Marcus Rajkowski, Head of Public Sector EMEA, and Leonard Schreij, VP EMEA, spend a lot of time inside European public institutions. What they keep finding is the same problem in different forms: talented people, buried in paperwork, unable to do the work they were hired for. This is their case for why that has to change.
Every week, 300 million hours of public sector capacity across Europe vanish into paperwork. The people hired to serve citizens spend their days reviewing documents, filling templates, and cross-checking files. AI will not replace that workforce, but it can free it to do the work it was actually hired for.
Every morning, close to 20% of Europe’s entire workforce clocks in at a public institution. In several Nordic countries that figure climbs to 25 or even 30%.[1] These people teach children, care for the elderly, process benefit claims, adjudicate disputes, maintain infrastructure, and make sure that when someone calls for help someone shows up.
The financial commitment is just as large. Social protection spending alone runs at about 19% of GDP across the EU.[2] This is the operating budget of a functioning society.
Yet the system that sustains all of this is in many places quietly breaking down. The sheer volume of administrative work is swallowing the people and the hours that were supposed to go somewhere else.
When the paperwork becomes the job
Think about what a social worker’s week actually looks like. The British Association of Social Workers found that 65% of it goes to paperwork, whether that is inputting data into templates, cross-checking documents, writing case notes, or drafting letters. Only 20% is spent in direct contact with families.[3] These are trained professionals, hired to protect vulnerable individuals and support people in crisis, who spend two thirds of their time on documentation. The aspiration should be 80% of time with the people they serve. The reality is nearly the reverse.
This is not a one-off. Across social services, healthcare, education, and public legal administration, employees routinely spend most of their day on administrative tasks rather than engaging with the citizens they exist to serve. The system was built for people, but somewhere along the way the paperwork became the job.
Scaled across the region the numbers are staggering. Roughly 300 million hours go to administrative work in the European public sector every single week. That is the equivalent of 8 million people working full-time on paperwork alone.
The burden is concentrated in high-volume and text-heavy processes that recur across agencies, from case files and correspondence to compliance checks, legal review, permit processing, and records management. These are not peripheral tasks. They are the machinery of government, and that machinery is currently running slower than the demands placed on it.
Pressure from every direction
Administrative overload alone would be serious enough, but it is hitting at the same time as demand is surging. Caseloads are climbing across the board. In the UK justice system, the Crown Court backlog reached an all-time high in 2025.[4] More citizens need more decisions faster, and the queue keeps growing.
At the same time, the workforce handling those caseloads is shrinking. Across Europe, public institutions face a generational turnover unlike anything in recent decades. In Germany alone, nearly a third of public sector employees are expected to retire within the next ten years.[5]
The talent pipeline is not filling the gap. Only 28% of Gen Z respondents express interest in public sector careers, compared to 41% for the private sector.[6] The old formula of hiring more people and adding more hours will not close this gap. The arithmetic no longer works.
Where AI meets daily work
The conversation about AI in the public sector tends to start at the level of strategy, with national AI plans, investment commitments, and regulatory frameworks. All of that matters. But the most immediate opportunity is not strategic, it is operational. It sits inside the workflows that consume the most time, touch the most people, and create the most friction. That means the administrative and legal processes that run beneath the surface of every public institution.
AI is well suited to exactly this kind of work; document review, summarization, compliance checks, correspondence drafting, data extraction, records classification, decision support. These are all tasks where current AI capabilities can deliver faster throughput, less manual handling, and shorter cycle times. The role of AI here is to prepare the ground so that humans can focus on the parts of their work that require judgment, empathy, and expertise.
Early results back this up. In the UK, a government trial of AI tools across 40,000 civil servants found that staff saved an average of 26 minutes per day, the equivalent of 13 working days a year.[7] The study covered roles across the board, not those with the heaviest administrative loads. For caseworkers and legal officers, the potential is substantially greater.
So what can this look like in practice?
Housing benefit assessments
A caseworker at a municipal social benefits agency starts each morning with a queue of housing subsidy applications. Each one means reading the applicant’s financial documentation, cross-referencing income against eligibility thresholds, checking household composition, and drafting a decision letter. Today, that takes 45 minutes to an hour per case, and most of that time goes to extraction and verification rather than actual judgment. With AI preparing each case file overnight, summarizing submitted documents, flagging discrepancies, populating the eligibility assessment, and drafting the correspondence, the caseworker arrives to a set of cases ready for review and revision, rather than from-scratch assembly. In practice, AI has enabled up to 60% faster response times on complex citizen correspondence.[8] For a family waiting on a housing subsidy, this means hearing back in days rather than weeks, and being able to plan their living situation accordingly.
Tax authority legal disputes
A legal officer at a national tax authority handles appeals from taxpayers contesting assessments. Each case requires reviewing the original tax decision, the taxpayer’s grounds for appeal, relevant case law, and internal precedent, then preparing a structured legal brief. Much of this prep work is repetitive. The same statutes, the same dispute types, the same formats. AI can compress preparation time for these disputes by up to 40%,[8] summarizing prior rulings, extracting the relevant legal arguments, and drafting an initial brief for the officer to refine. Speed matters, but so does quality. When preparation is more thorough and consistent, and when every relevant precedent is surfaced systematically, the legal reasoning improves. Taxpayers get well-grounded decisions that hold up on review, reducing the cycle of repeated appeals and the erosion of trust that comes with inconsistent rulings.
Immigration case processing
A case handler at an immigration agency reviews residence permit applications. Each file includes identity documents, employment contracts, rental agreements, language certificates, and prior correspondence, often in multiple languages. Today, assembling and verifying a single case can take hours of manual cross-checking. AI can extract and structure the key information, flag missing documents, and draft a preliminary assessment against the published criteria. The case handler then focuses on the judgment calls, the ambiguous situations, the interviews, the cases that genuinely need a human eye. The result: a human always makes the material decision. For applicants waiting on a residence decision, often unable to work, enroll in education, or sign a lease in the meantime, it means getting an answer months sooner.
Municipal citizen correspondence
A mid-sized municipality receives thousands of written inquiries each month, from questions about childcare placements and building regulations to parking permits and waste collection schedules. Each one is read, routed, and answered by a civil servant. Many follow predictable patterns, yet every response is drafted from scratch. The result is not just slow turnaround, it is also inconsistency. Two residents asking the same question about building regulations may receive answers that differ in completeness, accuracy, or tone, depending on who happened to handle it. AI-assisted drafting can ground responses in the same current information, use clear and consistent language, and cover the same essential points, while still leaving room for the civil servant to personalize and adjust where it matters. For residents, this means not just a faster reply, but a more reliable one, where everyone receives the same standard of clarity and completeness.
Environmental permit reviews
A case officer at a regional environmental agency processes applications for industrial discharge permits. Each application includes technical specifications, environmental impact data, and compliance documentation that must be checked against regulatory thresholds. The review is highly structured but labor-intensive. The officer spends most of their time extracting data points from lengthy technical reports and verifying them against fixed criteria. When the workload is high, the risk goes beyond delay, and it becomes an oversight problem. A threshold missed in a 200-page technical appendix. A cross-reference overlooked between two filings. AI can automate the extraction, systematically flag every parameter that falls outside normal ranges, and present the officer with a structured compliance summary, cutting the risk that something slips through under time pressure. The result is stronger environmental protection, following directly from the efficiency gain. The officer’s attention goes to the cases that genuinely need scrutiny, not to the mechanical extraction that precedes it.
The prize is not efficiency alone
Productivity gains matter, particularly when budgets are tight and demand is rising. Fewer hours per case, faster outputs, lower cost per decision. But the deeper opportunity is about outcomes.
When a benefit decision arrives in weeks rather than months, the family waiting on it can plan. When a caseworker spends less time on data entry and more time in conversation the quality of that interaction changes. When legal disputes are prepared more thoroughly and consistently, the rate of appeals drops. Evidence from real-world deployments suggests that AI-assisted case handling can reduce appeals by 20 to 30%[8] because the right information is reviewed, the right criteria applied, and the right reasoning documented the first time around.
Scale the impact across institutions and the numbers become genuinely significant. Improving efficiency in administrative roles by 50% could unlock capacity equivalent to four million professionals across Europe, more than the total number of practicing nurses in the entire EU.[9] The point is not to cut headcount. It is to redirect the time that already exists. The same caseworkers, legal officers, and civil servants, freed from hours of repetitive processing, can spend that time on the higher-impact work their institutions desperately need, like more thorough case assessments, more direct engagement with citizens, faster clearance of backlogs, and better quality decisions.
The private sector has already moved. Roughly 88% of organizations now report regular AI use in at least one business function, and workers using AI tools report saving 40 to 60 minutes per day on productivity tasks. The public needs this technology more than most. The question is whether it will adopt it at the pace necessary to get the most out of the opportunity.
Guardrails, governance, and getting it right
None of this can happen without trust, and trust requires more than good intentions. Public sector AI must operate within frameworks that are robust on privacy, transparent in how decisions are made, and clear about where human oversight is required. Citizens rightly expect that decisions affecting their benefits, their legal rights, or their applications are subject to source criticism and human accountability. The distinction between AI-assisted and AI-decided is essential, and governance frameworks need to enforce it through clear escalation paths, audit trails, structured feedback loops, and training for the people who work alongside AI.
Getting governance right also means getting the infrastructure right. Public institutions handle some of the most sensitive data in society, from tax records and benefit applications to health information and legal case files. European public institutions need AI solutions that are EU-native, built within European legal frameworks, processing and storing data on European infrastructure, and fully aligned with the data protection regimes that citizens already rely on.
The ambition is there. The EU has launched the InvestAI initiative to mobilize 200 billion euros in AI investment.[10] Sweden is targeting a top-ten global position in public sector AI.[11] As Sweden’s Minister for Public Administration, Erik Slottner, has stated: “AI has great potential to streamline the public sector and improve public service. By reducing the amount of time spent on administration, workloads can be lightened and more time can go towards core activities.” Denmark has defined world-leading public sector AI use as a national imperative. Finland, Norway, and the United Kingdom have each announced ambitious strategies to put AI at the center of public service delivery. These commitments signal real intent and a strong belief in what AI can do. What remains is execution.
Sources
KDZ Centre for Public Administration Research, based on Eurostat data. EU average: 18.8%; Sweden: 28.3%, Denmark: 30.2%, Norway: 32.2%. kdz.eu
Eurostat, Government expenditure by function (COFOG), 2023. Social protection expenditure: 19.2% of GDP across the EU. ec.europa.eu
British Association of Social Workers, 80-20 Campaign Report, 2023. basw.co.uk (PDF)
UK Ministry of Justice, Criminal Court Statistics Quarterly, October to December 2025. Crown Court backlog reached an all-time high of 80,203 cases. gov.uk
Serviceagentur Demografie, Monatszahl März — Öffentlicher Dienst, 2024. serviceagentur-demografie.de (PDF)
Cameron Kennedy, Attracting the Next Generation in the Public Sector, survey research. cameronkennedy.com
Peter Kyle, UK Secretary of State for Science, Innovation and Technology, speech at CityWeek 2025. Trial across 40,000 civil servants; average saving of 26 minutes per day. gov.uk
Based on Legora client deployment data and internal analysis across public-sector workflows. ↑
Eurostat, Healthcare Personnel Statistics, 2023. Approximately 3.7 million practising nurses across the EU (no data for Luxembourg). ec.europa.eu
European Commission, InvestAI initiative announcement, 2025. digital-strategy.ec.europa.eu
Government Offices of Sweden, press release, February 2026. government.se
