How Governments are already using AI in Policymaking
Artificial intelligence is no longer only something governments regulate. It is also becoming something they use. Across the UK, the EU, OECD countries and city governments, AI is starting to shape how policymakers gather evidence, analyse public input, improve services and evaluate what works. The biggest change is not that AI is replacing policymaking. It is that policymaking is becoming more AI-assisted. [6][7]
In practice, governments are using AI first in the parts of policymaking that are hardest to do manually at scale: processing large volumes of text, spotting patterns in service data, reducing administrative burdens and supporting evaluation. The OECD’s 2025 work on AI in government argues that AI can improve productivity, responsiveness and accountability across the policy cycle, provided it is paired with the right governance, skills and oversight. [6]
The UK: consultation analysis and civil service capability
The UK has some of the clearest recent examples of AI being used directly in policymaking workflows. In May 2025, the government announced that its in-house Humphrey tool, Consult, had been used on a live public consultation for the first time. According to the government, the tool reviewed responses to a Scottish Government consultation on non-surgical cosmetic procedures and produced results that were nearly identical to those reached by officials, but much faster. [1]
That matters because consultation analysis is a classic policy bottleneck. By October 2025, the government said the same technology would be used more widely across consultations, estimating that it could save around 75,000 days of manual analysis each year, equivalent to about £20 million in staffing costs. The government’s AI Exemplars programme also says the Humphrey Productivity Suite is being used to support policy professionals with tasks such as minute taking, consultation analysis and parliamentary handling, with the aim of improving policymaking. [1][4]
There is a broader capability story here too. The Civil Service’s One Big Thing 2025 programme is focused building build shared capability around a priority area across government, and the current programme is focused on innovation, encouraging civil servants to develop confidence in using new tools and ways of working to improve outcomes. In practice, that creates a useful bridge between AI adoption and policymaking capacity: it frames innovation not as a specialist hobby, but as part of the everyday job of improving public services and government effectiveness. AI works best in policymaking when officials know how to use it critically, safely and with a clear public purpose, and One Big Thing helps reinforce exactly that culture. [3][5]
The EU: building the foundations for AI-enabled government
At EU level, the picture is less about a single flagship tool and more about building the foundations for responsible adoption. The European Commission’s Joint Research Centre has been examining how generative AI could affect public administration and public services, highlighting its relevance for internal operations as well as delivery to citizens. That matters for policymaking because stronger digital capability, shared methods and clearer understanding of risk are all part of using AI effectively inside government. [9][10][11]
The broader European lesson is that AI in policymaking is not just about drafting faster. It is also about strengthening the institutional capacity that underpins evidence-informed decision-making: better data use, better evaluation and better understanding of where AI is useful and where it is not. That more cautious, capability-building approach is likely to shape how AI use grows across European public institutions. [9][10][11]
Across OECD countries: AI is moving into the policy cycle
The OECD’s 2025 analysis is useful because it shows this is not just a UK story. Drawing on dozens of governance approaches and 200 AI use cases, it argues that AI is increasingly being used by governments not only for back-office automation, but also for decision support, service responsiveness and accountability. It specifically notes that AI can add value at each stage of the policy cycle. [6]
One especially important area is policy evaluation. The OECD highlights AI’s potential to help governments understand what works, for whom and under what conditions. That is significant because evaluation is often the weakest part of policymaking. Used carefully, AI could help governments monitor implementation more continuously, spot patterns in outcomes and improve how policy is refined over time. [8]
The wider OECD framing is also useful because it cuts through some of the hype. The near-term opportunity is not machine-made policy. It is better government capability: stronger evidence handling, quicker synthesis, more adaptive public administration and a better chance of connecting operational reality back into policy decisions. [6]
City governments: practical use at the front line
City governments and local authorities are often where the most practical use cases emerge. The National League of Cities’ 2025 AI in Cities report frames AI as part of a wider push toward more efficient, equitable and resident-centred local government. In local settings, AI can help identify trends in service requests, complaints and operational data, which can then inform priorities, targeting and resource allocation. [12]
The UK’s council trials fit that pattern. In 2025, the government said 25 councils were testing AI tools to reduce administrative burdens in areas such as planning and social care, with early evidence suggesting time savings of around 60 minutes of admin for each hour-long meeting. These are operational gains, but they matter for policymaking too, because local policy choices are often shaped by the quality and speed of frontline information. [2] [12]
What this means for policymakers
A few patterns stand out across all of these examples. First, governments are using AI to process information at a scale that is difficult to manage manually. Second, AI is increasingly supporting not only administration but also evaluation, prioritisation and decision support. Third, the most credible use cases are narrow and practical rather than fully automated. And fourth, capability-building matters just as much as technology. Governments need public servants who understand where AI can help, where it can mislead, and how to apply judgement throughout. [6]
That is why initiatives such as One Big Thing are more relevant than they might first appear. If governments want AI to improve policymaking, they need people across the civil service to feel confident experimenting responsibly, learning from practice and focusing innovation on real public problems. The success of AI in government will depend less on a single tool than on whether institutions build the culture and capability to use these tools well. [5][6]
Conclusion
Governments are already using AI in policymaking, but mostly in a pragmatic way. In the UK, AI is helping analyse consultation responses and reduce friction in government workflows. In the EU, the focus has been on building the institutional and analytical foundations for responsible AI use in public administration. Across OECD countries, AI is increasingly being treated as a tool that can support the whole policy cycle, including evaluation and accountability. And in city governments, AI is being applied where operational pressure is strongest and policy consequences are immediate. [1][6][9][12]
The big shift is not that governments have handed policymaking over to machines. They have not. It is that AI is starting to change how governments handle complexity, evidence and administrative effort. The places getting the most value are not treating AI as magic. They are treating it as a tool, and pairing it with skilled people, sound judgement and a culture of innovation. [5][6]
Want to Learn More?
If you found this post useful, you can learn more about how to diagnose policy problems and how to strengthen policy at our upcoming course Artificial Intelligence in Governance and Policy Design.
This 3-day programme gives participants a critical, practical understanding of how AI can be integrated into policymaking — and the governance safeguards needed to manage risk. Drawing on real-world examples from around the globe, you will learn when, how, and why AI can complement policymaking while ensuring human judgement, critical thinking, and ethical responsibility remain at the core.
For a more bespoke training offering for your team or organisation, you can visit this page or contact us here.
Sources
- UK Government, Government-built “Humphrey” AI tool reviews responses to consultation for first time in bid to save millions (2025).
- UK Government, AI experiments see “Humphrey” help townhalls cut costs and improve services (2025).
- UK Government, Shake up of tech and AI usage across NHS and other public services to deliver plan for change (2025).
- AI.gov.uk, Government page, including Humphrey updates (accessed March 2026).
- A Modern Civil Service, One Big Thing (accessed March 2026).
- OECD, Governing with Artificial Intelligence (2025).
- OECD, How Artificial Intelligence is accelerating digital government journey (2025)
- OECD, AI in policy evaluation (2025).
- European Commission Joint Research Centre, Generative AI Outlook Report (2025).
- European Commission JRC / OECD, Evidence-informed policymaking: A pathway to increasing trust and improving public administration (2024).
- European Commission JRC / EU AI Office, framework papers on general-purpose AI model categorisation and impact capabilities (2025).
- National League of Cities, AI in Cities (2025).
