The UK's National Data Strategy must evolve to unlock AI in the public sector | Computer Weekly

"The reason is straightforward - the UK's current approach to AI is built on shaky foundations. The National Data Strategy, last updated in 2022, does not adequately prepare departments to manage, share, or secure their data in a way that makes it ready for AI. To achieve the cost savings outlined by the government, the UK needs a comprehensive strategy for unlocking data from legacy systems to be utilised in AI."

"Most AI projects in government today layer a large language model (LLM) on top of existing legacy datasets. This approach can deliver some useful outcomes. For example, an AI model can sift through hours of Hansard transcripts - the official record of parliamentary debates - and instantly summarise discussions, or scan government policy documents to pinpoint where a particular issue has been raised. These examples demonstrate AI's potential to cut through information overload and support faster decision-making."

"However, these use cases are limited and struggle when applied to large, disparate, and unstructured datasets spread across multiple departments. The underlying challenge remains - fragmented, incomplete, and siloed data. Without consistent data foundations, these proofs of concept cannot be scaled or integrated into critical public services. The infrastructure required to process and analyse this data - from cloud platforms and high-performance computing systems to modern networks and storage - must also evolve with growing demand"

In August, the Department for Business & Trade evaluated a trial of Microsoft's M365 Copilot and found employees saved time but productivity did not increase. The UK's current AI approach rests on weak data foundations; the National Data Strategy (last updated 2022) does not ready departments to manage, share, or secure data for AI. Most government AI projects layer large language models on legacy datasets, producing limited but useful outcomes such as summarising Hansard or locating policy mentions. Fragmented, siloed, and incomplete data prevents scaling proofs of concept. Cloud, high-performance computing, networks, and storage must modernise to meet growing demand.

Read at ComputerWeekly.com