Digital Skills for Public Good: Building Capacity in Local Administrations (deep-dive)

Why does digital sovereignty matter for public services?

Digital transition is no longer just a question of service modernisation. It now sits inside questions of power, sovereignty, resilience and public authority. As public systems become more dependent on cloud platforms, data infrastructure, AI capability, automation and connectivity, the issue is no longer only what digital technology can do. It is who owns the underlying systems, who governs them, who understands them and who remains accountable when things go wrong.

Citizens do not meet digital government as an abstract strategy. They meet it in the ordinary pressure points of life: booking appointments, accessing housing support, managing benefits, ordering prescriptions, checking records, applying for education or contacting a council. When those systems work, they reduce friction and make public services feel reachable and coherent. When they do not, they create delay, repetition, mistrust and exclusion. The public value of digital lies in whether it improves the lived encounter between people and institutions.

Digital access is already mass public behaviour, not a future aspiration. By the end of 2025, 61% of GOV.UK visits were made on mobile devices. Across the EU, 71.9% of people aged 16–74 used a website or app of a public authority in 2025, while 96.1% of assessed public services were mobile-responsive (Government Digital Service, 2026; Eurostat, 2026b; European Commission et al., 2025). Citizens increasingly expect digital access, and institutions are judged on whether those channels are usable, reachable and trusted.

That is why digital capacity for public good should not be reduced to buying tools or training staff. Local administrations need the judgement, leadership, delivery capability and governance required to ensure digital systems serve public value rather than administrative convenience, supplier logic or institutional drift. Through ComPAct, the European Commission links public-administration modernisation to resilience, innovation, skills, cooperation and the Digital Decade objective that all key public services should be accessible online by 2030 (European Parliament, 2023; European Union, 2023). The significance is not simply that more services will move online. It is that administrations need the internal capability to govern digital systems, manage interoperability, use data responsibly and maintain accountability as services become automated and interconnected.

Much of that challenge sits beneath the visible service layer. Citizens see the form, text message, dashboard, app, stronger mobile signal or quicker response. What they do not usually see is the workflow logic, data architecture, procurement dependency, automation layer, access controls, interoperability constraint or governance model beneath. Yet that unseen layer increasingly shapes what public services can do, what they cannot do, and whose interests they end up serving.

This brief uses the term digital fabric to describe the connected layer of infrastructure, data, workflow, sensing, modelling and governance that increasingly wraps around public systems and links people, places and services together. The rest of this brief looks at that fabric through four connected lenses: automation, participation, workforce capability and governance. If digital public services are the visible surface, digital fabric is part of the hidden architecture beneath. That hidden architecture matters because it is where public capability is either built or lost, with direct consequences for accessibility, trust and public value.

When does automation earn its place?

The debate about AI and automation in public services is often distorted by extremes. On one side sit inflated narratives of transformation, where technology is expected to solve structural problems simply by being introduced. On the other sits a blanket scepticism that treats automation as inherently dehumanising. The more useful lesson from practice lies between the two. Automation works best when it is applied precisely, inside real workflows, with clear governance and a firm distinction between tasks that can be automated and tasks that should remain human.

Drawing on professional practice in local government, robotic process automation has been used in Adult Social Care and Children & Families workflows to address repetitive, rules-based administrative tasks that create avoidable delays, inconsistencies and operational drag. These automations are not aimed at replacing frontline care, simulating judgement or automating the public relationship. They are designed for the quieter, less visible parts of service delivery: repetitive data entry, routine validation, workflow progression and administrative handoffs inside the system.

Some damaging service failures begin at exactly that level. Cases are delayed because a routine step is missed. Payments slow down because data have been entered inconsistently. Staff absence interrupts an already fragile workflow. Backlogs build because repetitive tasks are displaced by immediate frontline pressure. None of this is spectacular, but it shapes whether services function reliably for the people who depend on them.

By inserting software robots into repetitive parts of the process, services can reduce manual input, improve consistency and accelerate workflow. The public benefit is real but often indirect: fewer errors, fewer missed deadlines, more predictable progression and better continuity. Often the strongest gains come from improving the administrative machinery beneath the public interface.

The success of automation does not lie in removing people. It lies in changing where people add value. Robots can handle repetitive low-level processing, while staff remain essential for oversight, exception handling, professional judgement and higher-value work. In care-related services especially, the most valuable parts of the work are precisely those that should not be automated away: trust, contextual understanding, safeguarding awareness, interpretation, accountable decision-making and human care.

Automation also has to become trusted. Staff need to see that controls are robust, that feared issues such as incorrect outputs are not materialising, and that the technology reduces burden rather than introducing new risk. Digital capacity for public good is not built by automating everything that moves. It is built by identifying where repetition dominates, where workflow stalls, where human error is unnecessarily likely, and where administrative burden drains attention away from the work that most needs it.

What changes when robots become agents?

If robotic process automation handles repetitive, deterministic steps inside a known workflow, agentic systems point to the next operating layer. RPAs are strong where the process is stable, the rules are clear and the task can be executed predictably. Agents become useful where information sits across documents, records, emails, systems and knowledge bases; where context must be assembled; and where the task depends on finding and passing forward the right information.

Agents are not simply faster robots. They can search across data sources, extract relevant details, assemble context, draft a response, trigger the next workflow step and escalate to a human where needed. The shift from RPA to agents is significant because it extends automation from repetitive execution into retrieval, coordination and action across fragmented digital environments.

This creates obvious value for local administration, where work is often slowed not only by labour intensity but by fragmentation. Officers search repeatedly for the same details; records sit in different systems; documents are reviewed by hand; and context has to be rebuilt each time. Agents can reduce the burden of searching, extracting, triaging and assembling routine information, particularly where the bottleneck lies in fragmented data rather than a single manual step.

But this is also where governance becomes more serious. The more agents can move across systems, the more important it becomes to know what they can access, who gave them that access, how their actions are logged, what permissions apply, where escalation sits, and who remains accountable for the outcome. Emerging EU guidance under Article 50 of the AI Act reinforces this point by treating AI agents that interact with natural persons as systems requiring transparency and accountability (European Commission, 2026). UiPath’s 2026 trends model reflects the same shift: humans, robots and agents orchestrated across data, applications and workflows (UiPath, 2026).

The workforce implication is not that people disappear. RPA removed repetitive burden and freed staff to focus on exceptions, judgement and more meaningful work. Agentic systems extend that transition into searching, assembling, routing and preliminary drafting. That should free human capacity for the work that matters more: care, interpretation, contextual judgement, relationship management, safeguarding, complex case handling and accountable decision-making. The danger is that institutions misread this as a licence for hollowing out capability. Organisations still need staff who can supervise, challenge, interpret, audit and intervene, and leaders who understand enough of the hidden architecture to govern it.

How does digital fabric shape participation?

Digital capacity in local administration is not only about digitising transactions. It is also about building the conditions through which citizens, communities and local economies can participate effectively in a digital society. Fibre, mobile coverage, small cells, CCTV, IoT, workflow automation, data platforms and modelling tools are increasingly part of the enabling structure of place. They shape whether people can work, learn, communicate, trade, access support and remain connected to public institutions.

This is where digital fabric becomes especially useful. A stronger mobile signal is not just a telecommunications issue. It affects payments, navigation, service access and emergency communications. Fibre-connected CCTV affects operational visibility and public safety. IoT in housing can support earlier detection of damp and mould. Better home connectivity affects whether older people remain connected, children can learn from home, and citizens can engage with increasingly digital systems of health, welfare and work.

The public value here is broader than convenience. Digital infrastructure increasingly underpins participation: jobs, training, appointments, online banking, remote consultations, family contact and the digital economy. That is why digital exclusion is not a narrow service-access issue. It is a question of resilience, opportunity and inequality.

That participation is already visible at scale. By August 2025, the NHS App had 37.4 million registered users and averaged 11.4 million monthly logins; in the previous year, users ordered 61.5 million repeat prescriptions and viewed 87.4 million test results through the app. HMRC reported more than 136 million logins to its app in 2025. These are not marginal interactions. They show that digital public access is now tied to health, identity, tax, income and routine administration at mass scale (NHS England, 2025; HM Revenue & Customs, 2025).

Once local infrastructure is connected, the next layer becomes possible: civic intelligence. Authorities can connect information across planning, highways, emergency response, housing, public health and environmental services, not just to process demand but to understand patterns, anticipate risk and act earlier. Digital twins and predictive tools are potentially valuable where they help institutions ask what is happening, what is likely to happen next, who will be affected and which interventions might reduce harm. That is not mainly a technology story. It is a governance and public-value story (European Commission, 2025a).

What digital skills do public servants need now?

The public-sector workforce is not simply being asked to learn new tools. It is being pushed through a deeper transition in how value is created. Low-level repetitive administrative work is increasingly a candidate for automation by robots and agents. That changes where human value sits.

The more routine searching, routing, validation and extraction can be handled by machines, the more human capacity can be directed toward complexity, judgement, trust, communication, care and interpretation. These are precisely the skills public systems have often undervalued because they are harder to count than transaction speed or labour savings. Yet in many services they are the core of public value.

This is why workforce transition should not be framed mainly as efficiency. It is a social and organisational transition. Staff need trust, role clarity, confidence and progression routes. They need to know what is changing, what remains their responsibility and where their judgement becomes more rather than less important. If that does not happen, digital transition will be experienced as disempowerment rather than improvement.

International skills frameworks reinforce this point. The OECD Framework for Digital Talent and Skills in the Public Sector argues that digital government depends not only on technical skills, but on creating the right working environment, securing the right capabilities and evolving the workforce itself. UNESCO’s Digital Competency Framework for civil servants similarly treats digital capability as both a human and institutional capacity question. Together, these frameworks point away from narrow tool-based training and towards a broader model of public-sector competence: confidence with data and digital systems, cross-boundary working, understanding of risk and ethics, and judgement about where technology should support, rather than replace, human responsibility (OECD, 2021; UNESCO, 2022).

Practical capacity-building examples are emerging across Europe. Estonia’s e-Governance Academy shows how digital-government expertise can be structured into advisory support, training and international knowledge exchange. Latvia’s Digital Academy for Public Administration, supported through the country’s national Recovery and Resilience Plan, builds capability across state and municipal administration. Greece’s civil-servant training initiative in cloud computing and digital literacy brings together government, EKDDA and Microsoft, while its Future-Ready Skills work points to the next stage: using AI to anticipate skills trends and align public-sector roles with future needs. The lesson is not that one model should be copied, but that digital capacity-building is becoming more systematic, more institutional and more closely tied to the future shape of public work. 

The strongest public use of automation, then, is not to empty out the workforce. It is to remove repetitive burden so that human effort can move upward toward the work that most needs human presence, comprehension and accountability.

Who can govern the unseen architecture?

Interoperability is where the hidden architecture becomes practical. Digital public services cannot become genuinely joined-up if systems, standards, data and responsibilities remain fragmented. Interoperable Europe addresses this through public-sector interoperability, digital-ready policymaking and learning resources for public servants working on digital transformation (Interoperable Europe, 2026a; Interoperable Europe, 2026b).

The OECD Digital Government Policy Framework provides a useful foundation for understanding this shift. Its six dimensions - digital by design, data-driven public sector, government as a platform, open by default, user-driven and proactiveness - continue to underpin OECD digital government benchmarking. The framework is valuable because it moves the debate beyond e-government and online service provision towards the deeper institutional conditions required for digital government maturity (OECD, 2020; OECD, 2025a).

This distinction matters. A digital-by-design administration embeds digital technologies, data, leadership and governance into policy design, service delivery and internal operations from the outset. That is different from simply making services digital by default. Digital-by-default models may increase online access, but they can also risk excluding citizens with limited connectivity, confidence or digital skills. The stronger public-administration model is therefore not one that pushes every interaction online, but one that builds coherent, inclusive and accountable digital capability across the whole operating model.

The public sector is no longer at the stage where most leaders are unaware of digital. Many can see the benefits. The harder problem is that many institutions still lack enough leaders who can genuinely govern the digital systems now shaping public administration. Real digital leadership is not fluency in fashionable terminology. It is understanding enough of the underlying machinery to know when to challenge a supplier, when not to automate, where risk sits, what assumptions are embedded in data, what should remain human, and whether a proposal is genuinely problem-led or merely technology-led.

This is where many institutions remain weak. Procurement too often becomes a substitute for strategy. Supplier fluency fills the space where internal judgement should sit. Responsibility diffuses across committees and programmes. The result can be a performative version of digital leadership in which transformation is discussed confidently while the institution remains unable to see clearly inside the systems it is buying or relying on. The question is no longer simply whether a system works. It is who understands how it works, who can challenge it, who can stop it, and who can answer for its effects.

Equality is part of this intelligence. Digital transition is not only about who can access services. It is also about who shapes the systems, whose judgement is recognised, who progresses into authority and whose lived reality is understood early enough to influence design. OECD analysis on women, work and demographic change shows that population ageing, rising care needs and persistent inequalities in paid and unpaid work are directly connected to labour supply, productivity, public-service resilience and healthy ageing. It argues that gender-blind demographic strategies can reinforce inequality, reduce policy effectiveness and miss opportunities to strengthen care systems and workforce participation (OECD, 2025c).

This is not only an equality concern. It is a system-design concern. Women carry a disproportionate share of both unpaid and paid care work, while remaining underrepresented in many technical and digital roles. Skills for Care reports that the adult social care workforce in England is 81% female and 19% male. By contrast, Eurostat reports that women made up only 19.5% of ICT specialists employed in the EU in 2024. Where that gap is not understood, questions about usability, trust, workload, interpretation, safeguarding and human accountability surface too late. This affects system quality (Skills for Care, 2023; Eurostat, 2025).

Who leads and who lags in Europe?

The question of who leads and who lags in Europe cannot be answered with one hierarchy. Some countries are stronger on digital public infrastructure, some on basic digital skills, some on e-government usage, some on institutional coherence, and some on equality within the digital pipeline. The State of the Digital Decade reports are useful because they show that European digital progress is uneven: countries may perform strongly on connectivity or e-government use while still facing gaps in basic skills, advanced skills, interoperability, digital public-service maturity or inclusion.

Estonia is often cited as one of Europe’s most advanced digital states, and its experience is valuable, particularly in relation to digital identity, data exchange, cybersecurity and whole-of-government architecture. The UK has also engaged directly with Estonia on digital government. However, Estonia should be treated as a learning case rather than a simple template. Its model has developed in a smaller nation-state with a specific history, security context, administrative scale and long-term political commitment to digital government. The stronger lesson is that digital public administration depends on coherent vision, trusted architecture, legal safeguards, interoperability, cyber resilience and sustained institutional capability, while larger administrations must also manage balance, scale, public legitimacy and concern about surveillance (Cabinet Office and Government Digital Service, 2013; Ministry of Justice and Digital Affairs, Estonia, 2024; Foreign, Commonwealth & Development Office, 2025; OECD, 2025b).

Different measures produce different leaders. Through a Women in Digital lens, Sweden scores 76.2, Estonia and Ireland 68.0, and Finland 64.8, while Malta, Slovakia, Hungary, Italy, Czechia, Germany and Greece sit lower. If the lens is basic digital skills, the leaders shift again: the Netherlands stands at 84%, Ireland 83%, and Denmark and Finland at 81%, while Romania stands at 32% and Bulgaria at 38%. If the lens is e-government usage, Denmark reaches 98%, with the Netherlands, Finland and Sweden all at 96% (Women in Digital, 2025; Eurostat, 2026c; Eurostat, 2026a).

The more useful lesson is that stronger performers tend to be those that have built coherent conditions around infrastructure, shared capability, governance, data use and public-service design. They matter not because they use more digital language, but because they have made digital capability more operational, more governable and more embedded in the machinery of the state. That is the real dividing line: not rhetorical enthusiasm for innovation, but the ability to build, govern and sustain systems that improve public life without surrendering accountability or human judgement.

Where should the human boundary sit?

The strongest public case for digital and AI is not that they replace human systems, but that they can support them intelligently. They can reduce friction, widen access, improve forecasting, strengthen workflow, support earlier intervention and help institutions act more coherently. Those are real gains. But technology does not acquire moral authority simply because it becomes powerful, predictive or commercially influential. In a debate now stretching from papal warnings about human dignity to tech-sector claims about abundance and optimisation, public administration has to hold a steadier line: digital systems can support public value, but they cannot replace democratic accountability, human judgement or moral responsibility (Leo XIV, 2026).

That human boundary matters most in public life. Public services encounter people in vulnerable, dependent, transitional and deeply human conditions. A machine may be able to perform a function, but that is not the same as saying it should occupy the relational space around that function. Care, companionship, interpretation, accountability and dignity are not inefficient residues to be designed away. They are central to what public service is for.

Governance matters because the real question is never only what can be automated in principle. It is what should be automated in practice, under what conditions, with whose oversight, in service of what public purpose, and with what protection for the human beings affected. A well-governed digital society is not one that automates most. It is one that knows where the tool belongs, where accountability sits, and what must remain human.

Digital capacity for public good is therefore not just about skills, infrastructure or adoption in isolation. It is about building public systems that can use powerful tools without becoming subordinate to them. It is about making the unseen architecture legible, governable and publicly accountable. And it is about ensuring that digital progress strengthens the conditions of human life rather than quietly colonising them.

The purpose of digital progress is not to make human beings less necessary. It is to make public life more intelligent, resilient, inclusive and dignified, while keeping the human centre intact. 

***

Click here to view the full version of this paper, together with its references, in PDF format.