Burnham abolished the UK’s tech department and elevated AI in the same week
Prime Minister Andy Burnham moved quickly after taking office, shutting down the Department for Science, Innovation and Technology, or DSIT, on his second day as prime minister. The department had been created by former Prime Minister Rishi Sunak on Feb. 7, 2023, and lasted two years and five months as the UK’s only central government department explicitly branded around technology.
In the same week, Burnham appointed 36-year-old Kanishka Narayan as AI minister, with a role spanning both the Cabinet Office and the newly created Department for Business, Innovation, Science and Trade, or DBIST. Narayan was also given the right to attend Cabinet, giving AI a Cabinet-level seat in the UK government for the first time.
That timing mattered. During the same week, the National Audit Office, or NAO, published a report saying the government did not have a consistent methodology across departments for calculating workforce efficiencies tied to digital transformation and AI. That criticism landed directly against a figure ministers have repeatedly cited: the claim that digital government and AI could save the public sector £45 billion a year.
Most UK media treated the two developments separately, one as a machinery-of-government change and the other as a ministerial appointment. Put together, they describe a more revealing sequence: the government dismantled the department named after technology and, almost simultaneously, brought AI into the Cabinet room.
The political backdrop behind the reshuffle
The article places the reorganization inside the wider Labour leadership change. Keir Starmer won a sweeping victory for Labour in July 2024, but after two years his personal standing had fallen sharply, with polling showing many voters felt he had not delivered visible change. On June 22, 2026, after MPs in his own party had broadly lost confidence in him, Starmer announced he would resign as Labour leader and prime minister.
Labour rules created an immediate complication. Only sitting members of Parliament could run for the party leadership, and Andy Burnham, the strongest contender at the time, was serving as mayor of Greater Manchester rather than as an MP. The route back came through a by-election. An MP in the Makerfield constituency resigned, Burnham won the June 18 by-election, and on July 17 he was nominated by 379 MPs, more than 94% of Labour’s parliamentary party, becoming leader unopposed. He formally took office as the UK’s 59th prime minister on July 20. Since the 2016 Brexit referendum, he is the country’s seventh prime minister in 10 years.
On July 20, after meeting King Charles III at Buckingham Palace, Burnham returned to 10 Downing Street and pledged to end political instability. The next day he redrew the Whitehall org chart.
How DSIT was broken apart
DSIT itself had been created in a previous reorganization under Sunak. The government carved up the Department for Business, Energy and Industrial Strategy, then pulled digital policy away from the Department for Digital, Culture, Media and Sport to assemble a new department. Michelle Donelan became its first secretary of state.
Under Burnham’s reorganization, DSIT’s powers were split in three directions. Science and technology policy moved into the newly formed DBIST under Jonathan Reynolds. AI policy and the rollout of AI across the public sector moved into the Cabinet Office. Digital government functions, along with the Government Digital Service, or GDS, and the Central Digital and Data Office, or CDDO, were sent back to the renamed Department for Digital, Culture, Media and Sport.
Science minister Liz Kendall left office, and the position of technology secretary disappeared from the Cabinet structure. On the very same day, though, Narayan was elevated as AI minister and admitted to Cabinet. The article frames that contrast plainly: one individual was upgraded, while an entire institutional setup was downgraded.
Narayan’s rise and what he said on taking the job
Narayan’s biography has become part of the story. He was born in Muzaffarpur, Bihar, India, in November 1989 and moved to Cardiff in Wales with his family at age 12. They lived in a one-bedroom flat, and he walked to a state secondary school. He later attended Eton, studied Philosophy, Politics and Economics at Balliol College, Oxford, and earned an MBA from Stanford.
Before entering Parliament, he served as a civil servant in the Cabinet Office under David Cameron and in the environment department under Liz Truss. In July 2024, he was elected MP for Vale of Glamorgan, becoming the first parliamentarian from an ethnic minority background in Welsh history. In September 2025, he became parliamentary under-secretary for AI and online safety. Less than a year later, he moved up to Cabinet rank.
On taking the post, Narayan said AI may be the most important technology in human history and could help deliver a reindustrialized Britain, stronger national security and genuinely improved public services. He also acknowledged the risks and public anxiety around jobs and the pace of change.
The article’s objection is not to the substance of those remarks. It is to the context: he said them one day after the government shut the technology department.
Tech industry response: this is about attention, not just structure
The article argues that a government’s real priorities are revealed less by the agencies it creates than by the ones it is willing to close. Creating a department is a political statement. Closing one forces decisions on budget, staff and legal offices, and carries the cost of doing the work all over again.
Reaction from the UK tech sector was broadly aligned. techUK and Startup Coalition sent a joint letter to Burnham calling the plan “the wrong change at the wrong time.” Their concern was not simply sentimental attachment to an institution. Their argument was that breaking up active workstreams would create disruption when speed mattered most.
The letter specifically mentioned the AI Security Institute, the sovereign AI fund, GDS and UK Research and Innovation, or UKRI. Splitting those efforts across departments, they said, risked slowing delivery.
Dom Hallas of Startup Coalition put the point more starkly. Folding technology into a super-sized business department means UK tech companies must compete with sectors such as steel for the attention of the same secretary of state. Matt Clifford, who had served as an AI Opportunities adviser to Keir Starmer, called the move “a big mistake.” Ian Hogarth, a partner at Plural and chair of the UK AI Security Institute, delivered perhaps the bluntest judgment: a reorganization of this scale would, in practice, distract the departments involved for a year.
The Burnham government’s defense was that business, science and trade needed to be under one roof to produce a coherent industrial strategy. The article does not dismiss that logic outright. DSIT itself had originally been carved out of the business department, and once that split happened, technology policy and industrial policy did not always move in lockstep.
Still, the two sides were talking about different problems. Government officials were speaking in terms of coordination. The industry was speaking in terms of attention. Process can be coordinated. Attention is finite. A minister has only so many hours in the day, and in a department that must handle steel, trade negotiations, industrial subsidies and AI at once, AI becomes one more issue waiting in line.
Narayan’s new Cabinet chair was meant to offset that gap by giving him leverage across departments. That is an upgrade, but it is an upgrade for a person. The downgrade applies to an institutional structure, and institutions accumulate in ways individual appointments do not.
The week turns on two figures: £1.8 billion and £45 billion
The article says the most important feature of the week was not the org chart but the coexistence of two numbers.
The first is £1.8 billion. That was the Office for Budget Responsibility’s estimate, published in November 2025, for the three-year cost of Brit Card, a national digital identity program covering the 2026-27 through 2028-29 fiscal years. The proposed card would have carried a holder’s name, nationality, date of birth, immigration status and photograph, and was designed to verify the right to work as part of a crackdown on illegal employment.
Brit Card was announced in September 2025 and quickly triggered what the article describes as the second-largest petition in UK parliamentary history, with close to 3 million signatures opposing it. In October 2025, the government was forced to remove its mandatory provisions, making the scheme voluntary and allowing existing documents such as passports to remain valid alternatives. Throughout 2025, Starmer defended the proposal and even held private meetings on how to implement it. In July 2026, Burnham scrapped it outright after taking office.
Where would the savings go? The article says that point matters because some outside observers had assumed they might be redirected to AI. Instead, the Burnham government said the resources would be used to help cover household electricity bills as part of cost-of-living relief. That was a political decision, not a technology one.
The second figure is £45 billion. This is the headline number the UK government has used to describe how much digital transformation and AI could save every year. It comes from DSIT’s January 2025 report, State of digital government review, which said a fully digitized public sector could unlock more than £45 billion a year in unrealized savings and productivity gains, equivalent to 4% to 7% of public sector spending. Ministers and the prime minister repeated the number many times afterward, turning it into one of the foundations of the fiscal story around digital government.
The assumptions behind the £45 billion claim
How did DSIT reach that estimate? The department assumed that 100% of routine tasks and 10% of non-routine tasks could be automated.
Laura Gilbert, former head of i.AI, the government’s AI incubator, described the number this way in parliamentary testimony: “It is a reasonable guess, but it is of course a guess. It is very difficult to measure.”
The article stresses that Gilbert was not attacking the figure from the outside. She had spent more than four years in Downing Street, founded the 10DS government data science team and i.AI, and was one of the people inside government best placed to understand the issue. Her candor therefore made the weakness in the number harder to ignore, not easier.
Nick Davies of the Institute for Government was more direct. Unless there is a serious plan to produce substantial workforce reductions or lower capital spending, he said, achieving the full £45 billion would be very difficult. He added that “routine tasks” still had not been clearly defined and questioned whether the government itself knew what the term meant.
The House of Commons Science, Innovation and Technology Committee, in its report Rewiring the state, labeled the figure “worryingly optimistic” and cited it as an example of ministers “overselling” digital reform. In the committee’s view, that kind of overselling was not helping transformation. It was holding it back.
NAO: no common methodology, and not enough digital talent
On July 15, the NAO issued its report, Government workforce planning: lessons learned, and restated many of the same concerns in audit language.
The report said the UK public sector employs 6.2 million people, with staffing costs of £260 billion. It also said 76% of civil servants expect AI to change how they work within five years.
Yet the watchdog found that departmental efficiency plans “do not provide detail of how departments will deliver their expected workforce efficiencies,” and there was no consistent methodology across government. Cabinet Secretary Antonia Romeo had herself said workforce planning in the civil service “does not sufficiently reflect how AI and technology will reshape the civil service.” The NAO added another practical obstacle: the public sector lacks the technical and digital talent needed to deliver those efficiency gains in the first place.
Set side by side, the article argues, the two numbers tell the story of the week. The £1.8 billion figure is a cost. It has a clear time range, support from an independent institution, a place in the budget and a path to audit, so it can be cut. The £45 billion figure is a saving. It has no common methodology, no auditable denominator and rests on an assumption about routine work, yet it remains in circulation because no one has been required to fully prove it.
That, in the article’s framing, is the key point: the government cut the number that could be checked and kept the one that could not.
Costs must be defended. Savings only need believers
The article then draws a broader political conclusion. In accounting, spending and savings are symmetrical. In politics, they are not. Spending has to be defended. Savings only have to be believed.
When fiscal pressure is intense, a savings figure with no denominator becomes a convenient filler. It does not need a procurement process, a delivery report or an explanation for delays. It does not need to face a petition signed by close to 3 million people.
AI, the article argues, is now an especially useful vehicle for that kind of number. It is new enough that there is little historical baseline to test it against. It is broad enough that almost any efficiency claim can be attached to it. And it is technical enough that critics can be dismissed as people who simply do not understand the technology.
The piece closes the UK section by setting out a string of timelines: DSIT lasted two years and five months. Brit Card went from announcement to cancellation in 10 months. Labour went from a landslide to replacing its prime minister in two years. Since Brexit, the UK has had seven prime ministers.
The comparison with Taiwan
The final section turns to Taiwan. Taiwan’s Ministry of Digital Affairs was established in August 2022 and is approaching its fourth anniversary. Debate around it has never really stopped. Should digital governance sit in a dedicated ministry or be distributed across departments? Should resources be concentrated in one place or embedded across the state? The article says these are fundamentally the same questions now being argued in Britain.
For the author, the UK offers a ready-made case study, and a negative one. It shows that a ministry created to signal political priority can, under the next prime minister or president, become little more than a set of movable parts. It also shows that when technology is folded into a mega-department, the first concern from industry is often not policy direction but whether it can still get onto the minister’s calendar.
The article does not say Taiwan should therefore preserve its digital ministry or abolish it. It rejects both conclusions as too cheap. The more useful question, it says, is what the ministry has built that others cannot easily take away.
If the answer is durable technical talent, shared data and identity infrastructure across government, and projects that have been completed and shown to work, then the ministry becomes difficult to dismantle. If the answer is mostly a new org chart and a stack of annual policy plans, then Taiwan is facing the same governance problem the UK is now exposing.
The article ends with a sharp line about politics: for politicians, “controversial” means it may cost votes, while “courage” means it may cost an election.