This is the second article in the series. The first article examined how AI can reduce labour demand by automating tasks, compressing organisational layers, redesigning production and lowering the amount of human work required per unit of output.
The harder question is what happens next: if AI can eventually reduce the amount of human labour the economy needs, how should income, ownership and human purpose change?
1. “Learn New Skills” Is Not Enough
The standard advice is predictable: people should learn new skills. There is truth in it. People should adapt. But it is incomplete.
Imagine an economy in which AI can perform the work of 100 people with the output of 10. What happens if those 90 people all successfully acquire the new skill? The economy may simply automate the new skill too.
This does not mean learning is pointless. It means that individual retraining cannot solve a structural reduction in the amount of human labour required by the economy.
We therefore need to move from a worker-centric question to an economic question. The old industrial arrangement was broadly one in which human labour generated income, and income enabled consumption. The emerging arrangement could increasingly involve AI and capital generating production with less human labour. The missing link is straightforward: how does production become purchasing power when fewer people are required to provide labour?
2. We May Have to Separate Income From Employment
For centuries, the normal economic bargain has been simple: people work, earn money and use that income to buy necessities. But if machines eventually produce a large proportion of what society needs, maintaining that arrangement becomes increasingly difficult.
The solution does not necessarily mean one particular policy. Several mechanisms could contribute.
Shorter working weeks
If productivity rises dramatically, society could choose to produce the same amount with fewer working hours. A five-day week could become four days, and four could become three. The objective would not necessarily be to create more work. It would be to distribute the remaining work more broadly.
Broader ownership of productive capital
If AI systems, robots and automated companies generate most economic value, ownership becomes crucial. That ownership could increasingly be distributed through pension funds, sovereign wealth funds, employee ownership, citizen investment funds, public investment vehicles or broader equity participation. The question becomes not merely who has a job? but who owns the machines doing the work?
Social dividends or basic income
Another possibility is to distribute part of the economic surplus directly through mechanisms such as universal basic income, negative income tax, social dividends or universal basic services. The details are political and economic questions. The underlying principle is straightforward: if machines produce more of society’s output, some mechanism must allow people without traditional employment to participate in that output.
More human activity that is not economically necessary
Humans may continue doing things not because machines cannot do them, but because people want humans to do them. That could include caregiving, teaching, mentoring, art, sport, community work, research, entrepreneurship, leadership and the cultivation of relationships. The point would not be to invent artificial “jobs” for people. It would be to allow human activity to become less dependent on what the labour market happens to reward.
3. Three Possible AI Economies
Technology alone does not determine the outcome. Institutions matter.
The AI Renaissance
AI produces enormous productivity gains. Ownership becomes sufficiently broad. Working hours decline. Basic living standards rise. People spend more time on relationships, creativity, learning, care, exploration and voluntary activity. This would be the optimistic scenario.
The AI Oligarchy
AI produces enormous productivity gains, but ownership remains concentrated. A small number of companies and investors own the productive systems. Output rises while the number of economically necessary workers falls. The technology succeeds, but the distribution of its benefits becomes highly unequal. This would be technologically successful but socially unstable.
The AI Stagnation
AI becomes powerful but institutions fail to adapt. Companies can automate, but regulation, education, labour markets and social systems remain designed around the old economy. Productivity rises unevenly. Jobs disappear in some sectors while new jobs fail to absorb everyone. This may be the most politically difficult scenario because neither technology nor society fully adjusts.
4. The Paradigm Shift
This is why I call the scenario a new paradigm, rather than simply another wave of automation.
The industrial economy was built around an assumption: human labour is necessary for production. The information economy modified that assumption: computers can amplify human knowledge and productivity. The AI economy could challenge it much more directly: machine intelligence can perform an increasing share of production itself.
That changes how we think about productivity because the marginal unit of output may increasingly come from computation, automation and capital rather than additional human hours. It changes the meaning of expertise because machines can increasingly supply the “how” while humans specify objectives, exercise judgment and accept responsibility. And it changes the relationship between ownership and income because the owners of productive machines may capture a larger share of the value created by those machines.
The deeper issue is therefore not simply technological unemployment. It is whether an economy designed around the necessity of human labour can remain stable when that necessity begins to decline.
5. What Happens to Human Value?
If AI becomes capable of performing many economically valuable tasks, it does not follow that humans become worthless. Economic value and human value are different things.
A mother does not become less valuable because an AI can write a lesson plan. A friend does not become less valuable because an AI can provide advice. A teacher does not become meaningless because an AI can explain mathematics. A musician does not become meaningless because AI can generate music.
The deeper change is that economic necessity may become less tightly connected to human worth. That could be liberating, but it could also be psychologically difficult. For centuries, people have answered the question “What do you do?” with their occupation. Jobs provide more than income; they provide identity, status, routine, social contact and a sense of contribution. A society in which fewer people need to work for survival will therefore need to solve a cultural problem as well as an economic one.
6. The Real Question Is Not Which Jobs Survive
We can now return to the original question: Which jobs will AI take?
Some will disappear. Some will shrink. Some will become more productive. Some will become easier for non-experts. Some will become premium professions built around judgment and responsibility. Some physical jobs will remain difficult to automate, while others will become easier because AI removes their cognitive complexity. And some apparently safe occupations will become overcrowded because AI has destroyed alternative employment elsewhere.
So there is no permanent list of “safe jobs.” The better framework is to ask four questions:
- How much human labour does this activity actually require?
- How much of that labour can AI or automation remove?
- Can the underlying product or system be redesigned so that the work is needed less often?
- If the work remains, will there be enough economic demand to support the number of people capable of doing it?
7. The New World
The most important change may not be that AI takes your job. It may be that AI changes the relationship between production and employment.
If machines can increasingly perform cognitive work, physical work and eventually parts of maintenance and self-management, the central economic problem may move from “How do we create enough jobs?” to “How do we distribute the abundance created by machines?”
That question cannot be answered by teaching everyone another software tool. It requires us to think about ownership, distribution and institutions as parts of the same economic problem. If productive capacity increasingly belongs to machines and the capital behind them, then society needs a mechanism through which people can share in the output even when their contribution of traditional labour declines.
That is not primarily a technology problem. It is a question of how the economic system is organised around a new relationship between capital, labour and production.
And eventually it becomes a philosophical problem: If machines can produce most of what society needs, what should humans do with their time? And if machines create most of the wealth, who should own that productive capacity?
Those questions require us to rethink the economic bargain itself.
The biggest change AI may bring to the world is not a better way of working. It may be a world in which working for a living is no longer the central organising principle of human life.
Whether that becomes liberation or catastrophe will depend much less on how intelligent the machines become than on what humans decide to do with the wealth those machines create.