The Fourth Sector: When AI Does the Work, Who Gets the Wealth?

An automated economy could produce more with less human labor. Whether that creates widespread freedom or deeper inequality depends on who shares in the gains.

Imagine a business that keeps producing while almost everyone is asleep.

Software handles incoming inquiries. AI prepares proposals. Automated systems process orders, manage inventory, and coordinate deliveries. A small team supervises exceptions and makes the consequential decisions.

Now imagine that pattern spreading across thousands of industries.

The question becomes larger than whether AI will replace your job. It becomes a question about how society distributes prosperity when producing more requires fewer additional hours of human work.

For most people, employment is the bridge between economic growth and personal income. Businesses need workers. Workers receive wages. Those wages pay for the things businesses produce.

What happens if that bridge becomes narrower while the economy on the other side keeps expanding?

That is the provocative idea behind the “fourth sector”: an economy in which automated systems perform a growing share of production, and ownership becomes increasingly important to who benefits.

It is a useful lens for thinking about the future. But understanding it requires separating a powerful argument from some overly confident predictions.

An economy can become more productive without making prosperity more widely shared.

And technology alone cannot decide how those gains are distributed.

The Three Sectors—and What Automation Changes

The familiar three-sector model divides economic activity into broad categories:

Consider a loaf of bread.

Growing the wheat belongs to agriculture. Milling the grain and baking the loaf belong to production. Transporting and selling the bread involve services.

Automation can change every stage. Machines can help harvest crops, industrial equipment can process ingredients, and software can coordinate orders.

But an automated farm remains a farm. A highly automated factory remains a manufacturer. A service delivered through software remains a service.

That distinction matters because the “fourth sector” proposed here is not an established replacement for official industry classifications. Systems such as the North American Industry Classification System organize establishments by their production activities and processes, using much more detail than three broad categories. census.gov

The automated economy is therefore better understood as a dimension running across existing sectors.

One question asks what an organization produces. Another asks how much human labor it requires.

Calling automation a fourth sector makes the shift easier to picture. Treating it as a separate, precisely measurable slice of GDP is more complicated.

Why AI Makes This Debate Different

Automation has a long history. Machines have replaced physical effort, software has absorbed administrative tasks, and industrial systems have increased what a single worker can produce.

AI expands the range of activities that can potentially be automated or accelerated.

The exposure now reaches into work involving language, analysis, design, programming, and decision support.

In its January 2024 analysis, the IMF estimated that almost 40% of global employment was exposed to AI, rising to about 60% in advanced economies. Those figures describe exposure, not predicted job losses: some workers could become more productive, while others could face reduced demand for their tasks. imf.org

That distinction is essential.

A job consists of multiple tasks. Automating three of them might make an employee more effective. Automating most of them might reduce staffing needs. Lower prices might also attract enough additional customers to create new work elsewhere.

The outcome depends on adoption costs, reliability, customer demand, competition, and how organizations redesign their operations.

Still, a consequential possibility emerges: businesses may be able to increase output without increasing payroll at the same rate.

A company that once needed ten additional employees to serve its next thousand customers might eventually need two—or none for that particular expansion.

Multiply that possibility across the economy, and the relationship between growth and employment becomes less predictable.

“No Human Input” Is a Much Higher Bar Than It Sounds

An AI system generating an answer in seconds can appear to operate independently.

Behind that answer sit chips, electricity, networks, data centers, software, maintenance, and people responsible for building and operating the system.

A robot completing a factory task does not make the entire supply chain autonomous.

There is a substantial difference between:

• A task completed without immediate human intervention.
• A business operating with a small supervisory team.
• An entire economy sustaining itself without human labor.

The first does not establish the third.

This is why forecasts claiming that machines will soon produce 90% or 99% of all economic value should be treated as scenarios, not established measurements or dependable timelines.

The more useful question is closer to the ground:

How much human work is required to deliver a reliable unit of value—and how is that changing?

That can be investigated in a customer service operation, a manufacturing process, or an accounting workflow without assuming the arrival of a fully autonomous civilization.

The Biggest Shift Could Be From Wages Toward Ownership

Imagine two people participating in the same automated business.

One earns a salary for doing work that the company is learning to automate. The other owns part of the company.

If automation raises profitability while reducing demand for that employee’s tasks, their financial outcomes could move in opposite directions.

The worker faces pressure on wages or employment. The owner may benefit from higher earnings.

This does not mean every owner wins or every worker loses. Competition can eliminate profits. Investments can fail. Employees can gain from better tools, higher productivity, and new opportunities.

But it illustrates why ownership deserves more attention in the AI debate.

IMF researchers have explored how AI could affect inequality through both labor income and capital returns. Their analysis highlights the possibility that gains accruing to asset owners could increase wealth inequality, even when changes in wages produce a different pattern. imf.org

A person may become extremely proficient at using AI and still capture only a limited share of the value they help create.

If their productivity doubles, several things could happen:

Their wages could rise. Their employer’s profits could rise. Customers could receive lower prices. Their workload could increase. Their team could shrink.

The technology creates possibilities. Bargaining power, competition, ownership, and policy influence the distribution.

A Bigger Economy Does Not Guarantee a Better Household Budget

Consider a hypothetical economy producing 100 units of output.

Decades later, it produces 1,000 units, with automated processes responsible for most production.

That sounds like extraordinary progress. It could be.

But the production total does not tell us whether the typical household can afford housing, healthcare, food, or leisure.

We would still need to know:

Who owns the productive assets? How much do prices fall? What happens to wages? How are profits distributed? What services are publicly available?

It would also be incorrect to assume that 90% automated production necessarily means owners receive 90% of national income. Output attribution and income distribution are different questions.

Automation may benefit consumers through lower prices, workers through complementary jobs, owners through returns, and the public through taxes and services.

The proportions depend on how the economy operates.

GDP measures production. It does not, on its own, describe the distribution of purchasing power, financial security, or opportunity.

A society can post impressive growth figures while large numbers of people feel increasingly precarious.

What We Should Measure Alongside GDP

If automation becomes more important, our dashboard should become more informative.

Labor’s share of income is one useful measure. The U.S. Bureau of Labor Statistics defines it, for a given sector, as labor compensation divided by current-dollar output. It is not a direct measure of the percentage of work performed by machines. bls.gov

Other questions deserve equal attention:

Are typical households gaining purchasing power?
Productivity gains matter more to daily life when people can afford better goods, services, and living conditions.

How broadly is productive wealth owned?
An economy where gains reach many households through business ownership, pensions, or other institutions differs from one where ownership is highly concentrated.

Are people working fewer hours by choice?
A shorter working week with stable living standards is a different outcome from involuntary underemployment.

Are essential costs falling?
Cheaper digital content provides limited relief if housing, healthcare, and energy remain unaffordable.

How difficult is the transition between jobs?
New opportunities can coexist with serious hardship for people who cannot move, retrain quickly, or absorb an income interruption.

The quality of an automated economy will depend on these lived outcomes.

Human Value Will Not Disappear—But It May Be Priced Differently

Some activities matter partly because another person performs them.

A tennis lesson offers a useful example.

Software might analyze a swing, identify positioning errors, or recommend drills. Yet a coach can also notice frustration, build confidence, adapt communication, and help a player stay committed.

The technical information is one component of the service. The relationship is another.

Similar considerations apply to care, teaching, live performance, hospitality, leadership, and many other activities.

However, “people will always want people” is not a complete employment strategy.

Wanting a human service does not guarantee that enough customers can afford it, or that the people providing it will earn good wages.

A plausible future includes both inexpensive automated services and more expensive services built around human attention. Whether that becomes a healthy range of choices or a widening divide will depend partly on purchasing power.

Human connection can remain valuable while becoming unequally accessible.

The Opportunity for Small Businesses

The same forces that threaten established jobs could also lower some barriers to entrepreneurship.

Consider a small agency serving local businesses.

Its customers need inquiries answered, appointments coordinated, information kept current, and results explained. Some of that work can be standardized and automated. Some requires judgment, trust, and knowledge of the client.

The business opportunity is to combine those capabilities into a dependable service.

Simply generating more content is a weak proposition if competitors can do the same. Solving a recurring customer problem—and being accountable when something goes wrong—is more defensible.

For a small operator, useful assets might include a trusted brand, a direct customer relationship, an original resource library, a specialized workflow, or a product customers repeatedly use.

These assets still require maintenance. They can fail. They are not effortless income.

But they can allow revenue to depend less exclusively on delivering another hour of personal labor.

A sensible starting point is one narrow workflow:

Identify a repeated problem. Establish a quality standard. Automate the reliable steps. Keep human review where mistakes are costly. Measure whether customers receive a better outcome.

The business should earn its place through results, rather than through the novelty of using AI.

Ownership Must Mean More Than a Marketing Promise

As interest in AI ownership grows, so will the temptation to attach ownership language to products that provide little actual control or economic participation.

An account is not equity. A reward point is not a dividend. Access to an AI tool does not mean ownership of the infrastructure behind it.

A community token does not automatically provide enforceable rights to revenue, assets, or governance.

The practical questions are straightforward:

What does the participant actually own? What rights come with it? Who controls the system? What costs and liabilities remain? Can the claimed benefits survive outside promotional material?

Those questions matter because an economy increasingly organized around ownership could also become an economy increasingly saturated with misleading promises about ownership.

Access to technology and a durable claim on its economic returns are different things.

The Choices That Could Spread the Gains

There is no single policy that guarantees an inclusive automated economy.

Possible approaches include broader employee ownership, profit-sharing arrangements, support for new businesses, competitive markets, stronger transition assistance, and public investment that expands access to useful services.

Each involves tradeoffs.

Employee ownership can share upside, but tying both someone’s job and savings to one employer concentrates risk. Public investment funds require accountable governance. Retraining helps only when it connects people to work that actually exists.

Fiscal policy also matters. IMF analysis has argued for stronger social protection and attention to capital-income taxation as ways to address disruption and the concentration of AI gains. These are policy proposals, not automatic consequences of technological progress. imf.org

The underlying objective is straightforward: people need credible ways to participate in prosperity even if wages become a less reliable route for some households.

That requires institutions as well as inventions.

The Future Will Be Built Around Decisions We Make Before It Arrives

The “fourth sector” is most useful as a challenge to an assumption: that economic growth will always require enough additional human labor to distribute its benefits primarily through employment.

We do not need to accept a prediction of universal joblessness to take that challenge seriously.

We need to recognize that productivity, employment, wages, prices, and ownership can move in different directions.

For individuals, that means learning how their work is changing while developing judgment, relationships, and capabilities that remain useful.

For businesses, it means building systems that deliver reliable value and deciding how productivity gains will be shared.

For society, it means asking whether technological progress is translating into greater security, opportunity, and freedom for ordinary people.

An automated economy could reduce drudgery, expand access to services, and give people more time to live.

It could also concentrate economic power while leaving millions competing for a smaller share of the gains.

The machines may do more of the work. Who benefits will remain a human decision.