Dutch government plans 25,000 job cuts in AI efficiency push

By James Calloway

The Dutch cabinet announced plans to eliminate 25,000 full-time government jobs by 2035, a 16% workforce reduction as part of an AI efficiency push.

The same announcement sets savings targets of €900 million by 2027 and €3.5 billion by 2035. These figures make the proposal a material public-sector restructuring, rather than a limited office automation trial.

Yet the government does not say which positions will disappear or how AI will help remove them. Parliament received the headline targets without a defined deployment plan (actual use of specific systems), according to the cabinet’s post and state secretary Eric van der Burg’s letter.

What has the cabinet actually announced?

The post says the cabinet wants to eliminate 25,000 full-time jobs “over the span of a few years.” It presents the reduction as a cost-cutting measure and says the government will lean on AI to support productivity.

The announcement’s headline describes the change as a 16% workforce cut. The underlying text gives no starting workforce total, so readers cannot independently check that percentage or calculate the reduction’s pace.

The government estimates savings of €900 million, or around $1 billion, by 2027. It projects savings of €3.5 billion, or approximately $3.9 billion, by 2035.

Those checkpoints establish scale, but not the path between them. The announcement provides no starting budget, annual target, or breakdown by department and savings mechanism.

The cabinet has set staffing and financial targets without showing how AI will deliver either.

What will AI do inside the government?

The cabinet’s post explicitly says it remains unclear how the government will deploy AI to reduce costs. Its promised next step is to formulate goals and savings plans in the first quarter of 2027.

The government will “come up with goals and savings plans for deploying AI to work more productively,” according to the post. That wording commits the cabinet to a plan later, rather than presenting the systems, tasks, or timetable now.

This timing leaves the 2027 savings target without a disclosed AI method. The announcement does not identify any tool, supplier, pilot, government agency, workflow, or measure of productivity.

Bloomberg, citing earlier reports from Dutch news outlets, said the government had planned to incorporate AI into cost savings earlier in the year. The report also says several AI initiatives already exist across different government sectors.

Bloomberg further reports that Van der Burg wants AI applied more broadly. That supports the cabinet’s direction, but it does not establish which current activities will produce the advertised savings.

The known AI activity provides context for the ambition, not a costed route to 25,000 fewer jobs.

Which agencies and services escape the reduction?

The cabinet says not all government agencies will be affected. Police and defense are explicitly excluded from the workforce reduction.

The accompanying report frames the wider policy as a response to AI’s rapid growth and says no job is secure. The agency carve-outs qualify that broad warning rather than apply it uniformly across government.

That carve-out narrows the proposal’s direct operational reach. It also means the government does not present police and defense staffing as part of the AI-driven savings exercise.

Beyond those two exclusions, the source names no affected ministry, agency, service, or job category. It does not say whether the reductions come from administration, policy work, front-line services, vacancies, or early retirement.

The announcement gives readers no basis for concluding which public services may shrink. Any account that identifies a favored department or occupation goes beyond the cabinet’s stated plan.

Police and defense remain outside the plan, while every other affected service remains unspecified.

Can the job and savings targets be reconciled?

The two cost checkpoints are clear: €900 million by 2027 and €3.5 billion by 2035. The government supplies dollar approximations of around $1 billion and approximately $3.9 billion, respectively, but no exchange-rate date.

What the source does not supply is equally important. It gives no starting spending baseline (the budget against which savings are measured), no annual path, and no formula linking deleted positions to reduced costs. It also leaves open whether the euro figures represent annual or cumulative savings.

The source does not explain how the government will achieve a 25,000-job reduction. It gives no breakdown for the workforce mechanisms that would produce the change or distinguish staffing departures from positions that simply remain vacant.

The first-quarter 2027 deadline adds another gap. The cabinet has set a €900 million savings target for 2027, yet plans to produce AI goals and savings plans only during that quarter.

That sequence could work if existing initiatives already have measurable results. The source does not document such results, identify their scale, or show that they require no additional procurement, integration, oversight, or staff time.

The numerical targets are measurable, but the source does not establish the causal chain connecting AI, job losses, and savings.

What should Parliament test before the deadline?

The announcement leaves several questions that would turn broad intent into an auditable program. Parliament cannot test those answers unless the cabinet supplies them with the first-quarter 2027 plan.

• Which 25,000 positions fall within the reduction, and how does the government count them against a stated starting workforce?
• Which ministries and agencies contribute, beyond the stated exclusions for police and defense?
• Which AI systems, suppliers, data sources, and tasks will support the plan, and what will buying and integrating them cost?
• How does the government derive €900 million in 2027 savings and €3.5 billion in 2035 savings from measurable reductions?
• What service standards and oversight will apply to the AI systems, and who will answer for their effects?
• What workforce measures, including retraining and redeployment, will accompany the reduction in full-time jobs?

The source answers none of those implementation questions. It therefore provides targets for parliamentary scrutiny, but not yet a standard against which to judge delivery.

That distinction matters because cost reductions and productivity improvements are not interchangeable. A government can spend less without producing more, or deploy AI without cutting staff; this announcement assumes a connection without supplying its terms.

Parliament needs workload, savings, and accountability data before the AI rationale becomes measurable.

Can “simpler, smart and smaller” be tested?

Van der Burg wrote in his letter to Parliament that the effort forms part of a push to make the government “simpler, smart and smaller.” The slogan aligns with the staffing goal, but only one of its three terms has a published metric.

“Smaller” has the 25,000-job target and the two savings checkpoints. “Smarter” has no stated productivity measure, and “simpler” has no stated test for whether government will become easier to administer.

The announcement also does not say what AI will change or where it will operate. It provides no list of activities, systems, or agencies that will move from manual work to AI-supported work.

Nor does the post provide figures for implementation and oversight costs. That omission prevents readers from distinguishing gross savings (before those costs) from net savings (after them).

The slogan describes direction, while the missing measures leave its central productivity claim untested.

What happens by the first quarter of 2027?

The first concrete checkpoint arrives when the cabinet says it will publish AI deployment goals and savings plans. That document must identify the activities, systems, and agencies behind the promise.

It must also explain the starting point for both claims: 25,000 full-time jobs and €3.5 billion in savings by 2035. The current post supplies neither the starting workforce figure nor the calculation, despite using both in the headline.

Parliament will then have a clearer basis for assessing the €900 million target for 2027. Without those details, the number functions as a forecast rather than an independently checkable commitment.

If the plan does not name tools, tasks, owners, milestones, and savings methods, AI remains the organizing rationale rather than the operational program. The cabinet’s promise will remain credible only if the 2027 document shows how the government intends to reach it.

The Q1 2027 plan will show whether the cabinet can turn a broad promise into verifiable savings.