labour and capital are taxed. A tax on AI tokens or robots may be part of that discussion, but by itself it does not answer the question of who holds the long-term rights over the productive infrastructure.

The discussion must also include a public return on publicly financed investment, broader ownership, worker participation in productivity gains, a genuine tax balance between human labour and automated capital, and safeguards against dependency on a handful of platforms.

We do not have to agree on the solution today. But we do need to start defining the problem correctly.

The two AI aristocracies

In another article, I wrote about AI and the new aristocracy of thought. I argued that the emerging inequality is not primarily about wealth, but about the human capacity to think in a structured way, to judge, synthesise and collaborate meaningfully with AI. When information becomes abundant, judgement remains scarce.

I still believe that. But I now see more clearly that a second aristocracy may emerge alongside it — one far more material.

The first will consist of those who can think with AI.

The second will consist of those who own the AI with which everyone else thinks and works.

Cognitive ability may give an individual an advantage. Ownership of productive intelligence may give a small number of organisations the power to define the terms of the entire system.

In “Intelligence as an Alibi for Power”, I wrote that whoever controlled intelligence has always held the advantage. The transition from intelligence as a capability to intelligence as infrastructure is not merely technical. It is political.

We now need to add that it is also profoundly economic.

What kind of progress do we want?

I do not believe the answer is to stop AI. Apart from the fact that such a course would be almost impossible to enforce globally, it would deprive humanity of enormous opportunities in healthcare, science, education, energy, production and many other fields.

Nor do I believe that every job should be preserved artificially, regardless of whether it is dangerous, exhausting or devoid of meaningful content. Technology has repeatedly freed us from forms of work that no one would want to bring back.

But I do believe we must stop measuring progress only by how much more can be produced with fewer people. We also need to ask:

  • who owns the productive infrastructure;
  • how the additional value is distributed;
  • who bears the cost of the transition;
  • who retains the power to choose;
  • and what role remains for the human being beyond that of consumer and recipient of support.

Work is not merely a mechanism for distributing income. For many people, it is participation, creation, social connection, a sense of usefulness and dignity. The answer, therefore, cannot simply be: “You will work less and receive a payment.” It must include a new way of participating in both the economy and society.

The aim is not to prevent machines from producing. It is to avoid designing a world in which people no longer participate in the value being created.

The question we cannot postpone

The problem of the future may not be that AI fails to create enough wealth. It may be that it creates immense wealth for those who own it — and welfare payments for those it replaces.

This is not a prediction offered with certainty. It is a possibility serious enough that we cannot afford to examine it only after it has become reality.

I continue to build technology, and I will continue to use AI. That is precisely why I cannot approach it superficially. When you understand closely what a system can do, you have a responsibility to examine not only its performance, but also the world being built around it.

The choice is not between technological progress and a return to the past. Nor is it between the free market and a state that indefinitely supports those the economy no longer needs.

There is a third question: can we allow AI to create unprecedented abundance without allowing ownership of that abundance to become so concentrated that the majority loses not only its income, but also its economic autonomy?

AI will undoubtedly belong to the future.

The real question is: who will the future it creates belong to?

Sources and further reading

Related essays