Graduate
recruitment platform Adzuna published analysis in early 2025 showing that
advertised entry-level positions in the UK had fallen by 32 percent since the
public launch of ChatGPT in November 2022. The sectors with the steepest
declines included financial services, legal, marketing, and technology,
precisely the knowledge-economy roles in which graduates have historically
secured their first professional positions. The timing correlation with large
language model adoption is not coincidental. Hiring managers interviewed for
the Adzuna report were explicit: AI tools were handling tasks previously
assigned to junior employees, reducing the economic case for entry-level
hiring at previous volumes.

The 32 percent figure has attracted significant media attention
and equally significant dispute. Labour economists have cautioned against
treating a single dataset as definitive, noting that job posting volumes are
influenced by multiple factors including the broader economic cycle, changes
in employer posting behaviour, and shifts between formal job advertising and
informal recruitment channels. The Office for National Statistics employment
data for the same period does not show an equivalent collapse in youth employment
overall. But the sectoral concentration of the decline, in exactly the
industries where AI adoption has been fastest and most visible, gives the
correlation more explanatory weight than pure economic fluctuation would
suggest.

Which Roles Are Disappearing

The entry-level roles experiencing the steepest declines are those
characterised by high volumes of structured, repetitive cognitive work:
junior paralegal research, data entry and analysis, financial modelling
support, basic copywriting, social media management, and customer service
roles with clear scripts and limited exception handling. These are not
low-skill jobs in the traditional sense; they typically require degree-level
education and specific technical knowledge. But they involve task profiles
that large language models and AI automation tools can now perform to
acceptable standards at a fraction of the labour cost.

Law firms provide a particularly clear example. The research tasks
that junior associates and paralegals have historically performed, contract
review, case law research, document drafting and summarisation, are core
capabilities of current AI legal tools including Harvey, Clio, and similar
products. Several Magic Circle firms including Allen and Overy (now A&O
Shearman) have made significant investments in AI legal tools and have been
candid about the impact on junior hiring volumes. The Law
Society has acknowledged AI’s impact on entry-level legal
employment while arguing that the long-term demand for human legal judgment
remains robust, a position that many junior lawyers and recent graduates find
less reassuring than it is intended to be.

The Training Pipeline Problem

The concern that goes beyond individual career outcomes is the
disruption to the professional training pipeline. Entry-level roles in
knowledge industries serve a function beyond their immediate output: they are
the environments in which junior professionals develop the skills, contextual
knowledge, and professional judgement that make them valuable senior
practitioners years later. A junior lawyer who spends two years doing
contract review develops an understanding of legal risk and commercial
context that cannot be acquired through formal education alone. A junior
financial analyst who builds models from scratch develops an intuition about
financial dynamics that AI-assisted modelling shortcuts do not
provide.

If the entry-level positions that function as training grounds for
professional expertise are systematically eliminated by AI, the pipeline that
produces experienced senior professionals will eventually run dry. The
effects will not be immediate; current senior professionals took a decade or
more to develop their expertise, and their capabilities will sustain the
pipeline for years. But the disruption to the training pathway is a
structural problem that will manifest in professional quality and capacity
deficits a decade from now, in ways that are difficult to address once they
have materialised.

The Counter-Arguments

Those who argue that AI-driven entry-level job losses represent
transformation rather than threat point to historical precedent and emerging
evidence. Every previous wave of automation created new roles alongside the
ones it displaced; spreadsheets eliminated bookkeeping jobs while creating
financial analysis roles that did not previously exist. AI tools are creating
demand for prompt engineers, AI trainers, and AI-assisted productivity roles
that were not in the job market three years ago. The Graduate Employment
Survey published by the Higher Education Statistics Agency in 2024 found that
graduates with AI skills were commanding a measurable salary premium in the employment
market, suggesting that human-AI collaboration competence is becoming a
valued and compensated skill.

The honest assessment is that both perspectives contain truth. The
short-term displacement of entry-level roles is real and is causing genuine
hardship for graduates entering a more competitive market with fewer rungs on
the career ladder. The long-term emergence of new AI-adjacent roles is also
real, but the timeline for that emergence and the question of whether it will
compensate at equivalent volume and quality for the roles being lost are
genuinely uncertain. The policy response to entry-level job market disruption
from AI requires engagement with the speed of change rather than its long-run
direction. The historical pattern of new job creation following technological
displacement has operated over decades; the workers displaced in the near
term cannot wait decades for the labour market to rebalance. Active labour
market policy that provides genuine income support during transition,
retraining programmes that are effective rather than merely available, and
education policy that prepares the next generation for an AI-integrated
economy rather than a pre-AI one are all components of an adequate response.
The Resolution
Foundation has published analysis of AI’s impact on UK labour
markets that provides a rigorous evidence base for policy development, and
its recommendations for how policy should respond to near-term displacement
deserve the seriousness of engagement that the scale of the challenge
warrants. For related analysis of AI and labour markets, see our coverage of
the
AI automation divide and AI
in the gig economy.

What This Means for You

For students and recent graduates, the 32 percent figure is a
signal worth taking seriously even if its interpretation is contested. The
entry-level job market in knowledge economy sectors is more competitive than
it was two years ago, and it is likely to remain so. Building explicit AI
competence alongside domain expertise, understanding how AI tools work in
your target sector, and developing the skills that complement rather than
compete with AI, including complex judgment, client relationships, creative
synthesis, and ethical reasoning, are practical adaptations to a labour
market that is changing faster than career advice services have fully
recognised.

The international dimension of entry-level job market disruption
is significant and underappreciated in UK-centric coverage. AI tools that
reduce the cost of knowledge work are disproportionately affecting the
outsourced knowledge work sectors that provide entry-level employment
opportunities in countries including India, the Philippines, and Malaysia,
where large workforces have been employed to perform the data annotation,
basic content production, and business process tasks that were previously
uneconomical to automate. The disruption of these outsourced knowledge work
markets is proceeding faster than the disruption of directly employed
entry-level roles in UK organisations, and the workers affected have less
regulatory protection and fewer alternative employment options than their UK
counterparts. A complete picture of the entry-level job market impact of AI
requires attention to this international dimension, which rarely features in
domestic employment data. For related analysis, see also our coverage of
the
shadow AI workforce.

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.