The Government Digital Service (GDS) is urging public sector organisations to prioritise data maturity as the foundation for artificial intelligence adoption, following a joint project with The National Archives that explored how legal data could be prepared for AI use.
The initiative, which completed its discovery phase in April and is now moving into an alpha stage, examined whether legislation and case law held by The National Archives could be optimised for AI applications. Rather than beginning with a specific technology solution, the project focused on assessing the quality, governance and management of the underlying data.
According to GDS, the work demonstrates that successful AI adoption depends as much on organisational capability and governance as it does on technology itself.
‘Good AI starts with good data’
In a blog outlining the findings, GDS argued that AI readiness should not be measured solely by the availability of data or the deployment of AI tools.
Instead, organisations need confidence that their data is well managed, properly understood and supported by the right people, processes and culture before it can be safely used by AI systems.
The discovery project found that The National Archives’ legal datasets are already close to being AI-ready due to high levels of data quality combined with strong organisational practices, governance structures and leadership support.
GDS said this reinforced the principle that high-quality data alone is not enough.
“Good data managed poorly is not AI ready,” the organisation noted, arguing that public sector bodies must consider the wider organisational environment surrounding their data assets.
Focus shifts from AI tools to trustworthy outputs
One of the project’s most significant conclusions was that government may be able to add greater value by developing methods for evaluating and validating AI-generated outputs, rather than attempting to replicate AI tools already being developed by major technology companies.
GDS warned that the rapid pace of AI development means public sector organisations risk investing heavily in solutions that could quickly become outdated.
If you liked this content…
How data readiness is critical to responsible AI in government
DSIT spending plans put digital government and AI at heart of reform agenda
GOV.UK gets conversational as government launches AI chatbot
UK government must confront hard truths if it wants to lead in AI
Data-ready or not, here AI comes. How IAM can help
Had the project focused immediately on building an AI chatbot for legal information, GDS said it could have resulted in government attempting to solve problems already being addressed by global tech providers with significantly greater resources.
Instead, the discovery phase focused on identifying longer-term opportunities that are less vulnerable to rapid technological change, including work around governance, evaluation and standards.
Testing new approaches to government data
As part of the project, GDS worked with The National Archives, the Department for Business and Trade and the Ministry of Justice to test how legal datasets could be exposed using Model Context Protocol (MCP), an emerging open standard designed to connect AI systems with external data sources and services.
The organisations brought together around 40 lawyers, policymakers, engineers, data scientists and academics in a hackathon to explore potential use cases.
According to GDS, early evidence suggested MCP significantly improved the quality of AI-generated outputs when compared with approaches that relied solely on a model’s existing training data.
The alpha phase will explore these findings further and assess how data maturity can help reduce the risks associated with exposing government datasets to AI systems.
A model for wider government adoption
The project comes as government increases its focus on AI-ready data through initiatives such as the National Data Library and recently published guidance on preparing public sector datasets for AI use. The guidance emphasises that technical improvements alone are insufficient and that AI readiness also requires strong governance, legal compliance, metadata quality and organisational capability.
GDS said the National Archives project could provide a blueprint for other public sector organisations assessing their readiness for AI.
The organisation also confirmed that it will shortly launch an updated data maturity service, building on the framework first introduced in 2023, to help departments evaluate the quality, governance and usability of their data assets before embarking on AI projects.
