The
IMF’s 2024 assessment of AI’s labour market impact concluded that
approximately 40 percent of jobs globally are exposed to AI automation, and
that the technology is likely to increase inequality within countries even as
it potentially raises aggregate productivity. These are not projections for a
distant future. The displacement is happening now, unevenly, in ways that cut
along existing fault lines of class, race, geography, and education, widening
gaps that societies were already struggling to close.

The automation narrative has a long history of false alarms.
Previous waves of technological change, including mechanisation,
computerisation, and the internet economy, disrupted specific job categories
while creating new ones, typically resulting in net employment growth over
decades-long timeframes. The question with AI is whether the current wave is
categorically different: faster, broader, and less complementary to human
labour in ways that make historical analogies misleading. Economists are
genuinely divided on this question, and the honest answer is that nobody
knows with certainty.

Who Is Most Exposed

The pattern of AI-related labour market disruption differs from
previous automation waves in one significant respect: it disproportionately
affects white-collar and cognitive work rather than being concentrated in
manufacturing and physical labour. Roles involving document processing, data
analysis, customer service, routine legal work, basic medical transcription,
and administrative functions are all facing significant AI substitution
pressure that would have seemed implausible five years ago.

Research from McKinsey Global Institute estimates that activities
accounting for up to 30 percent of work hours in the US economy could be
automated with current AI by 2030. The sectors with the highest exposure
include financial services, legal services, information technology, and
professional services, historically well-compensated sectors whose workers
have not been the traditional focus of concerns about automation. This
reversal of the conventional wisdom about who automation affects is
significant and is only beginning to register in policy
discussions.

At the same time, the workers at the bottom of the income
distribution face different pressures. Retail, hospitality, and logistics
jobs are being automated through AI-enabled robotics, self-service systems,
and routing optimisation. These roles typically offer fewer transition
options than disrupted professional jobs and serve workers with less access
to retraining resources. The convergence of automation pressure at both the
top and the bottom of the skills distribution, while the middle holds
relatively firm, is a pattern that defies simple characterisation as either
good or bad news.

Geographic Concentration

AI economic gains are heavily geographically concentrated. The
companies developing and commercialising AI are clustered in a handful of
metropolitan areas, primarily in the United States, China, and Western
Europe. The productivity gains from AI adoption flow largely to shareholders
and highly skilled workers in these locations. The communities where
displaced workers live, which are often not the same communities where AI
companies are headquartered, face the consequences of automation without
commensurate access to its benefits.

This geographic dimension compounds existing regional
inequalities. Areas already struggling with deindustrialisation now face
additional pressure from AI-driven service sector automation, with limited
local alternatives for workers whose skills are displaced. The political
consequences of this concentrated disruption are visible in the populist
movements drawing support from economically marginalised communities in
developed economies. Research by the Brookings Institution
has mapped the geographic distribution of AI-exposed jobs and found that
rural areas and smaller cities face disproportionate displacement risk
relative to their capacity to absorb or adapt to it.

The Retraining Question

Policy responses to AI labour displacement have focused heavily on
retraining and upskilling. The idea that workers can transition to new roles
if given access to education and support is intuitive and politically
palatable, but the evidence that large-scale retraining programmes
successfully transition workers from disrupted industries is, at best, mixed.
Trade adjustment assistance programmes in the United States, which have
operated for decades, have consistently underperformed in terms of
participant employment and earnings outcomes.

The pace of AI-driven change may make the retraining approach even
less viable than historically. If the half-life of specific technical skills
is shortening rapidly, continuous retraining becomes an unrealistic
expectation to place on workers who also have families, financial pressures,
and limited time. Models distributing AI productivity gains more broadly,
through progressive taxation of AI-generated profits, sovereign AI funds, or
universal basic income, are attracting renewed attention as complements or
alternatives to individual retraining. The OECD has published
comparative analysis of these policy options that suggests no single approach
is sufficient and that combinations of targeted support, social insurance reform,
and revenue-sharing mechanisms are needed.

What This Means for You

If you work in a role involving significant document processing,
data analysis, customer communication, or routine professional judgement, it
is worth honestly assessing how your role might change over the next five to
ten years as AI capabilities continue to develop. This is not cause for
panic; the pace of actual deployment typically lags behind the pace of technical
possibility. But it is reason for strategic thinking about skills,
adaptability, and professional positioning in a labour market that is
changing faster than the careers of most people currently in it were designed
for.

The social insurance implications of AI-driven labour market
change are receiving increasing attention from economists and policymakers
who recognise that existing welfare systems were designed for a different
labour market structure. Unemployment insurance systems calibrated to
temporary job loss between similar roles are poorly suited to permanent
displacement into different occupational categories. Healthcare systems tied
to employment relationships become less viable as employment relationships
become less stable. Retirement systems that assume decades of continuous
full-time employment are challenged by career interruptions driven by
technological displacement. Redesigning social insurance for the AI era is a
complex, politically difficult undertaking, but the alternative, applying
twentieth-century social protection frameworks to twenty-first-century labour
market disruptions, is increasingly untenable and will produce growing
political instability as the gap between protection and need
widens.

The broader imperative is political as much as personal. The
distribution of AI’s economic benefits is not determined by the technology
itself but by the policy choices societies make about taxation, labour regulation,
social insurance, and public investment. The automation divide is widening,
but it is not inevitable.

The international dimension of the
automation divide is also significant. Countries at different stages of
economic development face very different AI labour market challenges. For
economies whose development strategy has depended on competitive labour costs
in manufacturing and services, AI automation of those same sectors removes
rungs from the development ladder that earlier-industrialising countries
relied on. The economic models through which countries like South Korea,
Taiwan, and China achieved rapid development may not be available to
countries attempting that transition today. International organisations
including the World Bank and the IMF are developing frameworks for AI-era
development strategy, but the honest assessment is that no established model
exists for achieving broad-based economic development in an economy where AI
is compressing the window of comparative advantage that labour-intensive
industries previously provided.

The question is whether democratic systems can move fast enough,
and with sufficient political will, to shape an AI transition that is broadly
beneficial rather than concentratedly so. For related analysis, see our
coverage of the
AI shadow workforce and AI
in the gig economy.

About the
Author

Stuart Kerr is a technology correspondent at LiveAIWire, covering
artificial intelligence, digital innovation, and the social impact of
emerging technologies. Follow LiveAIWire for daily analysis at liveaiwire.com.