Five things I think about GDS, data and AI being split up again

When the first briefings appeared over the weekend about abolishing the Department for Science, Innovation and Technology (DSIT), I decided to wait.

Machinery of Government changes have an awkward period in which the headline is certain but most of the consequential detail is not. Which teams are going where? Who owns the budgets, controls and responsibilities? Which minister is accountable? What has actually been abolished rather than renamed or moved? And what is this all supposed to make better?

The dust has not entirely settled and there is still detail to follow. But we know that the Government Digital Service (GDS) is moving into the Department for Digital, Culture, Media and Sport (DDCMS). We also know that responsibility for AI is being pulled towards the new Office of the Prime Minister and Cabinet, with the Minister for AI attending Cabinet.

The reason there’s so much dust to settle is that this is more than blowing the bloody doors off.

I can’t say that bodes well.

The other reason not to dive in was that wise people were quick to say much of what needed saying.

Gavin Freeguard’s original thread and subsequent update have kept track of what is being announced. Tom Loosemore has been characteristically clear that rearranging departments is an unserious use of a government’s limited time, and about the dangers of separating AI from the platforms and data it relies on. Paola Roccuzzo’s roads and buses analogy gives the infrastructure argument to a prime minister who understands devolution. James Plunkett found a good idea within the bad one: turning digital government outwards towards places, local innovators and frontline services. Dave Rogers reflected on what happens when moving the machinery of government. Theo Blackwell asked whether failing the place test was part of DSIT’s undoing, while Sarah Munby has pointed to the hard graft involved in building the thing now being dismantled. 

Sam Freedman has offered an important counterargument to the idea that this is simply random rearrangement. His case is that Burnham has a theory of power: a stronger prime-ministerial centre, a clearer set of priorities and a deliberate attempt to put more strategic and delivery capacity under his own control.

His piece is a useful warning for me. When digital government is your hammer, a lot of things will look like nails. It would be wrong to assess the whole Burnham project through what is happening to digital, data and AI, or to treat either continuity or change as inherently good or bad.

There is continuity I welcome. I’m pleased that the work Alan Milburn has led on young people and employment is continuing. I wrote about the importance of its diagnosis in Strong branches and good shade, and I’m pleased too that Pat McFadden remains at DWP to oversee the emerging delivery of the reimagined jobs and careers service (and I assume skills brief is staying there too and is not on the move again). 

Mind you, there is continuity I am less enthusiastic about, including an effective endorsement of the Home Office’s thinking behind “earned settlement”, which left me pondering what it tells us about belonging, character, and the country we are becoming.

So, as I now join everyone else who has committed key to board, I’m trying to ask a narrower question: does an apparently serious theory of political power include a sufficiently serious theory of how an AI era state will operate?

The last time the oversight of digital government changed, in July 2024, I wrote five things I thought about GDS, CDDO and i.AI moving into DSIT. I wanted to find the positives in bringing digital, data and AI under one roof, while recognising that a department outside the traditional centre of government might not have the clout to make the arrangement work across Whitehall.

Fast forward and the organisations were probably still somewhere in the middle of completing that integration. A new GDS took shape. A blueprint for modern digital government was published. Work on common infrastructure, data, identity, service transformation and AI continued. Things were done.

Now DSIT is gone. GDS is moving to DDCMS. The national digital identity proposition is being cancelled. AI is being given its own prominence at the centre while the services, platforms and data capabilities on which it depends are going somewhere else.

This is not a defence of the most recent status quo or of DSIT as a standalone entity; frankly, some of my interactions with DSIT and GDS over the last 18 months have left me nonplussed. 

But this is an (admittedly much longer) follow up to my piece of two years ago with questions that this news has put on my mind. 

We can see that Burnham has a theory of political power and is making deliberate moves to strengthen the centre. But it seems much clearer about where power should sit than about how the state beneath it should operate. Two years ago I worried that bringing digital, data and AI together might create coherence without clout. I fear this new government structure risks leaving us with less of both.

What did the 2024 change get right? What did it fail to resolve? What have we learned from it? And does moving the digital centre of government into the department responsible for culture, media and sport give us any confidence that somebody knows what future state we’re trying to build?

1. What have we learned from the last time we changed it all around?

In 2024 you could see the thought process behind the moves.

Digital services, data, technology and artificial intelligence are not separate concerns. Bringing GDS, the Central Digital and Data Office and i.AI together offered the prospect of a more coherent focus.

The subsequent State of Digital Government Review found that central capabilities had achieved important things but had not driven reform across the whole public sector. It identified a lack of sustained senior sponsorship, uneven funding and too much concentration on central departments.

The blueprint responded by describing a more integrated GDS with a wider public-sector mission, responsibility for common infrastructure and stronger political sponsorship.

In the two years that followed I don’t think that theory had obviously failed. 

Was it really working so poorly that the first thing a new government needed to do was abandon it? Or perhaps more worryingly, has the incoming government overlooked the Review or Blueprint and missed the important rationale they contain?

It would be too sweeping to say that the wider reorganisation has no diagnosis.

The diagnosis of the political centre is becoming reasonably clear. No. 10 has been underpowered. The Treasury has wielded disproportionate practical influence because it possesses institutional capacity the prime minister lacks. Strategic priorities have too often disappeared into bilateral spending negotiations, weak coordination and departmental bargaining.

Burnham appears to want a stronger centre that can set direction, manage performance and make economic regeneration, devolution and public-service reform prime-ministerial priorities rather than aspirations left with individual departments.

That is a serious argument and may produce a more capable centre of government. It still doesn’t explain what’s happening to digital government.

The new organisation only formally came together in January 2025. According to GDS’s own description of itself, it combined teams from GDS, CDDO, the Geospatial Commission, i.AI and the Responsible Technology Adoption Unit. Its responsibilities stretched from GOV.UK and One Login to digital standards, the government data function, AI products, geospatial data, procurement reform and the National Data Library.

That was a large integration. Organisations have people, cultures, systems and accumulated ways of working. Bringing those together takes time. Taking them apart does too.

Not long ago the government was publishing a roadmap for modern digital government built around the idea of GDS as an integrated digital centre. New units such as GDS Local and CustomerFirst had been launched. Work was under way on One Login, the app, the wallet, the National Data Library, and GOV.UK Chat.

Considering our expectations for evidence based policymaking and robust business cases elsewhere in government, reshuffles happen with remarkably little scrutiny or accountability. No detail had to be published for the original moves and no justification required for why they now need to be dismantled.

Which outcomes were not being achieved? Was it a problem that digital government sat outside the Cabinet Office? Had DSIT become too broad, leaving science, innovation, technology policy and public-service reform in competition for attention? Did the digital centre lack sufficient Treasury backing? Was ministerial sponsorship too weak? Was the integration itself not working?

Or is it simply that a new prime minister wanted a different shape and digital government does not feature in his idea of what must be close to the centre of power?

There may be good answers. But when we design good services, we expect to begin with needs and a problem. We look at evidence, understand where things were failing and find ways to test whether a change addressed the causes.

Even so, the digital part of the reorganisation still looks, as Tom so neatly put it, like deckchair shuffling. The government may have diagnosed the weakness of No. 10. What we’re left scratching our heads over is what failed, why GDS belongs in DDCMS and how its cross-government authority will now work.

Without understanding the government’s answer to that, we cannot tell whether the problem has been solved or merely relocated.

2. The digital centre still and always needs coherence and clout

One of my main concerns in 2024 has not gone away.

Sustainable and genuine transformation relies on bringing together services, data, identity, procurement and capability. The UK government nevertheless seems unshakeably wedded to the idea that these should be divided into tidy and separated portfolios.

A person does not experience a data function, an identity programme and a digital channel. They experience an attempt to find work, register a birth, see their doctor, pay tax, apply for a school place, or get help from their council. Whether that experience works depends on all those things coming together.

That requires coherence. But coherence without authority is just another recipe for frustration and internecine conflict.

At their best, GDS and the organisations that lead digital transformation around the world are not influential simply because of their assembled talent. That influence comes from combining that talent with senior political sponsorship, spending controls, standards, a mandate to intervene and a track record of delivery. Departments understood that technology choices, service quality and commercial decisions were subjects on which the centre could challenge, and if necessary overrule, them.

Moving the digital centre into DSIT may have offered the prospect of greater coherence but it also raised an obvious question about clout. A department outside the traditional centre might bring the relevant capabilities together while still lacking leverage over permanent secretaries, spending decisions, major programmes and departmental priorities.

If that was an unintended consequence needing correction then surely the prescription would be to give the digital centre more authority. Moving GDS to DDCMS seems to do the opposite.

This is not a criticism of the people working there or an argument that digital technology has nothing to do with culture, media or sport. DDCMS houses important parts of the digital ecosystem: telecommunications, online regulation, media and the creative industries. But digital sector policy is not at all the same thing as digital government.

Putting “Digital” into the department’s name does not give its secretary of state authority over HMRC, DWP, the Home Office, the NHS or the operation of local government. It does not provide leverage over permanent secretaries. It does not create control over departmental spending or make the Treasury’s funding model more compatible with service transformation.

Experience around the world shows why a digital centre matters to how the whole of government operates: setting standards, leading the digital and data profession and owning common products used across the public sector. It is hard to see how putting that function inside a department with a sectoral mission strengthens its hand.

I wonder whether anyone, politician or official, is aware of the historical echo reverberating around.

In 2018, responsibility for data policy and governance moved from GDS to the Department for Digital, Culture, Media and Sport. The written ministerial statement at that time presented it as a way of joining government data policy to the wider data economy (FWIW data is entirely absent from this ministerial statement this time around). 

But while those responsibilities moved, GDS itself remained in the Cabinet Office.

The consequences of that separation became part of the subsequent debate about why the centre had lost clarity and influence, and arguably why both data and identity stalled. The 2020 Organising for Digital Delivery report explicitly described confusion over the roles of the central functions and the different incarnations through which GDS had passed.

So yes, 2024 was a considered response to that: reunite digital, data and AI, then integrate them over the course of the Parliament. 

Now we’re moving the whole digital centre into the successor of the department to which one part was moved in 2018. Pity the poor saps trying to visualise Machinery of Government changes over time. 

What problem does this solve?

Who owns the common architecture? Who can require departments to follow standards or stop another closed, duplicative system being bought? Who can see across legacy technology and insist on action? Who is accountable for services that cross organisational boundaries? Who joins central government to councils, devolved governments and the wider public sector? And who makes sure Treasury funding supports iterative, multidisciplinary work rather than forcing everything into a conventional programme?

When the Blueprint was published I stopped writing my version of a Digital Government Review (leaving my data chapter still as an incomplete and unpublished draft) but it was asking similar things. 

Those questions remain largely unanswered, and it seems to me that DDCMS has no obvious authority to answer them.

3. A strong centre is not the opposite of devolution

I understand that this new government is passionate about devolution. 

Theo’s challenge about whether DSIT failed the “place test” is a fair one. The digital centre has historically been too focused on central departments and too inclined to assume that the rest of the public sector will eventually consume what Whitehall builds. Long-time visitors to this parish might remember my noodling on this back in 2014.

Local places should be able to design services around their own communities and institutions. But meaningful local freedom does not require each council to procure another identity service, payment platform, notification system or data-sharing mechanism. Nor should every place have to repeat the same arguments about security, assurance and interoperability before it can improve anything.

That is where Paola’s roads-and-buses analogy is particularly helpful. Local government can shape the routes, frequencies and experience of transport in a place because it does not also have to invent roads, traffic signals, vehicle standards and payment rails from scratch.

The digital equivalent should be straightforward enough.

The centre should not dictate the distinctive shape of every local service. It should help solve common needs once, establish standards, create a healthier market and make proven components available for reuse. Places should then have the freedom to use those foundations in different ways and design around the people and institutions they know.

To its credit, it felt like the previous settlement had started to respond to that. 

GDS Local was created as the strategic link between local government and the digital centre: opening common products to local services, tackling supplier lock-in, improving data sharing and developing a shared view of local-government technology. Its April 2026 update showed real work with councils and suppliers rather than another press release.

That had not magically solved the relationship between central and local government. GDS Local existed because so much had been neglected. But it had begun building something more coherent.

Perhaps there are some advantages to this move. A government serious about devolution should bring more local knowledge and authority into national policymaking. A department concerned with the wider digital economy may also have levers around technology markets, supplier reform and interoperability.

But GDS Local only works if it can connect councils to common infrastructure, the government data function, MHCLG, the Treasury and the big public-service departments. DDCMS has no natural authority over any of those things.

Unless this move comes with an explicit cross-government mandate and actual levers, it risks taking a team designed to connect places to the digital centre and putting it somewhere that is not the centre.

The wider point is that common infrastructure is not there to make local services more uniform. Government as a Platform was about equipping individual teams to build distinct services without rebuilding the basics each time. Devolution needs the same enabling logic for places: common foundations that mean local institutions can concentrate on what should be different. The common things need to be genuinely common precisely so that the distinctive things can be genuinely local.

4. Giving AI a seat at Cabinet does not do the hard yards of transformation

There is an understandable temptation to place AI close to the centre of government because that makes it look important. And giving the Minister for AI a seat at Cabinet is not something I want to dismiss. I have written enough about AI over the last couple of years that it would be strange to object when it receives greater political weight, not least when Kanishka Narayan has demonstrated a brilliant command of the brief.

There are huge questions here about national capability, sovereignty, economic growth, research, safety, security and the global role Britain can play in the AI era. They deserve senior attention.

But political elevation is not the same thing as operational power.

There is a tendency to favour the show of transformation over the work that makes transformation possible. AI intensifies that temptation. It offers something futuristic to announce and a language through which to promise growth, efficiency and better services.

But if the people working on public-service AI are separated from services, data, identity, common platforms and the people who run operations, its new prominence may not translate into very much.

AI cannot be applied to government from above.

Public-service AI needs the same foundations that high-quality digital transformation has always required: good data, purposeful services, resilient technology, multidisciplinary teams, sound commercial choices and accountability when things go wrong.

Those are not preliminaries to the AI work. They are the work.

But the biggest job of all, the thing AI really needs, is for government to be willing to change.

There is a real danger that AI becomes a way to avoid that. Put an assistant on top of a poor service. Automate a process nobody has properly questioned. Summarise information held across systems that should have been joined up years ago. Use a model to compensate for an operating environment that gives a work coach ten minutes every week or two to understand another human being and help them navigate a complicated life.

The technology may do something useful. It may even make a poor arrangement less poor. But it does not remove the need to ask whether the service is designed around the person, whether the process should exist in its current form or whether the organisation can act on what the technology reveals.

That is digital government.

So there was logic to moving i.AI out of No. 10 and into DSIT, and then integrating it into GDS. Teams working on GOV.UK Chat sat alongside those responsible for GOV.UK, data maturity, cloud architecture, One Login and the Service Standard. AI products were being built alongside the people responsible for the services, platforms, standards and data they would need in order to work.

That was not an incidental organisational convenience. It made sense. 

We know that ministerial leadership for AI and responsibility for public-sector adoption are moving towards the new centre, while GDS moves to DDCMS. We do not yet know what happens to i.AI, the other GDS AI teams, responsible-adoption functions or the underlying data capabilities.

Who owns the data and platforms on which that AI depends? Does i.AI move back towards the centre while the wider organisation into which it was integrated goes elsewhere? Where does responsible adoption sit? Are the National Data Library and the government data function now part of culture policy?

We have to hope that when the reshuffle beds in that it will preserve those working relationships. 

At present, though, it looks as though AI has been included in the theory of power while digital government has been left out of the theory of delivery.

This country literally cannot afford to keep making that mistake. We really should have learnt the lesson by now.

5. We keep doing digital identity without doing digital identity, and now we’re even stopping that

The cancellation of the digital identity proposal is politically understandable, but here is another lesson we stubbornly refuse to learn.

The original Labour Together Britcard paper framed national digital identity principally through migration control and compulsory checks. When that proposition went down poorly, ministers subsequently distanced themselves from parts of it. But they retained the political mechanic used to launch it: digital identity was presented first as a way of managing access to work and only later as infrastructure that might make ordinary life easier.

It was not something that could be repaired by appending a list of public service benefits. A system introduced through suspicion into Britain right now was always going to struggle to establish itself as trusted civic infrastructure.

But cancelling the latest foray into digital identity means choosing to stop even trying to answer a question Britain has spent far too much of this century tiptoeing around.

The claim that cancelling it saves money is classic ‘cost of something, value of nothing’ territory. Yes, it removes a visible programme cost, but it leaves behind much larger costs scattered across government, business and people’s lives. 

Services keep verifying the same people. Employers repeat checks. Individuals send copies of sensitive documents. Fraud and error remain easier than they should be. Cross-border services remain harder.

And I am not persuaded that the answer is simply to say the wider digital identity ecosystem survives.

GOV.UK One Login, the planned GOV.UK Wallet and the trust framework for digital verification services are useful components. We can build a login. We can create a regulated market. We can put credentials in a wallet.

But who makes sure everyone can obtain a dependable digital identity? What should the state be able to attest? How do public credentials relate to private providers? How should identity work across central government, councils, health services and the wider economy? How does somebody act on behalf of a child, a parent or a vulnerable relative? What happens when the information is wrong? What is the non-digital route? What makes British credentials useful across borders? What rights and redress make the whole thing legitimate?

And who owns it?

Those are political and civic questions, not just technical ones. An ecosystem can distribute delivery, but it cannot substitute for decisions about universality, rights, accountability and the role of the state.

I helped write the OECD Recommendation on the Governance of Digital Identity. I’m proud of what we put together and how we put it together. It was rooted in country level experience, refined through open consultation, and informed by experiences gathered from countries around the world. Those examples are now five years old, and some of the most striking progress is happening outside the G20, but the research still showed how many choices governments have.

The answer is not a binary between a compulsory physical card and permanent fragmentation. Identity can be user-centred, inclusive, privacy-preserving and supported by non-digital routes. It can minimise the information people disclose and provide transparency, consent and accountability.

Meanwhile, I’m not sure people appreciate how far off the pace we are. A few weeks ago, I was in Uzbekistan and encountered a version of the future that made the limits of the British conversation feel particularly stark.

People there are paying for the metro contactlessly, but not with cards or phones: with the pattern of veins in their palms. Cameras at compressed-natural-gas filling stations connect vehicles to government records and determine whether they are roadworthy, taxed and insured before allowing them to refuel. I was told that electronic patient records rolled out across more than 4,000 clinical settings handling a million appointments every day, in just 18 months. The result is that identity, payment, permissions and data are joined up in ways that make astonishing things pretty ordinary.

Now, before the obvious objections, I did not come home thinking we should copy everything I saw. Some things can be incredibly impressive and properly troubling at the same time. The ability of a state to join these things together can create real public value, but it also concentrates power. We have to be mindful of civil liberties, checks and balances and the legitimacy and safeguards surrounding that power.

Britain’s already catastrophic levels of public trust make this harder and help explain the decision to can this exercise in digital identity. But forever avoiding contentious decisions will do nothing to rebuild that trust.

We talk confidently about AI leadership and Britain’s place among the world’s digital powers. But on something as basic as identity, countries we think of as peers, and others we still too readily assume are catching up, are building at speed and debating how systems they already possess should be governed. We remain unable to settle whether, or how, to create the basic capability at all.

I cannot say these latest identity plans were the right answer. They plainly did not establish the legitimacy needed for something so foundational. But they spoke to a real problem and the government had decided it was worth spending political capital on trying to resolve it. Capital it can’t get refunded or use again.

That was an “easy hard” choice: politically difficult, but important enough to justify doing.

Instead, we’ve gone back to the comfortable position. Useful components, unresolved foundations and another decision not to decide.

So what is the state we’re trying to build anyway?

In my 2024 piece I described digital as mortar rather than another brick. That remains the simplest way I know to explain what troubles me about this reorganisation.

That does send me back to 2012 as a frame for comparison, which is obviously ridiculous. The answer cannot be to recreate it. Fourteen years later (and yes it is a shock to think GOV.UK will be turning 14 on October 17th), the state is different, let alone the world. The technology is different. Society and its expectations are different. 2012’s digital government agenda was incomplete too. 

It concentrated too heavily on central government and had too little to say about local government or health. An attempt was made to resolve identity (though it infamously did not succeed). It took too long to give data the sustained institutional treatment it needed (which was then unceremoniously dumped). Spend controls could be blunt (but went on to be blunted), and the powerful centre threw its weight around without building the bridges needed for what came next (and failed to sustain the leadership for what followed).

But there was clarity about the project. Maude launched GDS as the centre for digital government, and Martha Lane Fox called us to revolution. 

Digital was not principally a technology portfolio or an online-services programme. It was part of reforming how the state worked: how government organised itself, spent money, designed services, developed skills and built capabilities that could be shared.

Yet government keeps treating digital as a brick: something that can be picked up, put into a department, renamed, separated from the bricks next to it and moved again when the organisational fashion changes.

But digital government is the mortar.

It is the connective stuff between services, data, identity, technology, procurement, professions and operating models. It allows a person to move through government without having to understand its departmental structure. It allows a common capability to be reused rather than bought again. It makes it possible for local freedom to sit on top of national infrastructure. It is what turns an AI demonstration into something that can actually work inside a public service.

Mortar is not especially glamorous. Ministers do not get many photo opportunities from standards, interoperability, assurance, service ownership or fixing the funding model. You tend to notice it only when it is missing and the cracks begin to appear.

Burnham has a theory of power. The new Prime Minister seems to know which priorities he wants to hold close, where he wants political authority to sit and how he intends to make No. 10 more capable of driving them through government.

But power still has to travel through the state.

It has to cross departmental boundaries, become a service, draw on usable data, establish that somebody is who they say they are, interact with existing technology and reach people in places that do not map neatly onto Whitehall’s organisation chart.

That is the operating problem the 2024 settlement at least recognised.

Each individual decision can be given a plausible explanation. AI deserves political prominence. Digital-sector policy has relationships with GDS. Devolution should turn government outwards. The identity proposition was politically toxic. DSIT may not have worked as intended.

Yet Machinery of Government changes cannot be assessed one brick at a time. Somebody has to be thinking about what connects them.

The mortar does not decide what kind of building we want. That is a political choice. Digital government is not the answer to every political problem, and it will not generate growth, restore trust or repair public services by itself.

But without that connective capacity, political priorities do not become a state capable of delivering them.

And more of this government’s programme depends on that connective capacity than the machinery changes seem to acknowledge. AI will rely on usable data and services capable of changing. Devolution will rely on foundations that places can draw on rather than rebuild. Identity sits underneath rights, transactions and access to services. These are not technology projects. They are part of whether political intent can travel through the state and become something real.

That is the question left by this reshuffle: not simply where GDS sits or whether AI has a seat at Cabinet, but whether the new centre has an operating model beneath its theory of power.

At present, Britain appears clearer about who should wield power and which bricks it wants to lay than about what will hold the structure together.

What is the state we are trying to build? What does it need in order to work? And who is going to take responsibility for making it happen?