The
IMF published analysis in 2024 estimating that approximately 40 percent of
jobs globally are exposed to AI automation, a figure that has been cited in
thousands of articles, political speeches, and business strategy documents in
the months since its publication. The figure is real, carefully derived, and
significantly misrepresented in most of the coverage it has generated.
Understanding what the 40 percent figure actually measures, what it does not
measure, and what the evidence suggests about the pace and character of
AI-driven employment change is essential context for workers, employers, and
policymakers who are making consequential decisions based on how they
interpret it.

The IMF’s methodology defines job exposure to AI automation based
on the extent to which the tasks comprising a job involve the kinds of
cognitive, language, and pattern recognition activities that AI systems
currently perform well. A job with high AI exposure is one in which a large
proportion of the tasks involved could, in principle, be performed by current
AI systems. This is importantly different from saying that 40 percent of jobs
will be eliminated by AI, which is a claim the IMF did not make. Jobs that
are highly exposed to AI automation may be transformed by it, with some tasks
automated and others changed in character, rather than eliminated wholesale.
The distinction matters enormously for how individuals and institutions
should respond to the evidence, and it is a distinction that the apocalyptic
framing prevalent in much AI jobs coverage systematically
elides.

What Job Exposure Actually Predicts

The empirical research on what happens to jobs when AI tools are
adopted in the relevant sector provides more nuanced guidance than the
theoretical exposure framework alone. Studies examining the introduction of
AI coding assistance in software development, AI document processing in legal
services, and AI analysis tools in financial services all find patterns that
differ from both the apocalyptic and the dismissive interpretations of AI’s
labour market impact. Tasks that are highly routine and well-defined are
automated; tasks that require contextual judgement, client relationship
management, exception handling, and creative problem-solving are not. Roles
that consist primarily of the former category face substitution pressure;
roles that consist primarily of the latter are augmented. Most roles consist
of a mixture, and the outcome depends on how the task composition of the role
shifts as AI handles more of the routine elements.

The labour market data available in 2025 documents clear effects
in specific sectors and job categories that are consistent with the exposure
framework. Entry-level positions in financial services, legal support,
marketing, and software development have declined measurably since the
widespread adoption of AI tools capable of performing many of the tasks those
positions involved. Research from the University of Chicago and Stanford has
documented wage effects showing that workers whose jobs are highly exposed to
AI tools are experiencing slower wage growth than those in less exposed
roles, a finding that suggests AI exposure is already affecting labour market
outcomes even where it has not yet caused direct job elimination. The National Bureau of Economic
Research has published multiple working papers examining these
dynamics with methodological rigour that the headline-generating research
often lacks.

Which Workers Are Most Affected

The distribution of AI employment impact across the workforce
defies the simple narrative that AI primarily threatens low-skill, low-wage
workers. The current wave of AI automation is affecting knowledge work and
cognitive tasks in ways that previous automation waves, which concentrated on
manufacturing and physical labour, did not. This means that the workers most
exposed to AI-driven task automation include significant numbers in
professional, technical, and managerial occupations who have historically
been protected from automation by the cognitive complexity of their work. The
compression of the skills premium, the wage advantage that skilled workers
command over less skilled workers, is a potential consequence of AI reducing
the scarcity of cognitive task performance that has historically been the
basis of that premium.

At the same time, workers in highly exposed occupations who
develop the skills to direct, evaluate, and complement AI tools effectively
are experiencing different outcomes from those who cannot or do not adapt.
The polarisation between workers who use AI as a productivity multiplier and
those who are displaced by it is one of the most significant inequality
dynamics of the current period. Geographic concentration of this polarisation,
with workers in AI-hub cities having far better access to AI-adjacent
employment opportunities than those in other regions, compounds the
inequality dimension further. Research from the Resolution Foundation and the
Institute for Fiscal Studies in the UK has documented this geographic
polarisation in detail, providing evidence that the aggregate national
employment figures obscure significant regional variation in AI’s labour
market impact.

The Policy Response Question

The 40 percent exposure figure has generated policy responses
ranging from calls for universal basic income to proposals for robot taxes,
with most falling somewhere in between these positions without clearly addressing
the specific mechanisms through which AI is affecting the labour market. The
most evidence-based policy responses focus on three areas: education and
skills infrastructure that prepares workers for AI-augmented roles rather
than AI-substituted ones; social insurance frameworks that provide adequate
support for workers in transition between roles; and active labour market
policies that help match displaced workers with the growing demand for
AI-adjacent skills.

The UK
government’s AI Opportunities Action Plan published in 2025
acknowledges the workforce transition challenge while emphasising the productivity
and growth opportunities of AI adoption. The plan’s workforce provisions have
been criticised by trade unions and some economists as insufficient relative
to the scale of disruption that the government’s own AI optimism implies. The
gap between the ambition of AI adoption goals and the adequacy of workforce
transition investment is a structural inconsistency in AI policy across most
major economies that has not yet been resolved.

What This Means for You

If you are in a role that involves significant proportions of
structured cognitive work, document processing, data analysis, or routine
professional judgement, the 40 percent exposure figure is personally
relevant, but it does not predict your specific outcome. The most important
individual response is honest assessment of which components of your current
role are most and least exposed to AI substitution, combined with deliberate
investment in the skills that complement rather than compete with AI.
Understanding AI tools in your specific professional domain, developing the
judgement to evaluate and improve AI outputs rather than simply accept them,
and building the relational and contextual skills that AI lacks are the most
durable adaptations available. For related analysis, see our coverage of
the
automation divide and AI’s
impact on UK entry-level employment. The apocalypse framing is
unhelpful; so is complacency. The evidence points toward significant
disruption that requires active individual and policy response, distributed
unevenly across the workforce and concentrated in ways that existing social
insurance systems are not yet designed to handle.

 The most honest framing of AI’s
labour market impact is neither apocalyptic nor dismissive. It is that
significant disruption is already underway in specific sectors and job
categories, that the pace is accelerating, that the distribution of impact is
unequal in ways that compound existing inequalities, and that the policy
response in most countries including the UK is not yet commensurate with the
scale of the challenge. The IMF’s 40 percent figure is a useful frame for the
scale of potential disruption; it is not a prediction of mass unemployment but
a signal that active policy engagement, investment in education and
retraining, and reform of social insurance frameworks are urgent priorities
that cannot be deferred until the disruption is more advanced. The Institute for Fiscal
Studies has published UK-specific analysis of AI labour market
impacts that provides more granular evidence for UK policy than the global
IMF estimates.

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.