How Many UK Jobs Has AI Already Replaced? Evidence Gathered

AI isn’t a future threat to UK jobs — it’s already here, quietly reshaping hiring, freezing entry-level pipelines, and driving measurable displacement across white-collar sectors. While no official dashboard tracks “AI job losses” in real time, 2025–2026 data from government assessments, independent think tanks, academic studies, employer surveys, and labour market statistics paint a clear picture: silent displacement is underway. Vacancies have cratered, firms exposed to AI have cut headcount, and employers openly expect workforce shrinkage in the next 12 months. Yet this isn’t mass redundancy — it’s structural contraction through unfilled roles, reduced hiring, and automation of routine tasks.

This article synthesises the latest evidence to explain exactly how many UK jobs AI has already impacted, who is most affected, why causality remains debated, and what the projections mean for 2025–2026 and beyond. The data shows displacement is real but uneven: concentrated in high-exposure, high-salary occupations; hitting juniors hardest; and occurring alongside productivity gains that could ultimately create new roles. The UK faces a transition shock that demands urgent reskilling, not panic.

The Evidence of Silent Displacement: Vacancies and Employment Shifts

UK job vacancies collapsed 43% from 1.3 million in May 2022 to just 700,000 by May 2025. Roles most susceptible to AI — like software development — fell 37% since ChatGPT’s November 2022 launch. These figures come from aggregated job posting data and align with broader ONS labour market trends showing payroll employment declines of 121,000 (0.4%) between December 2024 and December 2025.

A landmark King’s College London study, analysing millions of job postings and LinkedIn profiles from 2021 to 2025, provides the sharpest firm-level evidence. Firms whose workforces were highly exposed to large language models (LLMs) reduced total employment by 4.5% on average after ChatGPT’s release. The impact hit junior positions hardest: down 5.8%. High-paying firms saw employment drop 9.6%, while low-paying firms were largely unaffected. Highly exposed roles experienced a 23.4% drop in job postings and a 6.3% (£2,951) cut in advertised salaries.

This isn’t broad layoffs — it’s a compositional shift. Companies post 16.3 percentage points fewer vacancies overall and pivot toward senior or customer-facing roles that require human interaction. Average pay within AI-exposed firms rose over £1,300 as junior headcount shrank, effectively upskilling the remaining workforce at the expense of entry points.

The UK government’s own January 2026 assessment confirms the pattern: a one-standard-deviation increase in AI exposure correlates with a 3.9% reduction in posting volume, statistically significant seven months post-ChatGPT and concentrated in high-salary occupations. McKinsey’s analysis (cited in the report) shows UK job adverts fell 38% for high-exposure roles versus 21% for low-exposure ones between 2022 and 2025. Early-career indicators are stark: the number of 16–24-year-olds in UK computer programming dropped 44% in 2024 alone, and youth unemployment (16–24) hit 15.9% in September–November 2025 — the highest since 2020 — with 729,000 young people out of work, up 103,000 year-on-year.

UK Government Research: Suggestive but Not Yet Causal

The Department for Science, Innovation and Technology’s Assessment of AI Capabilities and the Impact on the UK Labour Market (published 28 January 2026) is the most authoritative official view. It states plainly: “Data suggests hiring is falling faster in occupations more exposed to AI, but establishing whether AI is causing these patterns remains challenging.”

The report highlights rapid AI capability growth — task length and complexity for autonomous agents doubling every seven months in coding, cybersecurity, and research — yet adoption remains modest. Only one in five UK firms use or plan to use AI; within adopters, less than one-third of employees engage with it. Around 70% of UK workers are in occupations containing AI-exposed tasks (higher than the US), per IMF estimates, split roughly evenly between high- and low-complementarity roles.

Productivity upside is clear: OECD projections suggest UK labour productivity growth of 0.4–1.2 percentage points annually over the next decade. Experimental studies show AI boosting output in software development (56%), writing (59%), and consulting (25%). But the assessment cautions that firm-level productivity links are not yet robust, and observed hiring declines could stem partly from interest-rate sensitivity in overlapping sectors (tech, finance, professional services) rather than AI alone.

In short, government data for 2025–2026 treats AI as a contributing factor to subdued hiring and early-career contraction, not the sole driver. Causality gaps persist because exposure ≠ adoption, and macroeconomic headwinds complicate attribution.

Independent Research and Think Tank Projections

Think tanks provide the boldest numbers — and the clearest forward-looking scale.

The Tony Blair Institute (TBI) estimates 1–3 million private-sector UK jobs could ultimately be displaced by AI, with peak annual losses of 60,000–275,000. These are “relatively modest” compared to the UK’s historical average of 450,000 annual job losses. Using O*NET task-level analysis and GPT-4 modelling, TBI projects AI could save 23.8% of private-sector working time under full adoption, mostly via software on cognitive tasks. Displacement is gradual, peaking in the 2030s–2040s, with unemployment impacts in the low hundreds of thousands before new tasks and growth unwind it. Their “Tailwind” (most likely) scenario sees 1.5 million total displacements and a 340,000 unemployment peak around 2040.

The Institute for Public Policy Research (IPPR) paints a starker worst-case: up to 7.9 million roles at risk if AI fully displaces exposed tasks with no new job creation or GDP gains. Central scenario: 545,000 jobs lost but with 3.1% GDP uplift. IPPR emphasises white-collar knowledge work — up to 70% of organisational and analytical tasks transformable — and flags disproportionate risks for women, younger workers, and lower-paid roles.

The National Foundation for Educational Research (NFER) projects up to 3 million UK jobs in declining occupations (administrative, secretarial, customer service, machine operations) could disappear by 2035 due to AI and automation — faster than prior forecasts. Overall, the economy may still add 2.3 million jobs, but the distribution will be highly uneven.

Morgan Stanley research (shared January 2026) underscores the UK’s outlier status: British companies reported net job losses of 8% over the past year due to AI — twice the global average and the highest among major economies. UK firms cut or left unfilled about a quarter of roles but were far less likely to hire replacements, unlike US peers.

Employer Sentiments and Sector-Specific Impacts

CIPD’s Autumn 2025 Labour Market Outlook (survey of over 2,000 employers) is blunt: 17% expect AI to shrink their workforce in the next 12 months. Of those, 62% target clerical, junior managerial, professional, or administrative roles. Large private-sector firms lead at 26% anticipating cuts; nearly one in four predict reductions exceeding 10% of staff. Only 11% of British businesses planned headcount growth in Q4 2025 — the steepest decline among 42 countries surveyed.

At-risk sectors align across sources: administrative/clerical (document handling, data processing), customer service (chatbots), junior tech (coding, data analysis), legal research, and finance operations. Resilient roles require physical presence, complex interpersonal skills, or strategic creativity — construction trades, care work, senior leadership.

Public discourse on X reflects this tension. Posts from early 2026 highlight real-world examples (Klarna replacing 700 support reps, Amazon cutting 14,000 managers) while older predictions (WEF’s 85 million global jobs by 2025) are mocked as hype. UK-specific chatter focuses on youth unemployment spikes and “silent” entry-level freezes, with users noting AI agents now handling entire workflows — planning, execution, iteration — beyond single tasks.

Net Effects, Global Context, and the 2025–2026 Outlook

The World Economic Forum’s Future of Jobs Report 2025 projects 92 million global jobs displaced by 2030, offset by 170 million new roles — a net gain of 78 million. For the UK, short-term 2025–2026 data shows contraction: subdued hiring, youth unemployment pressure, and firm-level cuts in exposed areas. But productivity gains (already self-reported at up to 20% by adopters) and new AI-related jobs (projected to reach 3.9 million by 2035 per government skills modelling) suggest transformation over outright destruction.

The key 2025–2026 reality: displacement is front-loaded in entry-level and routine cognitive work, creating a “training pipeline” crisis. Without intervention, skills gaps widen — 3.7 million UK workers currently lack essential digital/problem-solving skills, potentially doubling to 7 million by 2035.

Reskilling as the Imperative

Policymakers, employers, and individuals must treat AI literacy and human-complement skills (creativity, strategy, interpersonal) as non-negotiable. Government reports call for better data, faster adoption tracking, and targeted support for at-risk cohorts. Employers using AI should pair productivity wins with internal mobility and training pipelines. Workers: double down on AI-augmented roles now.

The data is unambiguous for 2025–2026: AI has already replaced tasks equivalent to tens of thousands of UK jobs through hiring restraint and targeted reductions. Projections signal hundreds of thousands more annually at peak. But history shows technological shifts ultimately expand opportunity when managed proactively. The UK’s service-heavy economy and high exposure give it both the greatest risk and the greatest reward.

The future of work isn’t zero-sum. It’s adaptive. Those who reskill fastest — organisations and individuals alike — will thrive in an AI-augmented economy. The window for action is now.