As AI Decimates Jobs, Will a Universal Basic Income (UBI) Need To Be Introduced?

For most of modern economic history, there has been an implicit bargain at the heart of advanced economies: productivity rises, and living standards broadly follow. The mechanism has been wages. Firms innovate, output per worker increases, and over time that increase feeds through into employment and income growth. Nevertheless, artificial intelligence (AI) now threatens to disrupt that transmission mechanism.

The issue is not whether AI raises productivity. It certainly does. The issue is whether productivity gains continue to translate into broad-based wage growth, or whether they increasingly accrue to capital owners while labour demand weakens. If productivity becomes decoupled from wages at scale, the policy question shifts fundamentally. In that scenario, proposals such as a Universal Basic Income (UBI) move from the margins of political debate to the centre. They do not do so as ideological experiments, but as potential mechanisms for preserving purchasing power and economic stability in a labour market that may require fewer workers.

Productivity Without Jobs

The UK enters this transition with unemployment at 5.2%, vacancies at their weakest level since the pandemic period, and a sharp contraction in graduate and early-career roles. At the same time, survey evidence suggests firms adopting AI are reporting double-digit productivity gains alongside net reductions in headcount. That combination matters.

In classical labour economics, firms hire workers until the real wage equals the marginal product of labour. If a technology increases workers’ marginal productivity, labour demand expands. If it substitutes for workers, performing the same tasks more cheaply, the marginal product of those workers falls, and labour demand contracts.

The critical variable is the elasticity of substitution between capital and labour. The more easily capital replaces labour, the more employment shrinks when capital improves. That is what artificial intelligence is doing. Artificial intelligence is not simply augmenting labour; in many contexts, it is replacing it. When generative systems draft contracts, design marketing assets, analyse data or write code, they reduce the need for junior and mid-tier professionals. That is, the business can maintain output with fewer people.

That is no longer just theory, it is increasingly observable behaviour: hiring freezes, non-replacement of leavers, flattened hierarchies and workflow redesign around AI “agents” capable of executing structured professional tasks.

The effect begins at the bottom of the career ladder. Entry-level roles, the traditional training ground of the middle class, are the first to thin. Over time, that narrows progression pathways. This is not mass unemployment. It is labour-market compression.

The Domestic Middle-Income Trap

The term “middle-income trap” usually refers to developing economies that achieve productivity growth but fail to deliver widespread wage convergence.

However, a domestic analogue to the middle-income trap is emerging. Imagine an economy in which:

• Productivity rises steadily.
• Corporate profits increase.
• AI systems scale globally at negligible marginal cost.
• Yet stable mid-tier salaried employment shrinks.

The middle class is not merely a cultural category; it is the engine of consumption. Middle-income households spend a far greater proportion of their income than the wealthy. They sustain retail, housing markets, services and domestic growth. If a significant share of that cohort loses secure employment or experiences sustained income compression, aggregate demand weakens.

This is the uncomfortable arithmetic at the centre of the artificial intelligence and middle-income trap debate: If large numbers of middle-class workers lose secure income streams, who buys the goods and services businesses produce? An economy can produce more with fewer workers. However, it cannot grow if purchasing power narrows too far.

From Labour Share To Capital Share

National income divides into labour income (wages) and capital income (profits, rents and dividends). Over recent decades, labour’s share has already trended downward in many advanced economies. Artificial intelligence is now accelerating that shift.

Unlike traditional machinery, however, AI systems scale at near-zero marginal cost. Once developed, they can replicate tasks globally without proportional increases in labour. Returns concentrate around intellectual property, data control, platform ownership and network effects.

As output rises while labour input falls, capital’s share of income rises mechanically. GDP can expand while median wages stagnate. Accordingly, the dividing line in society shifts, with income security now dependent less on employment status and more on ownership status. Consequently, upward mobility will now increasingly require access to equity, scalable intellectual property, or asset participation. A job, by itself, will become increasingly a less reliable route as a pathway to advancement. That is not technological dystopia. It is distributional mathematics.

Government Policy As An Accelerant

As we highlighted in a previous article, Government policy by increasing employment costs has acted as an accelerant to this process (i.e. via increased employer National Insurance contributions and the Employment Rights Act 2025).

Whilst the cost of employing a human being is rising, the cost of automating their tasks is falling. From a business's perspective, the comparison is stark: a human employee represents recurring salary, pension obligations, compliance burdens and litigation risk. An AI system represents upfront investment, scalable output and no holiday entitlement.

Firms do not need to announce dramatic redundancy programmes to respond. They can simply stop hiring. They can allow roles to disappear through attrition. They can redesign teams around smaller cores supervising automated workflows.

This quiet restructuring is already visible in weakened vacancy data and declining entry-level recruitment.

Robotics And The Compression Of Manual Work

The common reassurance is that displaced professionals can retrain into trades or manual work. That comfort may however prove temporary.

Artificial intelligence is increasingly embedded into robotics, improving machine perception, spatial reasoning and dexterity. Warehousing, logistics and manufacturing are already automation-intensive, as service robotics continues to advance.

As elasticity of substitution between capital and routine physical labour rises alongside cognitive substitution, the compression is spreading across the labour market.

The risk is not universal joblessness. It is polarisation: a narrow elite of high-skill, high-judgement professionals; a concentrated group of capital owners; and a broader segment of precarious or lower-paid service work.

Some economists describe the long-term danger as the formation of a permanent underclass, not necessarily impoverished, but structurally excluded from meaningful social mobility.

The Demand Constraint

Modern advanced economies are consumption-driven. If income shifts upward toward capital owners, aggregate savings increase relative to spending. Without redistribution or expanded public expenditure, demand can weaken even in the presence of strong productive capacity.

This is how high-productivity economies drift toward secular stagnation: plenty of output potential, insufficient purchasing power.

UBI (a policy under which every adult citizen (and sometimes every resident) receives a regular cash payment from the state, unconditionally and regardless of employment status or income level) enters the debate here not as charity, but as macroeconomic stabilisation. It represents one mechanism by which productivity gains could be recycled into broad-based demand.

The argument is not ideological. It is mechanical. If labour’s share falls and capital’s share rises, either ownership must broaden dramatically, or redistribution must increase, or growth falters.

Even if income can be stabilised through policy, a deeper question emerges. For two centuries, technological progress reduced working hours gradually while raising living standards. If artificial intelligence meaningfully reduces the need for human labour across both cognitive and routine physical domains, societies may eventually confront whether full-time employment remains the central organising principle of adult life.

Experiments with shorter working weeks suggest that productivity does not always scale linearly with hours worked. If automation continues raising output per hour, work-sharing or reduced-hour norms may become rational adaptations rather than radical proposals. The alternative is to cling to a work-centric identity in a system that requires progressively fewer workers.

Redistribution, Ownership, Or Redesign?

If the wage mechanism weakens, societies face three strategic choices:-

• They can redistribute income directly through mechanisms such as Universal Basic Income (UBI) or negative income tax.
• They can broaden capital ownership through employee equity, sovereign wealth participation or universal capital dividends.
• They can redesign labour allocation by reducing working time and spreading available work more thinly.

In practice, a durable settlement may require elements of all three.

The Unavoidable Question

The UK is not facing imminent economic collapse. But it is entering a structural transition in which:

• AI substitution is observable.
• Entry-level labour demand is weakening.
• Labour costs are rising relative to automation.
• Capital scalability is increasing.
• The consumption base will narrow as mid-tier employment contracts.

Universal Basic Income is not inevitable. But the logic driving its re-emergence is rooted in classical economic distribution theory. If productivity growth becomes decoupled from broad wage growth, and if ownership remains concentrated, then the link between output and shared prosperity fractures. At that point, the debate is no longer about jobs. It is about the architecture of capitalism itself.

Yet the implications extend beyond income distribution. If artificial intelligence and robotics reduce the total quantity of human labour required to sustain high levels of output, then the role of work in society itself may change. For two centuries, productivity growth has gradually reduced working hours while increasing living standards. A further step in that direction would not be unprecedented. What would be unprecedented is the speed and scale at which such a shift might occur.

In such a world, economic participation may no longer revolve exclusively around full-time employment. Leisure, caregiving, creative production, independent research and community engagement may occupy a larger share of adult life. Education, accordingly, would have to evolve. It would no longer prepare individuals solely for employment within a wage-based system, but for functioning within a society where contribution, ownership, and adaptability matter as much as occupational status.

Whether such a transition proves stable or destabilising depends on how its gains are distributed. The central question remains stark: if work no longer distributes prosperity widely enough, what will? That is not a technological question. It is a political and economic one. And it will define the next decade.