AI in the UK is a political economy issue, contrasting private sector labour substitution with public sector capacity enhancement to boost service quality. While the private sector focuses on margin expansion and potential job displacement to concentrate wealth in fewer hands, the public sector can leverage AI to reduce administrative tasks and empower staff, aiming to improve productivity in human-facing roles and enhancing social value. Moreover, if the disappearance of well-paid jobs in the private sector profoundly disrupts the labour market and creates a pool of unemployed skilled workers, the public sector will need to step in by providing socially useful roles, most sensibly through a guaranteed job scheme.
From a Labour councillor’s perspective, the most important question about Artificial Intelligence is not whether it will increase productivity. It almost certainly will in many settings. The real question is who captures the gains, who carries the risks, and what institutions we build around it. In that sense, AI is not just a technology story. It is a political economy story.
The UK Government has plainly decided that AI is now central to both growth policy and state reform. The Government’s AI Opportunities Action Plan response was published in January 2025, and the Prime Minister’s foreword framed AI as a tool to “make it work for working people” while transforming public services and growth. A year later, the Government’s “One Year On” update reported that 38 of 50 actions had been met, and highlighted measures including 5 AI Growth Zones, more than 1 million AI courses delivered through industry partners, AI-assisted NHS chest X-rays (2.4 million scans, described as one third of NHS chest X-rays), and major compute and sovereign AI funding commitments. Alongside that, the AI Playbook for the UK Government (published February 2025) set out guidance for civil servants and government organisations, including 10 principles for safe and effective use.
That strategic direction matters because AI does not land on a level playing field. The private sector and public sector face different incentives, different accountability structures and different definitions of “value”. In the private sector, value is often realised quickly as margin improvement, labour cost reduction, and market share expansion. In the public sector, value is often realised more slowly and less visibly: fewer administrative bottlenecks, faster case handling, better targeting, fewer errors, more time for frontline work, and higher service quality. Those are economically significant gains, but they do not always show up as immediate cash savings in a spreadsheet.
This is why the public versus private contrast can become a genuine dichotomy. In the private sector, AI-enhanced productivity can absolutely lead to job displacement, especially in white-collar occupations that were once assumed to be relatively protected: administrative support, routine legal drafting, some finance functions, customer service, back-office processing, and parts of junior professional work. The IMF has warned that AI’s displacement risks extend further up the wage distribution than previous automation waves, and that while productivity gains may raise incomes overall, labour income inequality can rise and capital income and wealth inequality tend to increase with AI adoption because returns accrue to those owning the assets and platforms. That is exactly the kind of dynamic Labour politicians should take seriously.
Recent evidence is mixed in the short term but consistent in direction. ONS data published in January 2026 reported that around 15% of trading businesses said they were using some form of AI technology, and among AI-using businesses a small but notable share reported reductions in employee numbers linked to AI over the previous three months, with some expecting further reductions. Meanwhile, an NBER working paper using large-scale executive survey evidence found widespread experimentation but limited realised gains so far, with 89% of C-suite respondents reporting no material productivity gains from generative AI in the prior three years, even as respondents expected average productivity gains of 1.4% and employment reductions of 0.7% over the next three years. In plain English: firms are still learning, but many executives expect a future in which output rises and headcount falls.
On top of this sits market concentration. AI is not only an application layer; it is a stack of compute, cloud, models, data and distribution. When those layers are concentrated, the gains concentrate too. Ofcom’s market study noted that AWS and Microsoft together held an estimated 70% to 80% of UK cloud infrastructure services revenues in 2022. The CMA later concluded that competition in the UK cloud market was not working as well as it could and recommended targeted interventions, including strategic market status investigations into the largest providers under the digital markets regime. For a Labour councillor, that is not an abstract competition-law concern. It goes directly to local economies, procurement costs, resilience, and whether productivity gains flow outward to communities or upward to a small number of firms.
The public sector can tell a different story, but only if we choose to make it so. The strongest Labour case for public sector AI is not “do more with less” in the old austerity sense. It is “do better with what we have, and redeploy human time toward the work only humans can do well.” That means using AI to take minutes, draft routine correspondence, summarise case files, triage standard queries, detect anomalies, and automate repetitive data tasks, so that staff can spend more time on safeguarding conversations, social work judgement, teaching, housing casework, neighbourhood problem solving, and complex support.
There is now real evidence that this is more than rhetoric. A government-led trial involving more than 20,000 civil servants, published in June 2025, found average self-reported time savings of 26 minutes per day from generative AI tools, framed as nearly two working weeks per person per year; the release also estimated this as the equivalent of giving 1,130 people a full working year back when scaled to the trial cohort. The same release referenced Alan Turing Institute research suggesting AI could support up to 41% of tasks across the public sector. If even a fraction of that is realised responsibly, the economic implication for public services is profound: not simply lower wage bills, but a potential increase in state capacity.
And state capacity is the key phrase. The UK’s Spending Review 2025 explicitly linked reform of the state to a £3.25 billion Transformation Fund, and described a “step change” in digital and AI investment across public services. It also cited ONS estimates that total public sector productivity in 2024 remained 4.6% below pre-pandemic levels, with healthcare productivity 9.6% lower. In other words, the political case for public-sector AI is arriving at exactly the moment public systems are under pressure and productivity recovery remains incomplete.
But this public-sector optimism must be tempered by realism. Parliament’s Public Accounts Committee has been blunt: government adoption is constrained by weak data quality, legacy systems and governance gaps. The PAC reported that 28% of central government systems met the government’s own definition of legacy in 2024. It also highlighted that AI deployment remained uneven (37% of 87 government bodies surveyed had deployed AI, while 70% were piloting or planning use). And on trust and transparency, it noted that only 33 algorithmic transparency records had been published by January 2025 on the relevant government website. So the public-sector opportunity is real, but so are the implementation risks: procurement capture, opacity, bias, poor data, and “pilotitis” without scaling.
This is where the socialist lens matters most. If AI rollout is governed by a narrow efficiency logic, the public sector could simply imitate the private sector’s labour-substitution path and call it modernisation. If it is governed by a social-value logic, the same technology could reduce low-value admin and expand high-value human service. That would not be a contradiction, it would be a choice.
The distributional consequences of private-sector AI make that choice even more urgent. If firms raise output with fewer white-collar workers, local tax bases can weaken even while profits and capital values rise. You may see more corporate productivity and more insecurity at the same time. This is precisely why proposals like Universal Basic Income (UBI) return to the debate whenever a major labour-saving technology wave arrives.
UBI has some virtues. Economically, it can simplify parts of a complex welfare system, reduce administrative frictions, and provide a reliable income floor in a labour market with more churn, intermittent work and self-employment. Socially, it can reduce stigma, increase security, improve bargaining power for workers, and acknowledge forms of contribution that the labour market often under-rewards, including unpaid care. In a period of AI-driven disruption, those are not trivial benefits. Public attitudes also show there is appetite for the idea, even amid fiscal scepticism: a 2025 GLA polling report found 50% of Londoners supported introducing UBI in London and 59% supported at least a trial, while 64% said the UK government could not afford it. While affordability is obviously an important issue, a more fundamental problem stems from the fact that UBI decouples reward from productive work, providing compensation for exclusion from productive life rather than a platform for active participation. Apart from possible adverse social consequences of encouraging idleness, making cash payments perfectly divorced from any return in terms of productivity would inevitably drive inflation, especially in asset prices, the most obvious being land and property: a substantial part of any UBI would quickly capitalise into higher property prices and increased rents resulting in yet another transfer of wealth from the state to landowners and lenders, i.e. banks.
That is why many on the centre-left see more promise in a job guarantee, especially in the face of AI-related disruption to white-collar employment. A job guarantee says: if the market no longer provides enough decent work at the speed society needs, the public realm should. It treats work not only as income, but as dignity, skill, routine, social connection and contribution. It can operate as an automatic stabiliser in downturns and as a bridge in structural transitions. And crucially, it allows the state to direct labour toward unmet social need.
The UK is not starting from scratch here. Past and recent schemes offer design lessons. The Future Jobs Fund created just over 105,000 jobs between October 2009 and March 2011 at a programme cost of about £680 million, and the DWP-commissioned impact work found net benefits to participants, employers and society, even though there was a net Exchequer cost under baseline assumptions. More recently, the Kickstart scheme was explicitly designed as 25 hours a week of subsidised employment for six months for 16 to 24 year olds on Universal Credit at risk of long-term unemployment. And the current Government’s wider youth policy direction has moved toward a guarantee model, with commitments to a universal guarantee of support and a guaranteed offer spanning education, training or work for younger people.
If we were to build a modern, AI-era job guarantee, the “socially beneficial work” test should be explicit and locally grounded. In practice, ten categories stand out. The first is adult social care support roles that expand time for personal contact and prevention; the second is SEND classroom and inclusion support; the third is neighbourhood environmental maintenance and retrofit assistance; the fourth is public health outreach and community wellbeing connectors; the fifth is housing repairs coordination and tenant support; the sixth is digital inclusion coaches helping residents access services safely; the seventh is youth mentoring and youth club support; the eighth is cultural and library access roles; the ninth is community energy and home-efficiency advice; and the tenth is flood resilience, parks and biodiversity stewardship work. These are all labour-intensive, socially valuable, and in many places under-provided.
The division of responsibilities matters. National government should fund the programme, set employment rights and wage standards, build the legal framework, and maintain a national data and evaluation system. It should also ensure the programme complements rather than replaces the benefit system, with clear interaction rules for Universal Credit, childcare support, disability benefits and housing support. Local government should be the principal place-based coordinator: identifying unmet need, matching participants to projects, ensuring “additionality” (so guaranteed jobs do not displace existing council or contracted staff), convening local employers and unions, and integrating transport, childcare and training. The voluntary sector should be a core delivery partner, but not a cheap labour sink. Charities and social enterprises are often best placed to host trusted, community-facing roles, but they need supervision funding, safeguarding support and multi-year stability if they are to participate well.
From a Labour councillor’s perspective, the most judicious UK implementation would be gradual, targeted and rights-based. Start with cohorts most exposed to disruption and scarring: young people not in education, employment or training, long-term unemployed adults, and workers displaced from routine clerical and administrative roles. Begin in local labour markets with persistent unemployment or high economic inactivity, and in sectors with demonstrable unmet social need. Pay at least the real Living Wage equivalent (with pension contributions and normal employment protections), combine work with accredited training, and build a clear pathway into permanent jobs, apprenticeships or further education. Put local democratic oversight around project selection. Require independent evaluation from day one. And make “human value added” a design principle: the purpose is not to create make-work, but to meet real social needs that markets undersupply.
The broad lesson is this. AI can deepen inequality in the private sector if its gains are captured as profits, rents and market concentration while labour absorbs the adjustment. It can, however, strengthen the public realm if we deploy it to augment rather than hollow out human service, and if we treat productivity gains as a way to expand capability, not just cut headcount. The technology is the same. The outcomes are not. They depend on power, ownership, regulation, fiscal choices and democratic intent.
That is the Labour argument in one sentence: we should not be anti-AI, but we must be pro-distribution, pro-public value and pro-dignity of work. If we get that balance right, AI could help build a state that is more responsive and more human, while giving working people real security through transition rather than leaving them to bear the costs alone.
---
Anthony Molloy is a councillor for Kilburn ward in the London Borough of Brent and Press Officer of the Labour Land Campaign
All blog posts represent the views of the author alone and not necessarily those of Mainstream.
