The Public Accounts Committee (PAC) wants the government to tackle legacy IT problems if it hopes to succeed with AI in the public sector. Their report emphasizes that AI needs high-quality data, but the Department for Science, Innovation and Technology (DSIT) pointed out that much of the government’s data is poor quality and stuck in outdated IT systems.
About 28% of government IT systems are categorized as legacy systems, which means they’ve reached the end of their usable life and can’t be updated or supported. DSIT says these systems are no longer cost-effective and pose significant risks. In its Digital and Data Roadmap for 2022-2025, the government promised to put together remediation plans for its 72 highest-risk legacy systems, but the PAC is raising flags about 21 of those systems not receiving any funding for fixes.
The PAC’s report highlights the urgent need to resolve legacy system issues. DSIT stressed this is crucial not only for leveraging AI opportunities but also for addressing cyber security risks. They’re planning to prioritize systems with the most valuable data and highest security vulnerabilities.
On the hardware side, the PAC noted that moving to cloud services could help the government move away from large datacenters. However, there’s a trade-off between focusing usage with fewer providers for better value and diversifying to enhance resilience.
In a related inquiry, the Cabinet Office mentioned that many legacy systems rely on just a few cloud providers, complicating the migration off outdated services. DSIT aims to work with HM Treasury to review budget requests for legacy system updates, ensuring funds get allocated correctly, as they often shift elsewhere. They’re considering pilots for continuous funding to maintain IT systems over time.
Legacy IT isn’t the only hurdle for AI in the public sector; data sharing remains problematic. A new blueprint aims to lower those barriers, with plans for standards, frameworks, and the creation of a national data library to facilitate access across the public sector.
The PAC also stressed the importance of transparency in data to build trust. Despite this, only 33 records have been published on the government’s transparency website as of January 2025.
Alongside these issues, the PAC flagged the digital skills gap. They’re doubtful that recent efforts to strengthen digital skills in the public sector will succeed where previous attempts have faltered.
Among the PAC’s recommendations is a call to measure the costs of legacy IT. They suggest establishing a system for collecting and sharing information on pilot projects and their evaluations. They also advise the government to identify scalable AI products and support their widespread development.
On the procurement side, the PAC urges DSIT, with the Cabinet Office’s help, to clarify how its AI sourcing framework will benefit suppliers in a market mainly dominated by a few large tech companies, ensuring opportunities for smaller suppliers too.