Contributors: Oliver Large, Ryan Wain

Foreword

Progressive politics is at its strongest when it knows not only what it wants to change, but why and how; when it offers a coherent theory of change and reform, not simply a list of policies.

Britain has a long and proud tradition of purposeful renewal in public services. From Attlee’s creation of the welfare state in the post-war period through Thatcher’s introduction of market discipline to New Labour’s focus on equity and performance, each generation responded to crisis with a clear organising principle for reform.

The Blair government had such a theory. Its approach to public-service reform sparked fierce debates, but it was not accidental or episodic. It rested on a consistent set of principles that were applied across departments and adapted to circumstances. Whether in education, employment, health, housing or transport, the central intent was to offer high standards for all, a degree of choice and meaningful citizen engagement, and social policies geared to personal fulfilment and future economic prosperity.

None of this was predicated on dogma or ideological ends in themselves, but it was a necessary pathway to achieving social justice, addressing the world of modernity and supporting people through rapid change.

A shared formula – choice and competition can drive quality – was the route to increasing equity. It pushed poor providers to raise performance and improve efficiency and responsiveness, animating reform efforts and underpinning a strategic plan that delivered the biggest transformation in outcomes in a generation.

Today, that coherence has been lost. Progressive politics is marked instead by fragmentation: different sectors pursuing different reform instincts, often at odds with each other. There is no lack of commitment to public services, but in the absence of a shared theory of reform, rapid, visible, durable progress is absent too.

Renewing progressive politics demands clarity and coherence. But the conversation must go beyond relitigating old debates. The world has changed. The tools available to government have changed, too. Where choice and competition once drove improvement, now data, digital technology and artificial intelligence offer the greatest potential for reform. Used well and in the service of a common goal, they can enable a new model of public services, grounded in modern methods of public engagement and, where possible, citizens’ participation in decision-making. This new model can be transformative for the state, delivering services that are personalised, not standardised, preventative, not reactive, and always on rather than bound by labour alone.

The parallel is important. Just as choice and competition drove higher quality and better performance, data-driven and AI-enabled systems can support the delivery of better outcomes. They allow the state to identify need earlier, tailor support and intervene before problems become crises so it can finally achieve, directly and at scale, what earlier reforms pursued indirectly. Where competition disciplined poor performance, AI-enabled systems can analyse and predict performance in real time, allowing providers to course-correct quickly. Where choice depended on citizens having the information to navigate complex systems, technology can now adapt systems around citizens by default. This transformation requires deep and radical reform, not just adding technology to existing struggling services.

Without a clear view of how this can be accomplished, citizens’ expectations will never be met and their trust will be lost, with progress stalled and fairness foregone. Governments will end up paying a very high price – financially and socially – for failure.

And, in the end, effective delivery of services depends not just on systems and structures but on people who care. Someone’s brother, sister or mother is responsible for caring enough to challenge bureaucracy and to go the extra mile. That is why giving people the skills to meet the challenges of tomorrow is vital – as is ownership of the task of improvement, so that everyone can ask themselves: “Did I do a good job for someone else today?”

A renewed theory of reform is essential. But it must be a modern one. This paper makes the case for just such a modern theory of progressive reform – one that rises to the challenges of our time, matches today’s possibilities to enduring political values and can guide government action across the whole of the state.

Lord Reid of Cardowan and Lord Blunkett

Executive Summary

Despite record spending, people in Britain – as in other developed economies – are increasingly frustrated with public services. Long waits, overstretched staff and worsening outcomes have become the norm. Fiscal pressure is mounting, and trust in government is plummeting. It may be tempting to blame shocks dating back to the 2008 financial crisis, but the problem is not one of funding. It is one of vision. A new agenda of reform is needed to build personalised, preventative, always-on services.

Previous reform efforts, however different to each other, had a common foundation of political leadership: a deep analysis of what had gone wrong, clear answers on how to correct it and the will to act on those insights. Clement Attlee’s founding of a welfare state based on fairness through universal access, Margaret Thatcher’s pursuit of efficiency through market discipline, Tony Blair’s focus on equity and performance were each based on a well-articulated theory of change, giving coherence to the reforms they drove.

The conditions of today demand a new reforming agenda and zeal to match the best efforts of previous generations – perhaps even exceed them in boldness. But not since New Labour’s modernisation in the 2000s has a government pursued a coherent theory of change.

In the absence of a core vision, piecemeal promises, headline hiring pledges, pilots and patches can, at best, tackle symptoms, but they do not fix the underlying structural flaws. Government after government has settled for managed decline instead of disruptive but necessary renewal.

This is not a matter of abstract theory or lofty philosophy. Public services that work matter deeply to people – their prospects, their safety, their quality of life. Without ambition and coherence, improvements will be marginal, pressures will continue to build and disillusionment will grow. Without reform, governments face spiralling costs, failing systems and collapsing trust. With it, Britain can pioneer, once again, a new model of governance which restores legitimacy and proves that a state that delivers is an unstoppable force for good.

Only a reform agenda that radically changes the government’s ability to deliver, embracing the transformative potential of technology to overhaul the model itself, can meet the demands of this moment. At the Tony Blair Institute for Global Change (TBI), we call this agenda the Reimagined State.

Today, vast volumes of data coupled with artificial-intelligence capabilities make a fundamentally different kind of governance possible. Capacity can expand without necessarily expanding labour. Services can include an “always-on” layer, available around the clock. Personalisation by default and at scale becomes possible, replacing standardised provision with services that are tailored to each individual and learn continuously from every interaction. Providers can spot risks before they escalate, replacing today’s remedial model of costly intervention at the point of crisis with proactive prevention. In place of rigid command from the centre, professionals can be empowered at the front line with relevant real-time data and intelligent tools, restoring discretion and autonomy while unlocking innovation.

This reform agenda forms the blueprint of a data-driven state: always on, personalised, preventative. It uses intelligence rather than labour alone to deliver results. It learns in real time, adapts as conditions change and delivers better outcomes at lower cost. It delivers fiscal sustainability with higher productivity, expands capacity without expanding labour costs, shifts resources upstream to prevention and embeds continuous improvement into daily delivery.

To make this renewal real, public services need a new operating model built on three foundations. Universal digital ID can seamlessly and securely link every citizen to their entitlements and give them control over their collated data, embedding transparency and consent. Modular public-service platforms, such as a national health platform evolved from the NHS App or a national education platform for truly personalised learning, can give frontline staff smarter tools to work with and allow citizens to navigate services as one coherent system. And a real-time system-intelligence layer can transform millions of interactions into actionable insights for early intervention, continuous improvement and a return to genuine accountability.

Together, these three layers would create the architecture for a new kind of public services. But new infrastructure must be matched by new forms of governance.

In the AI era, government no longer needs to rely on command-and-control management or retrospective inspections. Instead, it must govern by “signal and steer”, using live data to set thresholds, maintain fairness and support continuous improvement. Funding too must evolve: locked institutional budgets should give way to modular, outcome-based allocations that flow to tools and interventions shown to work. The workforce must be re-empowered as innovators, freed from compliance for compliance’s sake, given the skills and autonomy to drive and own improvement through new methods at the front line, and incentivised to contribute to the system’s collective learning loop. Citizens themselves should not be passive recipients of services but co-creators, shaping their service journeys, giving real-time feedback and controlling their data.

This paper sets out how to achieve this transformative reform with a playbook built on four principles:

Use AI as a multiplier, to make services scalable and always accessible without simply adding labour. 1.

Rewire services around the citizen, through digital ID and data consent. 1.

Govern with foresight, using real-time intelligence to act preventatively rather than remedially. 1.

Foster public innovation, enabling the workforce to drive continuous improvement from the bottom up.

This is not about bolting new technology onto crumbling foundations. It is about replacing those foundations altogether, creating public services that are always on, personalised by default, preventative rather than reactive and powered by real-time data. Britain was among the pioneers of the 20th-century public-services model defined by universal access. The 21st-century model will be defined by mass customisation – an adaptive state that learns, protects and improves every day.

Here, too, our nation can show the way. This new blueprint can become core to a newly resurgent Britain – a country restored in capability, confident in its future and able to show that effective, empowering, citizen-centred government is possible in the age of AI. Reforming public services is not one policy area among many. It is the foundation of a renewed social contract and the clearest test of whether Britain can escape decline and build a better future.

Chapter 1A New Reform Agenda Is Needed

Public services are running on fumes. Across health, education, welfare and justice, citizens face the same frustrations: long waits, rigid rules, overstretched staff and outcomes that fall short despite record spending. People are paying more yet receiving less, even as their expectations are raised by seamless private-sector experiences. The result is a widening gap between what citizens expect and what government delivers – eroding trust in the state itself.

The root problem is structural. Public services are trapped in an outdated operating model that is labour-intensive, standardised, centralised and reactive. Scaling capacity means scaling staff, standardisation suppresses diverse needs and support typically arrives only after crisis has been reached. Investment has not translated into better outcomes because the model itself is flawed.

This mounting structural strain should have been obvious to anyone paying attention and has been evident in the repeated failures across public services – from prisons and courts cycling through emergency measures to the NHS’s annual winter breakdowns and the recurrent capacity collapses in councils, social care and the asylum system. But recent governments have lacked a coherent agenda for reform.

This absence of direction is striking when set against Britain’s longstanding tradition of renewal. Every generation of leaders has faced its own period of public-service failure and responded with a distinct governing principle.

Clement Attlee’s government confronted the social and economic devastation caused by the second world war and the moral bankruptcy of interwar neglect. It made universality the organising logic of the state, building provision at scale to guarantee a basic level of security for all citizens. Embracing the idea that the government’s overarching objective must be to beat back the “five giants” of the Beveridge Report – want, disease, ignorance, squalor and idleness – it established the National Health Service, expanded education, introduced social insurance and embarked on a vast programme of public housing. The guiding conviction was that no citizen should fall below a minimum standard of health, income, shelter or education, and that this kind of security was essential to a stable and cohesive society. Attlee’s government also set the parameters of the public-service operating model according to the abilities of the state at the time: a labour-intensive bureaucracy coordinating provision through central planning and mass employment.

Margaret Thatcher’s government confronted a different kind of crisis: the stagnation and inefficiency of the 1970s, when inflation, industrial unrest and fiscal strain had eroded faith in the state’s ability to deliver. To the logic of universality, she added competition and discipline, arguing that public services should mirror the incentives and accountability of the market. Through privatisation, deregulation and the weakening of producer power, her government sought to restore efficiency and individual responsibility. Its defining characteristic was a belief that the state should do less but do it better, creating space for enterprise, innovation and self-reliance to flourish.

Tony Blair’s government inherited creaking public services and a crisis of confidence in state performance after years of underinvestment. It sought renewal through modernisation – preserving the public ethos of Attlee’s post-war model while harnessing the mechanisms of market reforms. Competition was repurposed to drive standards rather than to shrink the state, and choice was framed as a route to fairness. Its defining characteristic was a belief that public services could be simultaneously universal in access and personal in experience, using targets, performance management and partnership to reconcile equity with efficiency.

What unites these different visions is that each brought a new level of coherence and direction to necessary reform. They were not incremental management tweaks but organising ideas rooted in a clear diagnosis of what had gone wrong and a leader’s conviction about how to put it right. Each vision reflected its time and its distinct challenges.

Today’s challenge may be the biggest of them all. The old operating model is breaking down under the weight of demand, complexity and patterns of need it was never built to meet. Just as the Victorians reconfigured the machinery of government in response to the Industrial Revolution, government must now adopt a long-term agenda to redesign how it operates for the digital and AI era – a framework TBI describes as the Reimagined State.

Transforming the state in this way requires a fundamentally different approach to reform: one that uses digital infrastructure, real-time data and artificial intelligence to overcome the structural limits of the old operating model. These tools make an entirely different model of service delivery possible. Instead of being bound by labour constraints, services can be always on. Instead of offering one-size-fits-all provision, they can be personalised by default. Instead of waiting for crises, they can act proactively and preventatively. Instead of relying on centralised decision-making, frontline workers can be empowered with live data and intelligent tools.

In this paper, we set out a reform agenda for the AI era: a new operating model and governing framework that can turn public services into self-improving systems – learning continuously, adapting in real time and delivering better outcomes with greater efficiency. Building on previous TBI work – including Governing in the Age of AI: A New Model to Transform the State and our recommendations of AI-era reform across the NHS, DWP, local government and the Spending Review – this paper consolidates that thinking into a single, system-wide reform playbook fit for the AI era.[_],[_],[_],[_]

Chapter 2The Operating Model of Public Services Is Broken

By any measure, the performance of public services in the UK has deteriorated over the past decade. According to the Institute for Government’s Public Services Performance Tracker, almost every service – with the exception of schools – is performing worse than in 2010.[_]

For millions, public services now feel more like a bottleneck than a safety net, as they find themselves stuck in a queue that barely moves – whether that is waiting hours in Accident and Emergency, months for treatment or years for justice (see Figure 1).

Figure 1

Backlogs and long waiting times are entrenched across public services

Public service	Size of backlog/typical wait	Target/pre-Covid baseline
NHS elective (planned) care (England)	7.39 million open “referral-to-treatment” pathways[_] Median wait 13.4 weeks; 180,000 waits over a year[_]	4.6 million pre-pandemic Median wait 8.0 weeks in September 2019[_]
A&E (England)	74.1 per cent of patients seen, treated or admitted within four hours[_]	Target: 95 per cent seen within four hours
Cancer-treatment start (62-day standard) (England)	67.9 per cent treated within 62 days of urgent referral[_]	Target: 85 per cent of patients commence treatment within 62 days of being referred[_]
Speech and language therapy (England)	68,185 children and young people waiting for speech and language therapy services[_]	Good practice emphasises early assessment and timely intervention
Adult social-care access (England)	418,029 people waiting for an assessment, care or direct payments to begin, or a review of their care plan[_]	List peaked at 542,000 (April 2022) and has since slightly fallen
Crown Court cases (England and Wales)	76,957 outstanding criminal cases[_]	Backlog was approximately 39,000 in 2018; government target of 53,000 by March 2025[_]
Personal Independence Payments (UK)[_]	826,609 people waiting for a PIP decision[_]	317,216 people waiting for a PIP decision in April 2019[_]

The queues, delays and frustrations that millions experience are symptoms of something deeper: an operating model that is no longer fit for purpose. By operating model we mean the structures and processes that govern how services are funded, organised and delivered. The model determines who makes decisions, how staff spend their time and how resources are turned into outcomes. In its current form it leads to backlogs, fails to deliver productivity gains, drains public finances and erodes trust.

First, the model struggles to convert more resources into better outcomes. Between 2019 and 2024, inputs to UK public services rose by almost 25 per cent but output increased only 14 per cent.[_] NHS productivity illustrates this: since 2013–14, resource use has grown 44 per cent, while activity rose only 35 per cent.[_] Even with more nurses and consultants since 2019, elective admissions increased less than 10 per cent. Waiting lists are longer and targets are consistently missed. Overall, public-sector productivity is 3 per cent below pre-pandemic levels.[_]

Second, it consumes ever greater sums of money. The tax burden is near its post-war peak, yet public-service deficits persist. NHS spending has grown 3.2 per cent annually in real terms since 2014–15. In 2023–24, NHS systems recorded a £1.4 billion deficit despite Treasury support. More money is being spent, with less return.

Figure 2

Taxes as a share of GDP are nearing the post-war peak

Source: OBR Economic and Fiscal Outlook November 2025

Third, it erodes public trust. Satisfaction levels are at historic lows: only 21 per cent are satisfied with the NHS, 40 per cent with GPs and 23 per cent with the police.[_],[_] People feel they pay more and receive less, creating a widening credibility gap that strains the social contract.

Covid-19 did not cause this failure; it exposed it. Services have struggled to recover, and many remain worse than pre-pandemic. The model is simply outdated.

Services Are Running on an Outdated Model

The operating model of public services was built first and foremost to deliver a historic breakthrough: universal access. In the mid-20th century, governments transformed provision that had been patchy and voluntary into a system that was national, reliable and fair. Every child was entitled to schooling through a common curriculum, every patient to health care, free at the point of use. Mass provision became the foundation of social equity.

But this achievement was shaped by the constraints of its time. The operating model was built around:

Labour intensity: Public services relied on large, affordable workforces delivering face-to-face interactions – consultations, lessons, home visits. Scaling access meant hiring more people.

Standardisation: Services were designed for the “average” citizen. In education, that meant a uniform curriculum in age-segmented classrooms; in health, standard pathways rationed by professional gatekeeping. Individual needs were often overlooked, leaving many to experience the system as unresponsive or even exclusionary.

Reactive demand management: Services intervened only after problems were visible. The NHS centred on hospitals treating illness rather than preventing it. Over time, this reactive mindset became hard-wired, with funding channelled into acute crises rather than prevention.

Slow information flows and weak feedback loops: With paper records and delayed data, the centre lacked real-time visibility of service performance. Alignment across a large system could only be achieved by centralising control – pulling decisions upwards and suppressing local autonomy. The result was a model that prioritised uniformity over responsiveness.

Productivity in labour-intensive sectors cannot rise easily: a nurse can still only treat one patient at a time and a teacher can only attend to so many pupils. As wages rise to remain competitive with more productive sectors, service delivery costs increase without matching output – a dynamic known as Baumol’s cost disease.[_]

This is why the record spending and staff growth today have not translated into better outcomes. A labour-intensive model drives costs up while productivity flatlines, leaving citizens paying more for less – and losing faith in the system itself. Recent public-sector pay crises underscore this reality. The government has faced immediate pressure to settle disputes simply to keep essential services operational.

The fiscal pressures facing public services are therefore structural, not cyclical. Baumol’s cost disease means that labour-intensive services inevitably become more expensive over time. As The Productivity Institute highlights, governments face a stark choice under this model: either spend ever more simply to maintain existing levels of provision, or rely on periodic efficiency drives that temporarily contain costs but ultimately erode capability, service quality and workforce morale.[_] Escaping this trap requires changing the operating logic altogether, rather than pushing the existing one harder.

Figure 3

How the current operating model impacts on public services

Source: TBI analysis

The Institute for Government identifies chronic underinvestment, rising and more complex demand, and workforce attrition as common pressures acting across services today.[_] Our argument is that these are not external shocks, but the predictable consequences of a system still trapped in the constraints of another age.

Underinvestment stems from the operating model’s failure to use resources efficiently: because it is labour-intensive and geared towards meeting immediate pressures, funding is consumed by day-to-day firefighting. Rising complexity of demand clashes with a model built for mass, standardised provision, which struggles to respond to diverse needs or provide personalised support. And the workforce, still treated as the primary engine of delivery, is stretched and burned out; it is expected to do more with outdated tools, rigid processes and growing caseloads. What look like external pressures are in fact symptoms of an outdated system.

Any serious public-service reform agenda must begin from this fact, because without replacing the model itself, no amount of additional funding, better management or piecemeal reform will reverse the decline.

The Private Sector Is Redefining Service Expectations

While public services remain locked in labour-intensive, standardised models, private-sector services are breaking free from these constraints. By embedding AI technologies into their operations, they have unlocked new capacity, raised productivity and transformed the customer experience. The lesson is clear: services are not condemned to stagnation. With the right operating model, technology can be harnessed to do more, better, for less.

Fast-food chains illustrate the shift. After decades of flat productivity, the sector saw sharp gains after 2020: sales per employee rose around 15 per cent, driven not by cutting staff but by augmenting them.[_] Mobile apps, online ordering and AI-driven demand forecasting streamlined transactions, reduced queues and enabled staff to serve more customers in the same time.

Other sectors illustrate how AI has built intelligence into everyday services. Navigation apps such as Google Maps and Waze use machine learning to turn billions of individual data points into a live map of the system, so that each journey updates the picture for everyone else. Streaming platforms such as Spotify or Netflix rely on recommendation algorithms trained on vast interaction data sets to personalise content, so every user receives tailored suggestions while the system as a whole keeps improving.

Amazon and Uber use real-time optimisation algorithms to dynamically allocate resources – whether that means routing stock through warehouses or matching drivers to riders in peak hours. Tesla vehicles continuously generate telemetry that feeds into AI models for autonomous driving, so the fleet learns collectively from every mile driven. Consumer-health ecosystems such as Apple Health or Fitbit depend on AI signal extraction: raw biometric data is filtered and interpreted into alerts or insights, shared in a privacy-preserving way that still adds value across populations.

These technologies not only lift output per worker but also create a more personalised, responsive service environment.

As a result, citizens’ expectations have shifted. People now take for granted speed, convenience and personalisation – same-day delivery, instant banking alerts and tailored streaming recommendations, for example. Over 70 per cent expect personalised interactions and feel frustration when they do not get them.[_] In that context, waiting weeks for a benefit decision or hours in an A&E queue feels not just inefficient but unacceptable.

The technologies that have transformed private-sector services are readily available to the public sector. AI, digital platforms and real-time data pipelines could help frontline staff work more effectively, personalise support, anticipate problems before they become crises and learn continuously from every interaction. The barrier is not capability but an operating model designed for another era – labour-intensive, standardised and reactive. Unless that model is reformed, public services will continue to fall further behind the standards that citizens already take for granted in the rest of their lives.

There Is No Public-Service Reform Playbook

Despite the scale of today’s challenges, Britain has no coherent playbook for reform. Health defaults to centralised performance management; targets and league tables meant to jolt the NHS into improvement sit alongside ambitions set out in the 10 Year Health Plan for England to shift care from hospitals to communities, move services from analogue to digital and reorient the system from sickness to prevention. Education, by contrast, is shaped by instinctive conservatism, with government wary of disrupting a system that has raised attainment but is now sustained by exhausted staff and has lacked a defined path to improvement since academy status became near universal. These contrasts expose divergent theories of change and the absence of a unifying philosophy of reform.

The last time Britain had a unifying framework for public-service reform was under New Labour’s modernisation agenda, which took shape in the Blair government’s second term. Faced with persistent underperformance in public services, the government developed a reform model built around four mutually reinforcing pillars:

Performance management: This meant clear national standards, measurable targets, rigorous inspection and corrective action for failing providers, underpinned by the Prime Minister’s Delivery Unit and Public Service Agreements. High-performing organisations earned greater autonomy, while corrective action was taken for failing providers. 1.

User choice: Families and patients were given more say in education and health care, challenging the producer-led model of the post-war settlement. 1.

Competition: Funding followed users, creating quasi-markets in health and education to incentivise efficiency and innovation. 1.

Diversity of supply: Charities, social enterprises and private providers were brought in to expand capacity of state monopolies.

This model could be applied across sectors to raise standards systematically. It worked: outcomes improved and satisfaction rose as delivery became the focus of government.

That coherence, however, was lost after the 2008 financial crisis. Fiscal retrenchment displaced reform zeal, and the architecture that had bound the reform agenda together was dismantled. The Prime Minister’s Delivery Unit was downgraded and eventually replaced by successor units, and Public Service Agreements were scrapped. Performance management became more fragmented and reactive, with successive governments cycling through delivery units, dashboard metrics and mission statements – none of which replicated the grip or coherence of the Blair-era model.

Other elements of the playbook survived in places, but no longer as part of a unified framework. In health, the Health and Social Care Act 2012 entrenched provider competition, only for it to be reversed a decade later in favour of integration. In education, the logic of earned autonomy was lost as academy status became near-universal. By the 2020s, the way in which improvement was sought across services had become incoherent: different sectors were governed by different logics, with no cross-government alignment.

Yet reviving the playbook wholesale would not work. Its core mechanisms were designed to raise performance inside the legacy operating model of the late 20th century. They could make that model work better for a time, but they could not change the model itself. The task today is not to restore the old architecture but to replace it.

In the absence of such a replacement, government has reached for alternatives. Programmes such as the £100 million Test, Learn and Grow initiative are deploying multidisciplinary innovation squads into communities to co-design services with councils, frontline staff and residents.[_] The government’s Blueprint for Modern Digital Government sets out a six-point plan to join up services, strengthen data infrastructure and accelerate AI adoption. Meanwhile, the ten-year NHS plan acknowledges the broken operating model and aims to rewire the health service around prevention, digital tools and patient empowerment, while bringing back performance-management elements from the New Labour era.

The test-and-learn ethos brings welcome agility, but its strength in piloting small-scale innovations becomes a weakness when there are no pathways to scale, diffuse lessons or embed change in the system’s operating logic. The Blueprint for Modern Digital Government recognises the urgency of tackling the state’s digital deficit, but it is fundamentally a technology strategy rather than a reform playbook: it outlines how to modernise infrastructure without reimagining the institutional architecture of public-service delivery.

As TBI has argued, the NHS plan has the right vision for transforming health care – a genuine attempt to disrupt from within through digital tools and prevention – but it has been criticised for lacking a credible delivery plan.[_],[_] Its ambition is confined to health.

Each alternative points in the right direction, but none offers a coherent vision for how to update the outdated operating model across public services so that they can deliver consistently, equitably and sustainably at the scale modern societies require.

Britain is stuck between nostalgia for past reform models and substitutes that cannot scale. The challenge of our time is not to squeeze marginal gains out of a legacy model but to replace the operating logic of the state itself. The task is to build a genuine playbook for the AI era – one that rewires public services to be adaptive, coherent and capable of continuous improvement across the whole system.

Chapter 3The Promise of AI for Public-Service Reform

The structural failures of the UK’s current public-service model are no longer in doubt. The question is not whether to reform, but how. Asking the system to do more, with less, inside an architecture that resists scale, learning and personalisation will not work.

AI, together with wider digital infrastructure, offers a way to break free from the constraints that have so far shaped public services.

Crucially, this is not just about automating existing processes or bolting AI onto outdated workflows. Using AI to speed up paperwork in the NHS or add a chatbot to Universal Credit still leaves the old operating logic intact. Instead, the true potential lies in using AI to replace the legacy operating model altogether – shifting from optimisation of the old to the creation of something fundamentally new.

AI enables a different logic of delivery: one that is scalable without proportional labour, personalised by default, proactive rather than reactive, and empowering for both citizens and professionals. In place of the legacy model, AI makes it possible to design a new operating model through these four core transformations.

Always on: From Constrained by Labour to Scaling With Demand

In the current model of public-service delivery, professional time is the ultimate bottleneck. A GP can only see a fixed number of patients in a day. Scale has therefore meant hiring more staff, an approach which is increasingly unsustainable under fiscal pressure.

AI enables a different path: augmenting professionals to both expand capacity and provide a layer of round-the-clock access. This allows existing staff to do more without increasing their workload, raising productivity, and enables service users to receive some forms of support at a time that works for them.

AI systems can help match resources to demand in near real time, as intelligent triage systems prioritise cases so that scarce professional time is directed where it has the greatest impact. Automated case-handling and documentation free staff from the administrative grind that often consumes hours of their day. Smart scheduling matches demand to availability in real time, ensuring that appointments or visits are allocated with far greater precision.

The result is a shift from labour scarcity to cognitive abundance. AI acts as a digital multiplier, taking on the repetitive, the predictable and the informational, so that professionals can focus on what only humans can do: exercising judgement, building trust, offering care and making ethical decisions. A GP supported by AI-powered triage can devote more attention to the most clinically complex patients. A probation officer aided by predictive analytics can monitor larger caseloads without losing sight of emerging risks. A teacher working alongside an AI tutor can deliver personalised reinforcement to every learner, without adding to their own workload.

The impact of AI extends beyond augmenting professionals to enhancing services themselves. Once intelligence is embedded directly into delivery systems, services can operate continuously rather than intermittently – becoming, in effect, “always on”. Citizens would no longer need to wait for office hours or appointments, or navigate queues to get help. AI-enabled systems, including in the near future agentic systems, can answer questions, guide users through applications, offer tailored learning support or provide initial health advice around the clock.[_] These systems supplement professional time rather than compete with it, extending the reach of services in ways that were once unthinkable. Embedding AI would not mean that every aspect of provision is accessible around the clock, but would add a layer of access that takes the strain off the system and provides immediate support where it is needed.

This shift is already under way. In health care, autonomous physiotherapy systems approved for clinical use are providing continuous musculoskeletal assessment, treatment and monitoring – operating day and night, without the scheduling limits of human staff.[_] By sustaining care between appointments and automatically updating recovery plans as new data arrive, these systems keep patients progressing even when professionals are offline. In this sense, they exemplify how AI can make a service “always on”: delivering guidance, feedback and adjustments in real time while clinicians focus on complex or high-risk cases.

This kind of early deployment shows how intelligence can be built into the service process so that capacity expands continuously rather than episodically – not by adding more workers, but by making professional expertise persistently available. Crucially, this is not just about optimising today’s workflows. By relieving the workforce bottleneck, AI unlocks new service models that were previously impossible within the cost and staffing constraints of the 20th-century architecture: a personalised tutor for every child, proactive welfare triage for every household, early intervention for every at-risk individual. In effect, AI allows services to scale in two directions at once – by boosting what professionals can do, and by making more possible without them.

Breaking the labour bottleneck is the foundation of fiscal sustainability. Once services can scale without proportional labour, they can deliver rising output without rising cost, addressing the structural cost disease that has long haunted public provision.

Personalised: From Standardisation to Customisation by Default

The opportunity goes beyond output measures, because quality can be improved at scale, too. Historically, public services were designed for the “average user” and delivered through standardised, one-size-fits-all models. This was not a failure of imagination but a function of necessity: industrial-era systems had to deliver for millions with limited professional capacity and slow information flows. Standardisation was the only way to achieve scale.

But this logic has costs. Services often over-deliver to some, under-serve others and intervene too late or in the wrong way. Citizens experience the system as blunt, inflexible and unresponsive to their actual circumstances. Professionals are forced to shoehorn diverse needs into rigid categories, knowing that the fit is poor.

AI makes personalisation the default mode of service delivery. With real-time data and adaptive algorithms, services can adjust intensity, timing and content to each individual. A student struggling with a specific concept can receive targeted reinforcement, while another who has mastered it can move ahead. A patient can follow a care pathway tailored not just to their diagnosis but to their lifestyle, risk profile and comorbidities. A jobseeker can receive proactive nudges and training opportunities customised to their needs and aligned with their local job market. Vulnerable citizens can be flagged through integrated service interactions, triggering tailored outreach before problems escalate.

This shift builds personalisation into the system at scale. Instead of treating everyone as average, services flex to diversity in real time. The benefits are twofold: citizens receive support that actually matches their needs, and resources are deployed with far greater precision, reducing waste and increasing effectiveness.

Just as importantly, personalisation fuels system-wide learning. Every interaction generates data: which interventions helped which types of students, which treatment pathways worked best for which patients or which employment-support strategies succeeded in which local labour markets. Instead of relying on infrequent evaluations or national averages, the system can continuously observe what is effective in real time.

Aggregate and anonymised data in these feedback loops allow frontline professionals, service providers and policymakers to understand what works, for who and under what conditions. They make it possible to refine interventions dynamically. A welfare programme can adapt its approach based on the characteristics of households it helps most. In education, teaching sequences can be improved by tracking which explanations unlock progress for different learners. A health service can adjust protocols by analysing how different patient profiles respond to treatment. Importantly, with the time and resources freed up by AI, public servants now have the space to reflect on, absorb and put into practice these lessons learned.

In this way, personalisation becomes not just a feature of delivery but the engine of continuous improvement. Services learn from every case, policy becomes evidence-led by default and the system as a whole gets smarter the more it operates – driving more targeted interventions, more efficient use of resources and better outcomes across the board. The current doom loop is replaced by a positive self-reinforcing cycle.

Preventative: From Reactive to Proactive

The existing model of public services is structurally reactive. Support typically arrives only once a problem has fully materialised – a patient is admitted to hospital, a pupil is already falling behind, a family is out of work or an offender is already back in custody. This orientation towards late intervention is not a matter of neglect but of design: with limited information and scarce resources, services are forced to respond to visible need rather than anticipate risk. By the time the system acts, harm has occurred, costs have escalated and options have narrowed.

This built-in reactivity is the main reason why, despite long-standing consensus on the value of prevention, public services have failed to move decisively in that direction. But AI capabilities, adopted at scale, enable a decisive move from reactive firefighting to proactive delivery. With predictive analytics, pattern recognition and cross-system data, public services can identify risks earlier and act before crises escalate. A child showing early signs of disengagement or rising absence in school can receive timely support to re-engage rather than waiting for exclusion. Health data from wearable devices can alert clinicians to deteriorating conditions before an emergency admission is needed. A welfare system can be on the lookout for households at risk of debt spirals and provide guidance or financial support before they fall into a crisis that is ultimately more expensive to resolve.

This proactive orientation has financial as well as social benefits. Public spending today is consumed by crisis response – hospitals operate at full capacity as preventable diseases escalate into acute conditions, schools intervene only once failure is visible and welfare payments flow only once people are already out of work. By redirecting resources upstream, AI enables an investment model that reduces long-term costs while improving outcomes. Forecasting demand, reallocating resources dynamically and addressing risks early prevents expensive acute interventions, aligning capacity with need and freeing up resources for prevention.

In health alone, modelling by TBI shows the scale of what prevention could deliver.[_] Cutting the incidence of six major diseases that keep people out of work – cancer, cardiovascular disease, chronic respiratory illness, diabetes, and mental-health and musculoskeletal disorders – by just 20 per cent would raise GDP by £20 billion within five years and £26 billion within ten. The resulting fiscal dividend – through higher tax receipts and lower benefit payments – would reach £10 billion a year by 2030 and £13 billion by 2035. When it comes to obesity, TBI modelling suggests that broader access to anti-obesity medications could generate cumulative fiscal benefits of around £52 billion by 2050.[_]

Proactivity is inseparable from real-time accountability. Services can only prevent effectively if they can see problems as they emerge. AI makes this possible by continuously tracking interactions across the system. Delays in case processing, spikes in absenteeism or uneven access to health care can all be detected early and seen in the context of benchmarking against similar providers, allowing corrective action upstream without imposing unrealistic expectations. For leaders, live dashboards and AI-powered suggestions provide system-wide grip and evidence-based advice; for frontline staff, they offer immediate feedback that supports learning and adaptation.

This shift is profound: accountability ceases to be retrospective and compliance-driven – based on audits or inspections after harm has occurred – and becomes a live, operational function of the system. In effect, services gain the capacity to see around corners and adjust course in real time.

The benefits are twofold. For citizens, problems are anticipated rather than endured, with support arriving earlier, more appropriately and often less intrusively. For the system, resources are used more efficiently, with costly acute interventions replaced by targeted, preventative action. AI therefore makes possible what the legacy operating model could not: a state that is not only more responsive, but proactively protective.

Empowering the Front Line: From Command and Control to Informed Autonomy

The original operating model pulled decisions upwards because the centre lacked visibility. Slow data flows and paper records meant one way to ensure alignment was to reduce autonomy on the front line. Teachers, doctors and caseworkers were expected to follow rigid protocols, while central administrators issued targets and dire