The Dutch government plans to cut 25,000 civil servant jobs, about 15%, to save €3.5 billion a year from 2035. Police and defence are exempt.

The plan is part of a wider push to use AI to make government smaller, but experts say public services are behind private firms and many tasks still resist automation.

AI may also create new problems, such as floods of complaints and data security worries. Union resistance and public concern could make the cuts hard to deliver.

AI generated The Dutch government’s AI-led efficiency drive to cut 25,000 civil servant jobs is perhaps the most ambitious government plan yet to use technology to reduce the state’s payroll. Making it work – and rolling out similar programmes in other countries – could prove tricky.

The Netherlands intends to lose roughly 15% of its civil servants to save €3.5 billion (S$5 billion) a year from 2035, the government said on Oct 26. Police and defence are among the exempted areas, while ministries, inspectorates and implementing agencies could see headcount reductions of as much as 60%.

Translated to the US, that would mean about 150,000 job cuts, according to calculations by Bloomberg. That is around half the number of federal employees who left their jobs in 2025 after efforts by Elon Musk’s Department of Government Efficiency in the early days of the second Trump administration to reshape the government workforce.

The equivalent in the UK would be approximately 75,000 jobs, while in France it would be slightly above 50,000, although it is hard to directly compare civil servant numbers and roles across countries.

The size of government bureaucracies is a political issue in many advanced economies. Dutch government headcount surged 41% between 2018 and 2025. The UK’s civil service has grown by more than a third in the last decade to more than 500,000. There, the opposition Conservatives promise to reverse that increase, while Reform UK has pledged to go even further.

Some argue such ambitious and controversial targets can be achieved. The Tony Blair Institute, set up by the former British prime minister, argued in a report published in July 2024 that artificial intelligence could partly automate over 40% of public sector tasks, saving a fifth of workers’ time. (The report itself was criticised for using AI in its analysis). “The speed with which these gains are realised is within the government’s gift to determine,” it said.

Blair this summer called for “not civil-service retraining, but a new cadre of workforce”. He has argued that AI can help to square the circle of being able to afford ageing populations and the increasing cost of health and pensions, and that a politics focused on issues such as tax, spending and nationalisation rather than AI is decades out of date.

While private companies have already announced waves of AI-driven job cuts since the launch of ChatGPT in 2022, governments are some way behind. Their AI use “trails the private sector” and initiatives “often remain in pilot phases”, according to a report by the Organisation for Economic Cooperation and Development in September 2025. They are behind on adopting the technology, getting more work from AI rather than less, and there are plenty of tasks they do not want automated.

Until AI Minister Kanishka Narayan intervened in October 2026, the UK’s official AI guidance recommended keeping prompts short, using smaller models, not saying thank you, and only using AI when needed, to lessen its environmental impact. Guidance given to Bloomberg by the country’s Environment Department still refers to Google Bard, a brand that was retired in favour of Gemini almost three years ago.

In a Treasury announcement last week detailing £16 billion (S$27.3 billion) of efficiencies and savings, AI gets relatively scattershot mentions, and no price tag. And in places, talking about AI is a bit like putting the cart before the horse. As recently as February 2025, almost 10% of English health trusts still lacked an electronic patient record.

The French government said in June it is rolling out a conversational AI assistant to all public sector workers after a 10-month trial. During the test phase, the assistant cut the time spent on writing tasks by about 12% and by 16% for summarising documents. However, France’s government has said it will not accept an AI that replaces civil servants or dehumanises relations between citizens and the administration.

Meanwhile, governments are finding that AI is creating more work, not less, as voters and businesses embrace the technology.

A study by a team including Chris Schmitz of Berlin’s Hertie School and co-authors published in August documented 84 cases across 11 countries of what they call “agentic flooding”: an AI-facilitated surge in complaints, court claims and other requests to government. Most involve chatbots drafting polished, legal-sounding text. The quickest fixes, such as fees, risk shutting out people with genuine claims, the authors warn.

Even where AI could technically do the job and respond to that deluge, there are tasks governments may decide it should not. Britain faced a massive backlash to proposals to curb trial by jury, let alone trial by AI judge.

“The big question is whether this government will be able to meet its targets,” ING chief economist Marieke Blom said of the Dutch plan, adding that previous cuts to civil servant headcount have been followed by increased hiring. “AI can lower the amount of people working in the government, but it can also lead to ‘needs’, such as when government institutions are bombarded by complaints as people use AI-agents to file them.”

Then there is sovereignty. With few cutting-edge AI models built in Europe, policymakers might be reluctant to outsource swathes of the civil service to American or Chinese products. (The Dutch say they will favour domestic and European suppliers.) Britain’s reliance on US cloud giants is already alarming its politicians, while a flagship £330 million data contract with Palantir for the health service has grown increasingly contentious over questions of value for money and data security.

“The thing the British public really doesn’t like is their private sensitive data, particularly their healthcare data, going to foreign billionaires,” Narayan said in September.

Governments are also seeking to reassure their citizens that AI will not lead to widespread job cuts, and to encourage them to adopt rather than fear the technology. Narayan regularly talks about developing and adopting technology “in line with Labour values”.

Even if a government tries to cut swathes of jobs, civil servants are often highly unionised and likely to resist. Across high-income countries analysed by Bloomberg, union membership among civil servants was on average more than double that of the workforce as a whole. More than a quarter of US federal employees belong to a union, versus a tenth of all workers. In South Korea, it’s two-thirds versus 13%.

Perhaps AI is simply a convenient cover. Labour MP Anneliese Dodds told Bloomberg last week of a growing number of “performative mentions of AI as a reason for redundancy” among companies. “It seems like perhaps the redundancies were already in train, but there’s a belief that using the term AI will somehow make the company look more cutting edge,” she said.