UK Migration — The AI Labour-Market Frame

One of seven companion framings, the most rapidly evolving evidence base in the document. AI is currently displacing high-paid white-collar work faster than low-paid migrant-dependent sectors. The King's College London October 2025 study found firms with high AI exposure cut total employment 4.5% and junior positions 5.8% (2021-2025). The AI frame complicates restrictionist assumptions about automation replacing migrant labour and supports adaptive sectoral planning. Presented at full strength.

Migration policy in an age of labour-market automation

Framing: This article addresses a variable largely missing from the master document and the three companion framings: the impact of artificial intelligence and automation on the UK labour market over the next decade, and what that means for migration policy. The data is moving fast and the picture is more complex than the headline framing allows. This article uses the most current available evidence (to May 2026) and tries to represent the picture honestly rather than reaching for the obvious story.

Health warning: This is the framing where evidence is changing fastest. Some figures cited here will be out of date within twelve months. The structural points should remain useful for longer.

The headline story is partly wrong

The intuitive narrative goes: AI is coming for low-paid work. Robots will replace warehouse workers, self-checkout will replace cashiers, automated kitchens will replace hospitality staff, AI will replace care work. Therefore, the UK has less need for low-paid migration than current sectoral demand suggests, and migration policy can tighten without creating workforce shortage.

The current evidence does not support this story. It supports a more uncomfortable one.

Where AI is actually hitting the UK labour market hardest, right now (2024-2026), is in higher-paid white-collar work. A King's College London study published October 2025 analysed millions of UK job postings and LinkedIn profiles 2021-2025. Firms whose workforces are highly exposed to AI capabilities reduced total employment by 4.5% on average between 2021 and 2025, with the effect concentrated in junior positions which fell by 5.8%. Higher-paying firms saw job losses of about 9.4% across the same period — much of it after ChatGPT's release in November 2022. Professional roles in management consultancy, psychology, and legal services have been flagged as particularly exposed.

Government's own Assessment of AI capabilities and the impact on the UK labour market (January 2026) reports: "A separate analysis from McKinsey found that between 2022 and 2025, UK job adverts fell by 38% for high-exposure occupations compared to 21% for low-exposure roles." The decline concentrated in high-salary occupations, with no significant change observed in low-salary roles in this period.

Most strikingly: UK digital sector employment dropped for the first time in a decade in 2024. The number of 16-24-year-olds in computer programming fell 44% in a single year. Coding — once thought to be the safe, high-value, future-proof career — is being directly displaced by AI coding tools.

Meanwhile, hospitality (a low-AI-exposure sector) accounted for 53% of UK job losses between October 2024 and August 2025. Those losses are not AI-driven. They reflect cost pressure, reduced consumer spending, post-pandemic structural change, and the impact of immigration restrictions reducing labour supply in some sub-sectors.

The current picture, then, is the opposite of the intuitive story. The graduates in £40,000-£80,000 jobs are being displaced. The Filipino health care worker in £30,000 social care work is not — yet.

But the medium-term picture is different

The picture changes when you look beyond the next five years.

UK roles overall: studies suggest 25-30% of current UK jobs could be affected by automation by 2030. The British Safety Council research projects up to 3 million low-skilled UK jobs displaceable by 2035.

Where the displacement is moving — robotics rather than pure LLM displacement — is into specifically the sectors that currently depend on migration:

Warehouse and logistics. Autonomous Mobile Robots (AMRs), automated sortation systems, robotic picking, depalletising and packing. UK robotics sector saw £4.3 billion in investment in twelve months to mid-2025; warehouse automation is one of the fastest-growing sub-sectors. The implication: warehouse labour demand falls through the late 2020s.

Agriculture. Robotic harvesters, weeding robots, milking systems, autonomous tractors. The Seasonal Worker Visa scheme handles approximately 45,000 workers per year, predominantly in soft fruit and vegetables. Strawberry-picking robots are now operational at commercial scale; broccoli harvesting is being automated; lettuce picking is being mechanised. The implication: seasonal agricultural labour demand falls through the early 2030s.

Food processing. Hygienic robots for case packing, palletising, portioning, and quality inspection are operational. Slaughterhouse and processing roles — historically migrant-dependent — are being automated faster than the white-collar wave because the physical infrastructure is already mature.

Healthcare and assistive robotics. Robots supporting clinical workflows, surgical robotics (the NHS surgical-robot procurement boom is significant), and assistive technology in elderly care. Care robots are not yet operational at scale but research and pilot deployment are accelerating.

Hospitality. Cleaning robotics, kitchen automation, automated front-of-house systems. Slower deployment than warehouse but moving.

Construction. Robotic bricklaying, autonomous diggers, precision construction technology. Slower than other sectors due to site-condition variability but moving.

By the early 2030s, the pattern reverses. The cohorts being displaced shift from white-collar professionals to the migrant-dependent low-paid sectors. Workers currently in social care, agriculture, food processing, warehouse, and hospitality face significantly different career trajectories than those in skilled white-collar work because robotics technology is maturing at different rates than LLM technology has.

The gap between exposure and displacement

A subtlety worth flagging because it confuses the debate.

"Exposure" — the proportion of a job's tasks that AI could in principle perform — is not the same as "displacement" — the proportion of jobs actually lost. Eloundou et al. (2024) estimate up to half of all jobs may have the majority of their tasks "exposed" to generative AI in the near term. But task exposure does not mean workforce reduction at the same rate. Jobs adapt. New tasks emerge. Productivity gains are reinvested. Workers shift to tasks AI cannot do.

The Centre for British Progress's analysis (April 2026) of AI usage versus displacement notes: "Professional and managerial occupations show the widest divergence: experts rate them as highly exposed, but actual AI usage lags well behind." The gap matters. High exposure with low displacement is one outcome; high exposure with high displacement is another; both are happening in different sectors at different rates.

The empirical pattern in UK data 2024-2026:

• High-exposure white-collar work is showing actual displacement effects, not just task automation
• Low-exposure low-paid work is showing minimal AI displacement (though it is showing other forms of job loss)
• Mid-skill technical work (programming, junior consulting, junior legal work) is showing the sharpest displacement

This is not the picture the Frey-Osborne 2013 work predicted. That work emphasised routine manual tasks. Generative AI displaces routine cognitive tasks faster than routine manual tasks.

Why this matters for migration policy

The intersection with migration policy operates on multiple time horizons and goes in different directions at the same time.

If AI is displacing graduate-entry white-collar work and junior professional roles, the UK domestic labour force has more pressure on it, not less. Young UK-born graduates entering a labour market with fewer junior roles in their fields cannot easily move into the migrant-dependent sectors (social care, agriculture, food processing) because those sectors require different skills, different geographic locations, and offer different wages. The Conservative-Reform argument that migration restriction will produce domestic workforce expansion is harder to make when the displacement is happening in the white-collar sectors that domestic workers were already in, not in the low-paid sectors that domestic workers have largely left.

The hospitality sector accounting for 53% of recent UK job losses is partly a labour-supply story (some of it driven by post-Brexit migration restrictions) but those displaced hospitality workers cannot easily fill social care vacancies. Sectoral mismatch is structural.

The Government's own May 2025 Immigration White Paper takes a position on this: it explicitly cites "investment in automation" as part of the rationale for reducing overseas recruitment. The implicit assumption is that automation will displace migrant labour in low-paid sectors, allowing migration restriction without sectoral collapse. The current evidence suggests this is at best premature and at worst back-to-front: automation is currently displacing the kind of workers who were not migration-dependent, while the migrant-dependent sectors are showing limited automation displacement so far.

When robotics maturity reaches social care, warehouse, agriculture, and food processing at scale — projected through the early 2030s — the workforce demand in those sectors changes substantially. Migration policy that maintains current sectoral recruitment levels through 2030 may find itself with surplus migrant labour in the displaced sectors by 2035.

This is a real planning question. Workers admitted on Health & Care visas in 2026 will be eligible for ILR in 2036 (under the 10-year qualifying period) or 2031 (under the previous 5-year rule). If automation has displaced significant proportions of the social care sector by then, those workers are settled with limited route-relevant labour-market demand. The Green Card analogy in the US — workers admitted for industries that subsequently shrunk — shows the political and economic complexity of this trajectory.

Whether displaced white-collar workers in the 2025-2028 cohort retrain, what new roles emerge that AI cannot perform, whether robotics maturity is faster or slower than current trajectory, what happens to productivity and wages, what happens to fertility (a longer-term automation pattern interacts with the demographic frame discussed in the demographic article) — all of these are genuinely uncertain. The OBR fiscal sustainability work is currently being updated to incorporate AI productivity assumptions; the next iteration will likely change the migration-need projections substantially.

Specific implications for the policy options menu

Several of the options in the master document's Part II look different through the AI lens.

Option 4 (Earned Settlement extension) becomes more politically defensible if AI displaces low-paid sector demand. The 10-year qualifying period gives time for sectoral demand to shift before settlement; if a Health & Care Worker is admitted in 2026 and qualifies for ILR in 2036, by which point the sector has been substantially automated, the settlement decision can be made on lifetime contribution evidence including the AI-displacement effect. Under the previous 5-year rule, settlement decisions are made before AI effects are visible.

Option 8 (Danish-model integration investment) becomes more important under AI displacement. If white-collar workers are being displaced in the near term and migrant-sector workers are likely to be displaced in the medium term, integration investment that focuses on transferable skills, retraining capacity, and language fluency becomes essential for both populations. The Danish-model framing — substantial public investment in language and labour-market integration — addresses the AI-displacement question alongside the migration-integration question.

Option 11 (Returns capacity expansion) becomes harder to scale. If voluntary returns are disproportionately taken up by workers who could find employment elsewhere, AI displacement of UK low-paid work may reduce the willingness of migrants to return voluntarily — there are fewer "back home" jobs to return to as automation displaces global low-paid work too. The 11x cost differential (£4,300 voluntary vs £48,800 enforced) may compress over time.

Option 12 (Restoration of HMRC non-UK nationals tax statistics) and Option 1 (Migration Fiscal Ledger) both become more important under AI displacement. As the labour market changes, the data needed to track outcomes by route, sector, age, and time becomes essential. A government making migration policy in 2030 without comprehensive data on which 2026-cohort migrants are still in employment and which sectors are still functional cannot make rational policy.

Sectoral migration design becomes more important than aggregate caps. If different sectors face different AI displacement timelines, generic numerical caps make less sense than sector-specific workforce planning that integrates with industrial strategy and skills strategy. The Labour Market Evidence Group structure created by the May 2025 White Paper is the right institutional shape for this; whether it is given sufficient resource and authority remains to be seen.

Specific implications for party briefings

Labour's automation-and-upskilling framing becomes more empirically supported. Labour's Industrial Strategy and Skills England framework is structurally aligned with what AI labour-market evidence suggests is needed: domestic workforce capability building paired with controlled migration. The current implementation is criticised for being insufficiently funded but the framework fits the evidence.

Conservative and Reform "deportation as labour-market policy" becomes harder to defend. The implicit assumption — that removing migrants creates jobs for British workers — was always contested empirically. Under AI displacement of white-collar work, it becomes much weaker. British workers losing junior consulting and legal jobs to AI cannot replace migrant social care workers. The sectoral mismatch is structural and cannot be solved by deportation.

Reform's £14.3 billion fiscal saving claim becomes more questionable. The model behind that claim does not account for AI displacement effects on either side. Removing migrant workers from sectors that may be automated by 2035 produces less fiscal saving than assumed if the workers had paid tax for years before automation reduced their contribution. Conversely, if white-collar AI displacement reduces UK tax revenue more than projected, the relative fiscal contribution of migrant workers in unautomated sectors increases.

Lib Dem coordination framing becomes stronger. International coordination on AI governance, labour-market transition, and skills frameworks is a serious requirement of the AI age. Pure-restrictionist national policies cannot address the scale of the labour-market change AI is producing.

Green economic-justice framing becomes more salient. AI displacement creates substantial winners (capital, AI infrastructure providers, automation-deploying firms) and losers (displaced workers across sectors). The economic-justice framing that the Greens hold is structurally aligned with this distributional question. Universal Basic Income proposals, AI productivity dividend ideas, and tax-on-automation proposals fit naturally.

SNP and Plaid demographic case becomes more complicated. The case for managed migration to address demographic decline is partially weakened if AI productivity offsets workforce shrinkage. But the regional sectoral dependencies (NHS Scotland, Welsh agriculture and processing) remain, and may even sharpen if AI displaces other workforce sources first.

Reform's net-negative migration target becomes more defensible at long horizons but more dangerous at short horizons. If AI productivity gains are realised, smaller workforce can support same population. But the short-term mismatch — migrants in displaced white-collar work versus migrants in essential low-paid work — produces specific shortages that net-negative policy cannot address.

What the data does not yet tell us

The honest version of this article includes substantial uncertainty:

The pace of robotics deployment in low-paid sectors is contested. Optimistic estimates suggest substantial automation by 2030; pessimistic estimates push it to the 2040s. The differences are not minor — they materially affect what migration policy should do.

The displacement-versus-augmentation question is contested. Brynjolfsson, Klein Teeselink, and others argue current AI is producing displacement; Acemoglu, Restrepo, and others argue it is more often producing augmentation (workers using AI tools become more productive rather than being replaced). Both effects are happening; the balance is uncertain.

The new-job-creation question is contested. Some of the displaced white-collar work is being replaced by new AI-related roles — prompt engineering, AI quality assurance, AI infrastructure. Whether new role creation matches displacement scale is unclear and varies by sector.

The productivity-wage transmission is uncertain. AI may produce significant productivity gains. Whether those gains translate to higher wages, higher profits, lower prices, or some combination depends on labour-market institutions, competition policy, and political choices.

The spatial geography is uncertain. AI displacement of white-collar work concentrates in cities (where white-collar work concentrates). AI displacement of physical work could either concentrate in cities (where automation infrastructure deploys first) or in rural areas (where agricultural automation matures). Different geographic patterns produce different migration policy implications.

The interaction with global migration trends is uncertain. If AI displaces low-paid work globally — which is the medium-term trajectory — labour-source countries face their own employment crises, with implications for migration pressure on UK and other receiving states.

What follows: recommended additions to the policy options menu

The master document's Part II should include three additional options that emerge from the AI framing:

Option A: AI-aware sectoral workforce planning. Empower the Labour Market Evidence Group with authority and resources to model AI displacement timelines by sector, integrating with Skills England and Industrial Strategy. Migration policy cap-setting, visa-route specification, and settlement-condition design should reflect AI projections, not just current labour market state. Cost: £20-50m per year for analytical capacity. Saving: substantial avoided cost from policies that would otherwise produce sectoral mismatch.

Option B: Productivity-displacement linked migration policy. Where AI productivity gains are realised in specific sectors, migration intake for those sectors reduces in step. The Migration Fiscal Ledger (Option 1 in the existing menu) should track sectoral productivity alongside fiscal contribution to allow this.

Option C: AI transition fund linked to migration policy. Some proportion of Immigration Skills Charge revenue, currently directed at general upskilling, should be specifically directed at retraining workers in sectors facing AI displacement. This creates a fiscal link between migration revenue and domestic workforce resilience that addresses both the fiscal and political case.

These options sit alongside, not in place of, the existing fourteen.

Conclusion: the AI question and the political moment

The intersection of AI displacement and migration policy is genuinely complex. The intuitive narrative — AI replaces low-paid work, so we need less migration — is empirically wrong about the current period. The reverse is happening: AI is displacing high-paid work first. The medium-term picture, with robotics maturing into low-paid sectors, eventually produces the intuitive picture, but on a timeline of 5-15 years rather than now.

This timing matters politically. Migration policy decisions made in 2026 with assumptions about 2035 sectoral demand have to navigate substantial uncertainty. Decisions taken now to admit Health & Care workers on 10-year settlement paths are being made in a labour market that may look very different by the time those workers reach settlement. Decisions taken now to restrict migration on the assumption that automation will replace migrant labour are being made before the automation has actually happened.

The honest position: AI labour-market displacement is real, is happening unevenly across the economy, is currently affecting different demographic and migration cohorts than the intuitive narrative suggests, and will substantially reshape what migration policy is for over the next 10-15 years. None of the parties in the master document have fully integrated this into their packages. The framework that comes closest is the Labour-government Labour Market Evidence Group structure, but its delivery capacity and political authority are uncertain.

For voters and policy-makers thinking about the 2029 general election: the AI variable is one that will move significantly between now and then, and the parties' positions will need to evolve with it. A migration policy framework that does not engage with AI labour-market change is a policy framework that will be obsolete by 2030. A framework that engages with AI seriously requires capacity for adaptation, which the more rigid restrictionist proposals lack and the more flexible coordination-focused proposals have.

The policy debate has not yet caught up with the labour-market reality. This article tries to help close that gap; the data will continue to move; the conclusions will need updating.

This article is one of seven companion framings to the master document, addressing variables the master under-serves. The other six — cohesion, protection, demographic, capacity, emigration, sovereignty — apply the same evidence base from those framings. Together they offer a more balanced view than the master document's fiscal-balance privileging would suggest.

Compiled using public sources to May 2026. AI labour-market evidence is moving fast; readers checking later than mid-2026 should verify current figures. Errors are the author's; data is sourced.

Sources

• King's College London, Bouke Klein Teeselink (October 2025), New study reveals early impact of AI on job market in UK
• Government Office for Science (January 2026), Assessment of AI capabilities and the impact on the UK labour market
• McKinsey UK labour market analysis (2024-2025)
• British Safety Council (2024), Navigating the Future: Safer Workplaces in the Age of AI
• ONS analysis of jobs at high risk of automation (2019, with updates)
• Centre for British Progress (April 2026), AI and the UK labour market: the evidence so far
• Brynjolfsson et al. (2025), Beyond Automation: Redesigning Jobs with LLMs
• HM Government (May 2025), Restoring Control over the Immigration System
• Eloundou et al. (2024), AI exposure analysis
• Acemoglu and Restrepo (2018), automation and labour augmentation
• Frey and Osborne (2013), The Future of Employment
• techUK and DLA Piper analyses of the May 2025 Immigration White Paper
• Labour Market Evidence Group establishment documentation
• UK robotics sector reports (2025-2026)