By
Stuart Kerr, Technology Correspondent,
LiveAIWire
Headline predictions that AI will
eliminate half of all jobs have circulated since ChatGPT launched in late
2022. The unemployment rate in the United States, the UK, and most developed
economies has not moved materially in the direction those predictions
implied. Yet something significant is happening in the labour market that
does not show up cleanly in headline unemployment numbers, and understanding
it requires reading the data more carefully than either the doomsayers or the
dismissers typically do.
The most authoritative recent
assessment comes from the World
Economic Forum’s Future of Jobs Report 2025, which projects 92
million roles will be displaced globally by 2030 alongside 170 million new
roles created, for a net increase of 78 million jobs. Goldman Sachs research
estimates that approximately 300 million full-time jobs globally are exposed
to automation by AI, a broader definition covering significant task changes
rather than full elimination. These numbers are large but they span five
years and they describe exposure, not confirmed loss. The near-term picture,
what is actually measurable in 2026 data, is more specific and more
instructive.
For anyone trying to assess their own
position, the aggregate numbers are the least useful part of this analysis.
The distribution matters far more than the total, and the distribution is
anything but uniform.
What Is Actually Happening Right
Now
Goldman Sachs researchers estimate that AI has reduced
monthly payroll growth in the United States by approximately 16,000 jobs over
the past year, while simultaneously adding around 9,000 jobs per month in
roles where AI augmentation increases productivity enough to drive new
hiring. The net effect on total employment is small and sits well within
statistical noise at the aggregate level. But the composition of that net
figure conceals a significant distributional shift: the jobs being suppressed
are concentrated in administrative and entry-level roles, while the jobs
being created require AI fluency alongside domain
expertise.
The hiring suppression mechanism is the most
important and least discussed dynamic in the current AI and jobs debate. AI
is not primarily replacing existing workers by firing them. It is changing
how firms manage headcount growth: employers are integrating AI to accomplish
existing workloads with fewer additional hires, meaning entry-level positions
that would previously have been created are simply not being opened. The
downstream consequence for a generation of workers who cannot find
entry-level positions to begin building career capital is potentially as
severe as direct displacement, just slower and harder to measure in
conventional unemployment statistics. A Yale Budget Lab analysis using actual
AI usage data found no clear upward trend in AI-task exposure among
unemployed workers, confirming that the disruption is suppressing hiring
rather than creating redundancies.
The Roles Actually
Growing
LinkedIn’s 2026 Jobs on the Rise report found that
AI Engineer is the number-one fastest-growing job title in the United States,
with postings rising 143 percent year over year. Four of LinkedIn’s top five
fastest-growing roles in 2026 are AI-related. The Stanford
HAI 2026 AI Index confirms that AI skills now appear in
approximately 2.5 percent of all US job postings, up 55 percent year over
year, with agentic-AI skill mentions growing more than 280 percent in a
single year. The labour market is repricing for AI fluency well ahead of any
aggregate job losses, which is consistent with the pattern where the demand
for AI-capable workers precedes rather than follows the broader deployment of
AI in those workers’ industries.
The WEF identifies the
fastest-growing role categories as technology and engineering specialists,
green transition roles, and care economy workers. The manual and interpersonal
roles that earlier automation waves left largely untouched, including
frontline service, delivery, construction, and healthcare delivery, are
projected by the WEF to see the largest absolute job growth by 2030. This is
consistent with the pattern that large language models and generative AI
tools perform worst on tasks requiring physical presence, real-time judgment
in unpredictable environments, and genuine interpersonal care relationships.
Trades and care work are structurally protected in ways that
information-processing work is not.
Who Faces the Real
Risk
The Goldman Sachs estimate that only 2.5 percent of
US employment is at near-term displacement risk is sometimes cited to argue
that AI job fears are overstated. That interpretation misses the
concentration. The 2.5 percent is not evenly spread across the workforce: it
is concentrated in administrative support, junior professional roles in law
and finance, certain categories of content production, and entry-level
positions in sectors early to adopt AI tools. For workers in those specific
categories, the risk is substantially higher than 2.5 percent of their
personal employment would suggest.
The structural concern
about career pipelines is separate from and in some ways more serious than
direct displacement. When AI absorbs the entry-level volume tasks that
traditionally provided training grounds for junior professionals, the
long-term supply of expertise in those professions is affected. Legal
associates, junior financial analysts, and early-career developers in firms
that have significantly reduced entry-level hiring are not being fired. They
are not being hired in the first place, and the experience that would have
made them valuable mid-career professionals is not being accumulated. For the
broader
framework of how AI is dividing workers into those who benefit and those who
face displacement, this pipeline effect is the most consequential
development that current employment statistics are not capturing. The answer
to “will AI take your job” depends less on the aggregate numbers
and more on whether your specific occupation sits at the intersection of
structured task processing, information synthesis, and communicable outputs,
because those are precisely the conditions where AI tooling currently
performs best.
What to Do About It
The
evidence is sufficiently clear to support specific preparation rather than
generic anxiety. Workers in high-exposure roles should develop the ability to
use AI tools effectively within their domain, shifting their personal value
proposition toward the judgment, client relationships, and contextual
reasoning that AI handles poorly. Understanding how
automation is affecting where and how people choose to work adds
context to the geographic dimension of this shift. Workers in low-exposure
roles face different pressures, primarily the question of whether AI fluency
will eventually become expected across roles that are not currently
AI-exposed. The data in 2026 suggests the answer to that question is yes, on
a five-year horizon, rather than imminently. Using that horizon as a
preparation window rather than a comfort to avoid preparation is the more
useful interpretation. Watching how AI
companies are preparing to meet public market accountability also
illuminates which sectors the industry itself believes will see the highest
continued AI investment, which is useful signal for workers making
medium-term career decisions about where AI capability is developing
fastest.
The geographical picture adds a dimension that
national averages obscure. AI adoption in the labour market is concentrated
in specific metropolitan areas, industries, and firm sizes. The firms
reducing entry-level hiring due to AI tools are disproportionately large
enterprises in major cities in professional services, technology, and
finance. Smaller firms in distributed locations are adopting AI more slowly
and seeing fewer of the hiring suppression effects. For workers deciding
where to develop their careers, the local employer mix matters as much as the
occupation classification in assessing personal AI risk
exposure.
About the Author
Stuart Kerr
is Technology Correspondent at LiveAIWire, covering artificial intelligence,
cybersecurity, and the social impact of emerging technology. He publishes
daily at LiveAIWire.com.