The new government faces an urgent challenge: revitalising the UK’s crumbling public services without major increases in public spending. Technological change could be the solution, but digital government initiatives aimed at this challenge have tended to disappoint. The latest generation of ‘data-intensive’ technologies, including data science and AI, offer new possibilities for ‘progressive public services,’ which play a key role in improving people’s lives.
The best opportunity for AI and data science is to increase public sector productivity. There are two ways that AI and data science might do that. First, AI can be used to automate bureaucratic ‘micro-transactions,’ such as triaging documents, checking photos, and drawing out salient points from text. Of the one billion citizen-facing transactions that the UK government makes every year, around 120 million are highly automatable with AI. Automating micro-transactions would not only speed up processes, but also make public sector professionals’ roles more satisfying by reducing their administrative load.
Second, AI can be used to optimise budget decisions, guiding policymakers on how they can get the most out of the £1.2 trillion of public funds that the government spends in a year. Data science models can capture the interdependencies between policy domains and understand to what extent investing in one area, such as health, boosts outcomes in another sector, such as education.
If increasing productivity is the greatest hope for AI, growing inequity is the greatest fear. Technological change can reinforce inequities of power and reward if the benefits are narrowly shared, and the harms fall disproportionately on the most vulnerable. For example, while digital chatbots could revolutionise the way that people interact with public services, over a million children and their families do not have adequate access to a device or connectivity and home to use them. Furthermore, AI can introduce biases into bureaucratic decision-making and reinforce existing inequities. AI systems are designed, developed, and deployed by humans. So humans’ biases and discriminatory practices can – and do – creep into all stages of AI systems’ lifecycle.
The public sector needs to chart a careful path through the complex trade-off between productivity and equity. Technological capacity building is a must if the public sector is to navigate this path successfully. Over the last 50 years, government has become heavily dependent on global computer services providers, so-called ‘Systems Integrators,’ who custom-build systems under long-term, high-value contracts. These contracts have progressively stripped technological expertise out of government agencies. While the systems integrators retain influence over the so-called ‘legacy systems’, government has also become dependent on the huge technology platforms that originated in Silicon Valley (SV) for cloud provision and AI technologies such as large language models (LLMs).
In the UK, we see concrete steps to amass expertise inside government. The Central Digital and Data Office (CDDO) leads on ‘digital, data and technology’ while the Government Digital Service (GDS) manages government websites and information provision. The Incubator for Artificial Intelligence (i.AI) was set up more recently to develop ‘products’ for government, such as data-sharing solutions and AI-powered co-pilot tools. The AI Safety Institute (AISI) was created after the UK AI Safety Summit to work on safety testing and evaluation of the latest ‘frontier’ models. All these initiatives are doing important and innovative public sector work, and provide a needed counterweight to the powerful systems integrators and SV platforms. But they have been scattered across the centre of government in the Department of Science and Innovation (DSIT), the Cabinet Office, and No. 10.
We propose a roadmap for how to develop the expertise and capacity needed for a more productive and equitable public sector, and pinpoint the organisational changes necessary to develop progressive, technologically enhanced public services:
1. Bring central agencies together with shared goals, joint working, and radiation outwards from the centre to the front line of public service delivery. Excitingly, this first recommendation is part of the new government’s plans. The DSIT Secretary of State, Peter Kyle, has announced a “digital centre for government” that includes CDDO, GDS and i.AI. This Digital Government Centre (pending its naming) should fly under the banner of ‘public services innovation.’ It should act as a centre of expertise to support cultural change and be a trusted source of help, guidance, and prototyping, growing the policy, managerial, and specialist digital community in government. It should also draw on the cutting-edge expertise of i.AI and AISI to counterbalance the lingering dependence on systems integrators and re-calibrate new relationships with SV firms.
2. Develop AI capacity in large departments (e.g., Department for Work and Pensions, the Home Office), who face productivity and equity challenges which require domain expertise. We recommend a digital innovation appointment at the highest level leading a Departmental Digital Centre for every large department and agency. The DSIT Digital Government Centre might work with these Departmental Centres in a federated way, seconding personnel and helping to domesticate and develop the products produced by i.AI.
3. Meet a minimum standard for citizens’ digital capability. Within the DSIT Digital Government Centre, a dedicated Unit should ‘own’ the issue of digital access and connectivity, collecting data and evidence on digital inequality and inclusion; monitoring the costs of not pursuing digital inclusion, and gathering research on best practices across all levels of government. Digital access, that encompasses the connectivity, appliances and skills needed to live in a digital world, should be considered a ‘critical public service’. Recommendations for taking this forward are laid out in a recent British Academy policy briefing.
4. Guide public sector professionals who are already using AI in a ‘bottom-up’ way. Research across public agencies shows that LLMs are being used widely by public sector professionals, including in schools, hospitals, social care, and emergency services. But worryingly, only 32% feel there is clear guidance on generative AI usage in their workplaces. So the Digital Government Centre needs to work across the public sector to roll out and develop the CDDO’s Generative AI Framework for HMG. It is particularly important to remember local government here, as owner of some of the most important citizen facing public services.
5. Initiate R&D in public sector data-intensive digital technology, incentivising new research partners to offer independent, neutral advice. Potential solutions include setting up a government innovation fund or developing new funding models to enable government departments and agencies to collaborate with university partners and research institutes. UKRI could align research funding calls with the departments’ published Areas of Research Interest.
Download the new paper, ‘How to Build Progressive Public Services with Data Science and Artificial Intelligence‘, by Helen Margetts, Cosmina Dorobantu and Jonathan Bright, published open access, by the journal The Political Quarterly.
About the Authors
Professor Helen Margetts is Professor of Society and the Internet, Oxford Internet Institute, University of Oxford, and Director of The Public Policy Programme at The Alan Turing Institute.
Dr Cosmina Dorobantu is the Co-Director of The Public Policy Programme at The Alan Turing Institute and a Turing OII Fellow at the Oxford Internet Institute.
Dr Jonathan Bright is Head of Public Services and Online Safety at The Alan Turing Institute and Research Associate at the Oxford Internet Institute.