So here we go again. Just under three-and-a-half years since its creation, the Department for Science, Innovation and Technology has been abolished.
Science policy bounces back to an expanded Department for Business, Innovation, Science and Technology; DSIT joins BEIS, BIS, DIUS and the DTI in the alphabet cemetery of departments responsible over the past 20 years. Reviews on that are mixed: the Science, Innovation and Technology Committee is worried about fragmentation of policy, while other experts are more positive.
Reviews are much less positive on the shifting and slicing of responsibility for the use of data, digital and AI across the public sector. The Government Digital Service, which moved to DSIT in July 2024 and took in other components (such as the Incubator for AI) from January 2025, is dispatched to DCMS, once again the Department for Digital, Culture, Media and Sport.
But responsibility for public sector adoption of AI moves both to the Cabinet Office, and to the new Office for Prime Minister and Cabinet (part of Cabinet Office). The OPMC will include an AI Taskforce (chaired by Lord Vallance, reporting to cabinet-level AI minister, Kanishka Narayan, and cabinet secretary, Antonia Romeo) and a prime ministerial adviser on AI. The Incubator for AI and the recently-appointed government chief AI officer apparently stay with GDS. I say ‘apparently’; the government hasn’t published the full details and a paywalled report by Politico gives the most comprehensive breakdown of the changes.
From a public sector digital transformation viewpoint, the abolition of DSIT is baffling – and bad. First, any machinery of government change brings disruption and cost: incompatible systems, different contractual terms, working out how teams fit together, turf wars over responsibilities, new arrangements bedding in, and stakeholders having to rebuild relationships are a recipe for policy paralysis, delayed delivery, and opportunity costs all the way down. DSIT was not without problems but is being split just as it was starting to cohere – and there’s not that much time before the next election (August 2029 at the latest).
Second, the specifics of the split are damaging. ‘AI’ being fragmented four ways (BIST, as well as DCMS, OPMC and Cabinet Office) is not a recipe for success. Most worrying is separating AI from data and digital foundations at GDS. ‘AI’ is only one set of technologies and techniques in the digital toolbox, and relies on good quality data and a wide range of digital skills (from data engineering to service design to user research) to work well. GDS is where all that sits; hiving ‘AI’ off from these foundations heightens the risk of it being a shiny bauble at best, and tech solutionism at worst.
Heightening and widening that risk is what’s happening with civil service reform. The Starmer government announced an oddly-overlooked ‘Review into the Organisation, Performance and Transformation of the Permanent Civil Service’ in June. One assumes work on it continues; and one hopes it will now take account of devolution and local government, an anachronistic omission under any PM and an unsustainable one under Andy Burnham [editor’s note: this morning's statement from the cabinet on rewiring the government to shift power, funding and accountability from Whitehall to local leaders" said the new governing framework will "form a major part" of the end goal for the civil service’s future operating model, which the review is shaping].
The second bullet point on what the review will cover is ‘AI and technology as enablers of delivery and transformation’. The cabinet secretary’s objectives talk about “ramp[ing] up AI and technology adoption”, alongside the familiar buzzwords of “efficiency” and “productivity”. Considered deployment of AI can improve how the civil service works, and talk of efficiency is inevitable in a constrained spending environment.
But there is a risk that “civil service reform” comes to mean tech-first AI-driven headcount reduction in the name of efficiency and productivity, rather than taking a better starting point such as “how can public service better serve the public and where could new approaches help?” Throw in those machinery of government changes around public sector AI – giving the cabinet secretary the lead; splitting it off from GDS and all its expertise in the data and digital foundations required for AI to be deployed responsibly – and it heightens the danger that AI is used to replace rather than reform, slashing rather than serving.
Amid this digital disruption, the government needs to reassure public and public servants alike that it will continue to make progress in using data, digital and AI responsibly to better serve the public. We need to know exactly what bits of government are moving where, and how they will work together. We need to know what the remit of the AI Taskforce is, who’s on it and how they will be appointed. And we need to know if the government’s roadmap for digital government is changing route or destination, and how it will avoid diversions caused by these changes.
Gavin Freeguard is director of the new State of the Future think tank, and former programme director for data and digital at the Institute for Government. This article draws on a longer piece about the new government