Uphill struggle ahead for Government’s use of AI as PAC report reveals the scale of the challenge

Significant challenges lie in the way of Government's plans to support public sector adoption of artificial intelligence (AI). In its report on Government's use of AI, the Public Accounts Committee warns that while AI has the potential to radically change public services, the scale of the task ahead in grasping these opportunities is concerningly great.

The Government is seeking to scale up AI adoption in the public sector. For AI to be used well, it needs high quality data on which to learn. The PAC is warning that too often Government data are of poor quality, and often locked away in out-of-date, or ‘legacy', IT systems, which are partially defined as "an end-of-life product, out of support from the supplier, [and] impossible to update…” An estimated 28% of central government systems met this definition in 2024. Approx. a third of Government's 72 highest-risk legacy systems still lack remediation funding. The report warns that there are no quick fixes here, and calls for funding for the remediation of this kind of technology to be prioritised.

The report further finds slow progress in ensuring transparency in how AI is used by Departments. This jeopardises public trust in its use – key to its successful adoption. By January 2025, only a relative handful of records had been published on a Government website set up to provide greater transparency on algorithm-assisted decision-making. The Committee is calling on Government to address public concerns over the sharing of sensitive data in AI's use.

The Government also has a long way to go to ensure a thriving market for AI suppliers. The PAC's inquiry highlights a number of concerns that the dominance of a small number of large technology suppliers in the AI market risks stifling competition and innovation. The Government's approach to procurement also risks over-reliance on the services of specific companies, and an inability to adapt.

Another barrier to the safe and effective adoption of AI by Government are longstanding and persistent digital skills shortages. Around half of roles advertised in civil service digital and data campaigns went unfilled in 2024, and 70% of Government departments report difficulty recruiting and retaining staff with AI skills. The PAC has long raised concerns about digital skills gaps in Government, and is sceptical that the Department for Science, Innovation and Technology's (DSIT) planned digital reforms will address the problem.

On DSIT's role in holding responsibility for wider AI policy, while the PAC welcomes the recent creation of a new digital centre of Government in the Department, it has serious concerns over whether DSIT has sufficient leverage to drive change across the public sector. The Committee's report is calling for a senior digital officer to be embedded at the top table with senior management at every department, on the boards at each Department, and their respective agencies.

Notes to editors

The PAC has previously reported on the use of AI by the Department for Work and Pensions (DWP) to scan customers' written communications so that it could provide support quickly – you can read that report here, which was drafted prior to media reports that a number of AI pilot programmes aimed at identifying claimants in need of support are no longer in use.

Sir Geoffrey Clifton-Brown MP, Chair of the Committee, said:

“This Committee's role is to help ensure efficient Government policy delivery through our recommendations. The potential for AI to secure widespread efficiencies is obvious, and the need for digital improvements is a theme that will run through all our inquiries as through a stick of rock. The AI industry needs a Government that is on its side, while making sure that any opportunities for improvement are seized in a safe and ethical way. Transparency is key here, as public trust that AI will work for them is central to any successful use of it. We still have a long way to go in this area.

“The Government has said it wants to mainline AI into the veins of the nation, but our report raises questions over whether the public sector is ready for such a procedure. The ambition to harness the potential of one of the most significant technological developments of modern times is of course to be welcomed. Unfortunately, those familiar with our Committee's past scrutiny of the Government's frankly sclerotic digital architecture will know that any promises of sudden transformation are for the birds.

“A transformation of thinking in Government at senior levels is required, and the best way for this to happen is for digital professionals to be brought round the top table in management and governing boards of every Department and their agencies. I have serious concerns that DSIT does not have the authority over the rest of Government to bring about the scale and pace of change that's needed. We hope the recommendations in our report aid the Government in succeeding in bringing public sector systems into the 21^st century for their users, where other efforts have failed.”

PAC report conclusions and recommendations 

Out-of-date legacy technology and poor data quality and data-sharing is putting AI adoption in the public sector at risk. AI relies on high quality data to learn, but too often government data is of poor quality and locked away in out-of-date legacy IT systems. Progress tackling these issues has been too slow, and 21 out of the 72 highest-risk legacy systems in government prioritised as part of the 2022-25 digital and data roadmap still lack remediation funding. DSIT has set out its ambitions to address these issues in A blueprint for digital government. These include more focus on addressing legacy technology including identifying remediation funding, and a commitment to reduce the barriers to sharing data (which is important for unlocking access to data to train AI models) by implementing standards, frameworks and tools. However, there are no quick fixes and addressing these issues will take time, resources, prioritisation and sustained focus. We remain concerned at the scale of the challenge, government capacity to tackle the problem and the significant risk this represents to successful adoption of AI across the public sector.

Recommendation 1. Within six months, DSIT should set out publicly how it intends to:

1. a) Prioritise and ensure funding for the remediation of the highest-risk legacy technology;
2. b) Establish an approach for measuring the costs associated with addressing legacy technology, as well as the costs of failing to act, to increase transparency  and improve decision making
3. c) Track funding allocated for remediation work and take action where progress is slow
4. d) Address the risks to AI adoption resulting from barriers to data-sharing and poor data quality

Public trust is being jeopardised by slow progress on embedding transparency and establishing robust standards for AI adoption in the public sector. Public confidence that the AI technology used by government is fair, accurate, secure and safe is key to successful adoption. Transparency is fundamental to building that trust but as at January 2025, only 33 records had been published on the government website set up to provide greater transparency on algorithm-assisted decision making in the public sector. Some 67% of government bodies responding to the NAO's survey said that support from the centre of government in fostering public trust in AI was very important. DSIT acknowledges that it has more to do to communicate effectively with the public and to be 'demonstrably trustworthy'. Its work to strengthen its digital and data spend controls for high-risk AI and develop technical standards and guidelines must be progressed quickly to build public confidence that AI adoption by public sector bodies is safe and responsible.

Recommendation 2. DSIT should write to the committee in six months to update us on:

• Departmental compliance with the Algorithmic Transparency Recording Standard and further action it is taking to tackle gaps in transparency to strengthen public trust, including to address public concerns over data privacy and the sharing of sensitive data.
• How its strengthened spend controls for high-risk AI use cases will support safe and ethical AI roll-out.

There are persistent digital skills shortages in the public sector and DSIT's plans to address the skills gap may not be enough. 70% of government bodies responding to the NAO's survey identified difficulties in recruiting and retaining staff with AI skills as a barrier to AI adoption. In 2024, around 50% of roles advertised in civil service digital and data campaigns were unfilled. In January 2025, the government's State of digital government review set out the skills challenge faced by government. These include persistent difficulties recruiting, civil service pay levels that are uncompetitive with the private sector and the need for more technical roles within the profession. Its Blueprint for modern digital government sets out a series of priority reforms including plans to adopt a digital-first operating model, assess the optimum employment models, strengthen digital leadership, and assess competitiveness of the overall package for digital and data staff. However, the Public Accounts Committee has repeatedly raised concerns about the digital skills gap in previous examinations of digital transformation in government and we remain sceptical that these reforms will address the issue where previous attempts have failed.

Recommendation 3. DSIT and Cabinet Office should write to the committee alongside publication of the Digital and AI Roadmap to:

• explain how the planned reforms are expected to resolve the skills gap and by when, including how they will encourage the flow of digital skills between the private and public sectors, ensure digital leaders are more influential in decision making and embed a digital-first ethos into the civil service
• explain how they will monitor and evaluate the reforms so they can take action swiftly if reforms are not successful, and 
• set out their plans for reporting progress publicly

DSIT has no systematic mechanism for bringing together learning from pilots and there are few examples of successful at-scale adoption across government. At the time of the NAO report (March 2024), levels of AI use in government were low, but 70% of surveyed government bodies were piloting or planning AI tools. Examples of pilot activity include use of AI to analyse digital images to identify and classify objects, use of natural language processing to summarise or draft text, and use of AI to assess trends and patterns and monitor live data.  More recently i.AI has developed a suite of tools to support civil servants with tasks including minute taking and analysis of consultation responses. However, there is so far little evidence of successful adoption at scale. To grasp the opportunities of AI, government must learn from these pilots, identify the most promising examples, and where appropriate, help drive adoption at scale so the whole of the public sector can take advantage. However, there is no systematic mechanism for bringing together the learning from all the pilot activity across government and disseminating it. We welcome the recent publication of the Evaluation Taskforce's guidance for evaluating the impact of AI tools, and DSIT's commitment to piloting an AI Knowledge Hub to bring together good practice and learning in one place. However, without concerted effort and leadership from DSIT, government risks duplication of effort and cost from siloed pilot activity. 

Recommendation 4. To learn from AI pilots and support the scaling of the most promising use cases DSIT should:

1. a) set up a mechanism for systematically gathering and disseminating intelligence on pilots and their evaluation
2. b) Set out how it will identify common and scalable AI products and support their development and roll-out at scale.

DSIT and Cabinet Office have a long way to go to strengthen government's approach to digital procurement to ensure value for money and a thriving AI supplier market. The technology market is rapidly evolving with a shift from upfront capital purchases to subscription-based models, and greater consolidation of the supply chain. Technology suppliers, charities and academics have raised concerns with us that the dominance of a small number of large technology suppliers in the AI market risks stifling competition and innovation. We also heard concerns that government procurement approaches are ill-suited to a dynamic, fast-paced market, increasing risks to value for money from vendor lock-in and the inability to adapt and take advantage of rapid technological development. The government is aware it has a lot to do to adapt its procurement approach to keep up with the pace of change in the sector, and to leverage its significant buying power. It has committed to setting up an AI sourcing and procurement framework, including rapid procurement and national tenders where appropriate. It has also confirmed it will establish a digital commercial centre of excellence to deepen its expertise. The success of these measures will be critical if we are to see a vibrant AI market in the UK and value for money in the procurement of AI for the public sector.

Recommendation 5. DSIT, in collaboration with Cabinet Office, should set out publicly how its proposed AI sourcing and procurement framework will:

• get the best from all suppliers in a market dominated by a small number of global technology companies and ensure opportunities are available for small suppliers
• align with the overarching digital technology sourcing strategy (on how government builds, buys and partners) that the Blueprint for modern digital government promises
• leverage government's spending power to achieve value for money for citizens
• Ensure those taking procurement decisions across government have access to the right digital skills and knowledge

Realising the benefits of AI across the public sector will require strong leadership from DSIT. Addressing the skills shortage, tackling the persistent issues with poor data quality and out-of-date technology, and providing the support and guardrails the public sector needs to build public trust and adopt AI safely and responsibly all require a clear strategy and strong leadership. In its report, the NAO raised concerns about unclear accountabilities and limited integration of governance across Cabinet Office – which was primarily responsible for AI adoption in the public sector, through CDDO, i.AI and the Government Digital Service (GDS) – and DSIT, which held responsibility for wider AI policy. We welcome the move of CDDO, GDS and i.AI to DSIT in 2024 to create a digital centre of government. This new Government Digital Service, and the proposals in A blueprint for modern digital government to require all executive teams and boards to include a digital leader by 2026, offer an opportunity to strengthen leadership and governance and support digital transformation. However, the scale of the task should not be underestimated and we are seriously concerned whether DSIT has sufficient leverage to drive change across the public sector. One significant way to do this is to ensure that every department and their agencies have a senior Government Digital Service officer embedded at senior management levels in their organisation and agencies to oversee digital policy and procurement change. Equally the relevant procurement and main boards should include people with excellent digital policy and procurement skills.

Recommendation 6.

1. a) DSIT must ensure that the proposed Government Digital and AI Roadmap is underpinned by a clear implementation plan with:

• clear accountabilities at the centre, across government and the wider public sector
• delivery milestones to drive change and maintain momentum
• metrics to track progress and spend over time, to identify whether further levers are needed to support implementation.

1. b) The Cabinet Office should consider the inclusion of a passage in each organisation's annual report and accounts on their progress on digital and AI transformation alongside ensuring the commitment to placing senior digital leaders, with suitable expertise, on all executive teams and boards is met promptly.
2. c) The Cabinet Office should consider whether a Government Digital Service officer with digital and procurement skills could be embedded at a senior level in each Department and agencies. Equally the procurement and main board should contain individuals with these skills.