Building better
government

Consultations are essential to effective policymaking, embedding public voices into government decisions, but manual analysis of consultations is a slow and costly process.

Central government receives around 23 million individual consultation question responses per year. Analysing these responses manually requires roughly 75,000 working days and costs around £20 million.

Consult quickly sorts responses into key themes. Analysts can focus on understanding these insights, instead of categorising responses, which means better decisions and faster policymaking.

We have run several evaluations of Consult and found that it consistently matched human performance. Read more on our evaluations.

Consult is also the first tool to receive a ‘Social Readiness Advisory Label’ - a new approach led by Nesta to assess public attitudes to AI tools.

Organisations across the public sector are racing to adopt AI. But despite having similar ambitions, teams often work in isolation, duplicate efforts or simply do not know where to start.

The same questions come up across senior leaders, delivery teams, and AI users: What’s happening in AI across government? What am I allowed to do? What tools can I use, and how? Teams need practical steps to get projects off the ground, not just policy frameworks.

The AI Knowledge Hub is a curated platform that focuses on practical, evidence-based guidance rather than policy. It turns real government AI projects into reusable “how-tos”, tools, and prompts, so teams can build on what has already been learned instead of starting from scratch. One team’s experience becomes shared knowledge that helps many others move faster and with less risk.

Searching through legislation is complex, often requiring specialist knowledge, familiarity with legal language, and knowing exactly where to look. For many policy teams, this means hours spent searching across multiple sources, interpreting dense legal text. 

Lex is a data layer that provides semantic search capability over almost all legislation data. By connecting Lex to your LLM agent (e.g Copilot, Claude Code, etc), you have the ability to search through millions of legislation documents in seconds for relevant information.

Through the Lex API interface, users can ask questions in plain English and receive clear summaries and analysis of the relevant laws, with precise citations to the original legislation.

Lex draws on a large, trusted legal datasets, including 1.5 million sections of legislation (primary, secondary, and EU) and 63,000 court cases from the National Archives.

To learn more or get involved, contact lex@cabinetoffice.gov.uk

The record of parliament - including debates, committees, votes - is publicly available on Hansard. But with 500 pages of transcripts added to the record every day, it is difficult to access, search and use. This is especially true for teams taking bills through parliament, preparing for committee hearings or creating new policy.

Parlex is a research assistant for parliamentary information. It can quickly find and analyse key contributions from members of parliament, parliamentary questions, and committee hearings. 

Parlex helps bring the views of parliamentarians directly into the policymaking process, prepare ministers more effectively, and refine policy before it is introduced.

To learn more or get involved, contact parlex@cabinetoffice.gov.uk

With the rise of LLMs, civil servants needed a secure tool to "talk" to for general tasks such as drafting emails and brainstorming. 

We rolled out Redbox to over 6000 users, spanning 150,000 chats and 1.3 million messages. 

With enterprise tools like Copilot and Gemini now widely available in government, we sunset Redbox to focus on projects delivering more directly for the public.

We have written on a blog on the lessons learned.

Major projects in government, from railways to national IT systems, represent over £805 billion in spending. The National Infrastructure and Service Transformation Authority (NISTA) ensures these projects deliver value, quality, and timely outcomes through detailed assurance reviews.

These reviews are manual and labour-intensive, requiring teams to sift through hundreds of project documents. 

We helped NISTA build Scout, using AI to analyse large document sets, summarising content, identifying key patterns, and surfacing potential risks, all while supporting, not replacing, expert reviewers.