Government needs ‘realistic plan’ for adopting AI, says NAO

Artificial intelligence (AI) presents the government with opportunities to transform public services, but harnessing the benefits will require a “realistic plan” to change business processes, address skills gap issues and manage risks, the National Audit Office (NAO) has said.

In 2023, the Cabinet Office’s Central Digital and Data Office (CDDO) began working with the Department for Science, Innovation & Technology (DSIT) to start developing a draft strategy for wide-scale public sector AI adoption.

The draft strategy commits all central government departments to create costed and reviewed AI adoption plans by June 2024, and sets out a plan to identify common capabilities that can be used by multiple organisations and adopted at scale once tested.

In its Use of artificial intelligence in government report, published 15 March 2024, the NAO explores how effectively the government has set itself up to maximise the opportunities of AI, finding that the adoption strategy requires further development in a number of areas. “Adoption plans that are sufficiently ambitious to deliver the scale of transformational benefits envisaged will require both changes in technology as well as significant business process and corresponding workforce changes,” it said.

NAO chief Gareth Davies added that delivering improved outcomes will depend on tackling longstanding issues such as data quality and ageing IT, as well as building in effective governance of the risks. “Without prompt action to address barriers to making effective use of AI within public services, government will not secure the benefits it has identified,” he said.

While the NAO noted the technology has the potential to deliver billions of pounds in productivity savings – and that the government recently announced funding for a number of AI-related productivity initiatives in its 2024 Spring Budget – it added that the current government’s plan “is at an early stage and does not set out an implementation plan with performance metrics, funding, or [outline which department has] overall ownership and accountability for delivery”.

The significant business process changes envisaged by the NAO include establishing clear accountability structures for senior leadership over AI deployments; addressing the public sector’s reliance on legacy systems and data; and ensuring the right skills are in place to successfully implement the technology.

The report said the government must also assess the potential workforce impacts of greater AI adoption. “Realising the benefits of large-scale adoption of AI will require changes in the roles of civil servants,” it said. “The implications for the overall composition of the workforce and the skills required are not yet considered in detail in the strategy for AI adoption in the public sector.”

Linked to this, an NAO survey of 80 government bodies included in the report identified that many already have difficulties attracting and retaining staff with AI skills.

“In April 2022, there were 3,900 digital, data and technology vacancies, rising by 7% to 4,100 by October 2022,” it said. “One way in which government bodies address this skills shortage is by using contractors, agency workers, and temporary staff, with estimates from autumn 2023 indicating that approximately one-third of digital and data professionals in the civil service is made up of these groups.”

However, the NAO noted the government has committed to reducing the civil service’s reliance on this kind of “contingent labour” to reduce costs and grow long-term capability.

The NAO added that to further reduce the possibility of programme failure, the government must also identify and understand the business need – including the attendant risks of unreliable or inaccurate data and risks to privacy, data protection, or cyber security – before it jumps to AI as a solution.

“The controls in place for digital and technology spend across government do not provide assurance that AI risks have been mitigated: government is still developing its standards, guidance and assurance processes to support adoption of AI and manage risks,” it said. “It has plans to publish further guidance, update its digital and technology spend controls, and further embed assurance into processes for procuring AI.”

Responding to the NAO's report, a government spokesperson said, “As the deputy prime minister set out in his speech on AI for public good, artificial intelligence has the potential to revolutionise public services and boost productivity.

“We have invested over £3.5bn in the technology in the last 10 years, and are currently more than doubling the ‘Incubator for AI’ team, recruiting the best of British talent to drive AI integration across the public sector.”