Will digital transformation lead to the end of the national welfare state?

By John Storm Pedersen, Professor Emeritus, Department of Political Science and Public Management, University of Southern Denmark, Denmark

Will digital transformation lead to the end of the national welfare state?

Does digital transformation make it impossible for national welfare states to fulfil their historical task of helping vulnerable citizens who need help the most? In the following, I argue that a specific combination of digital transformation, the digital divide, failed public policies on digital inclusion, dependence on Big Tech, and geopolitical changes can make it impossible for national welfare states to fulfil this historical task. I also argue that governments can develop and implement data-driven provision of public welfare services to create fair, balanced, and documented win-win situations for all key stakeholders in service provision, including vulnerable citizens. I link these arguments to a presentation and expansion of some of the analyses put forward in my book, Digital Transformation and Public Welfare Services – the Opportunity, the Challenge, and the Wildcard (Elgar, 2025).[1]

The national welfare state and the provision of public welfare services

The national welfare state is a political construct. Governments – meaning politicians – decide which groups of citizens should receive which help and how the help should be financed. The welfare services provided consist of seven main elements or building blocks (Pedersen, 2025, pp. 12-13): 1. Communication between citizens and welfare professionals, who act as both gatekeepers and core employees of welfare states, 2. Collaboration between citizens and welfare professionals; 3. Decisions about services, 4. Delivery of services, 5. Welfare professionals as bureaucrats; 6. Feedback for welfare professionals, and 7. The presence of citizens in public service provision.

Since the 1960s, these seven elements/building blocks have gone through a process of digitalisation (Pedersen, 2025, pp. 13-18). This digitalisation has contributed to the emergence of New Public Management (NPM) version 2.0 (Pedersen, 2025, pp. 75-78), and NPM 2.0 has made American Big Tech crucial to the digital provision of public welfare services. In the 2010s, American Big Tech consisted of the ‘big five’ – Google; Facebook; Apple; Amazon; and Microsoft (Dijck et al., 2018, p. 12) – which has now, in the 2020s, become the ‘magnificent seven’: Alphabet (parent company of Google); Amazon; Apple; Meta Platforms (parent company of Facebook and Instagram); Microsoft; Nvidia; and Tesla.

E-services

The digitalisation of the seven main elements/building blocks of service provision enables welfare professionals and citizens to replace physical presence with digital presence. The replacement is primarily enabled by various forms of e-services such as telemedicine, e-learning, e-counselling, DIY^[2] e-service administration, and others.

During the COVID-19 pandemic, most governments strongly promoted e-servicesdue to social restrictions. Post-COVID, what drives the development and implementation of further e-services is a desire to substitute the traditional ‘handheld’ services provided by welfare professionals with cheaper and better services, revitalising the phrase ‘don’t work harder, but smarter’. This post-COVID ethos is reinforced further by the current trend of deploying AI-driven technology in public welfare service provision, putting e-services on a path to becoming the dominant type of service. However, what is at stake in public e-services extends far beyond providing cheaper, better and smarter services.

The digital divide

The digital transformation has created a digital divide between citizens. In the context of public welfare service provision, the digital divide results in unequal access to and use of e-services. The immediate cause of this inequality is the difference in citizens’ IT equipment, IT skills, ability to access, understand and utilise e-services, and their motivation to do so. Meanwhile, the root cause of these differences must, first and foremost, be found in the differences in citizens’ socio-economic living conditions. These include factors such as age, gender, disabilities, education, career, income, as well as citizens’ connection to the job market, whether they live in rural or urban areas, as well as their access to IT support from family, social networks, and public institutions (Gomes and Dias, 2025, p. 167; OECD, 2024, pp. 70-79; Agency for Digital Government, Denmark, 2026, Hvem oplever udfordringer ved det digitale? [Who needs help with e-services]; Minimum Digital Living Standards, 2026, p. 3 About | MDLS Minimum Digital Living Standard for UK Households in 2025: Briefing Paper (June 2025) .Based on this, citizens can be divided into three main groups which, in the following, I will term the winners, the excluded, and the greys.

The winners are citizens who have strong digital competencies and are motivated to deploy these competencies with the aim of receiving as much welfare as possible from public e-services. The excluded are citizens who are excluded from receiving public welfare e-services due to a lack of IT equipment, IT skills, and motivation to use e-services in their everyday lives. The greys are citizens who have several limitations regarding IT equipment, IT skills, IT support, and their motivation to use e-services. As a consequence, the greys struggle in their everyday lives to access and utilise the public e-services needed to maintain an acceptable living standard in terms of health, education, social support, job access, and other essential areas. If their existing limitations are not reduced significantly, the greys are constantly at risk of falling into the excluded group.

The winners will exert pressure on politicians and governments to develop public welfare e-services to the highest possible AI-supported level, because they themselves benefit from this development in the form of lower taxation, additional welfare, and a better work-life balance (Pedersen, 2025, p. 92). If further developments in AI technology deliver even half of what has been promised, e-services will be taken to a new, unprecedented level and this will necessarily raise the bar of what is required of citizens in terms of their access to IT equipment, IT skills, available IT support, and motivation to utilise e-services. According to the OECD, such developments are likely to have disproportionately negative consequences for citizens with relatively disadvantaged socio-economic backgrounds (OECD, 2024, pp. 70-82). Furthermore, women are identified as one of the groups most negatively affected by this development (OECD, 2024, p. 10). To reduce the likelihood of citizens falling into the excluded group, the OECD recommends upskilling and reskilling.

Do the excluded and the greys pose a substantial problem for digital welfare states in Europe?

Two unknowns must be clarified in order to answer whether the excluded and the greys present a substantial problem for the European digital welfare states’ provision of public welfare services to the citizens who need help the most. Firstly, it must be determined how many citizens can be defined as belonging respectively to the group of greys and the excluded. Secondly, it must be defined how the levels of greys and excluded can be influenced in a positive manner.

To clarify the first unknown, the EU Digital Economy and Society Indicators – DESI -index can be applied. The data in the following are retrieved online from this database.[3]

The excluded, the greys, and the winners do not exist as categories in the DESI-index. The index, instead, categorises three types of users: 1. Internet users, 2. Users with at least basic digital skills[4], and 3. Users with above-basic digital skills. To identify the excluded group, we may, at a first instance, take this to be the difference between the total population and the percentage of internet users. Unfortunately, this definition has one major problem. In the DESI-index, the category of internet users refers to citizens who use the internet at least once a week. Using the internet once a week, however, is far from enough to use e-services at an acceptable level in digitalised welfare states. In fact, 8 percent of internet users can be said to belong to the excluded group, as will be shown in the following.

In their cluster analysis of the digital divide in EU countries, Gomes and Dias (2025, p. 157) found six categories of internet users: Non-users (21 percent), basic users (8 percent), information exchangers (12 percent), instrumental users (24 percent), socialisers/entertainers (15 percent), and advanced users (20 percent). The relevant categories, in this context, are the first two: non-users and basic users. The share of non-users, at 21 percent, is consistent with the DESI-index 2017 figures. Internet users within the EU made up 79 percent and, consequently, the non-users – the excluded – stood at 21 percent. Basic users, at 8 percent, are defined as users who use the internet only three times per week, and mostly from home, and since low-level usage of this kind is not enough to access e-services at an acceptable level in most digitalised welfare states, 8 percent points of internet users in the DESI-index can be defined as belonging to the excluded group.

When it comes to the greys, this group may be defined as the difference between internet users in the DESI-index, minus 8 percent points, and the level of users in the DESI-index having at least basic digital skills. Meanwhile, the winners may be defined as users in the DESI-index having at least basic digital skills.

Based on these definitions, the estimated percentages of excluded, greys, and winners across the EU, and in EU member states with the highest and lowest shares of each group, are as follows:

The excluded. In 2017, the highest level of excluded citizens, at 47 percent, was found in Romania. In the EU overall, the excluded represented 29 percent, while Luxembourg had lowest level, at 12 percent. In 2025, Croatia had the highest level of excluded citizens, at 22 percent. In the EU, the same group made up 15 percent, and the lowest level, at 8 percent, was found in Ireland.

As the figures show, percentages differ significantly across EU countries. However, differences have diminished significantly in recent years. In 2017, the highest level was at 47 percent and the lowest at 12, compared to 22 and 8 percent respectively in 2025. Furthermore, bottom levels have been raised overall, both in the countries representing the highest and the lowest numbers of excluded citizens.

The greys. Unfortunately, the DESI-index only holds data on internet users having at least basic digital skills from 2021 to 2025. Consequently, the percentage of greys cover only this period.

In 2021, Romania had the highest level, at 46 percent. In the EU, the greys made up 25 percent, and the Netherlands had the lowest level, at 7 percent. In 2025, Bulgaria had the highest level, at 41 percent, compared to 25 percent in the EU, while the Netherlands represented the lowest level at 7 percent.

Again, percentages differ significantly across countries in the EU. The main reason for this is the differences in the number of users having at least basic digital skills. In 2021, in Romania and Bulgaria respectively, 28 and 31 percent had at least basic digital skills, compared to 79 percent in both Finland and the Netherlands, and 54 percent in the EU overall. In 2025, the numbers in Romania and Bulgaria were 32 and 38 percent respectively, compared to 60 percent in the EU, and 84 and 83 percent in the Netherlands and Ireland respectively.[5]

The winners. Winners can be divided into two groups: A-winners and B-winners. Citizens that have at least basic digital skills can be defined as B-winners because they are able to utilise most of the e-services offered in digitalised welfare states. Citizens with above-basic digital skills can be defined as A-winners because they can utilise advanced e-services such as, for example, e-hospital care (home admissions rather than hospital admissions), advanced e-learning, periodical e-teaching of children at home, and continuing education. In the EU as a whole, A-winners made up 26 percent in 2021, and 31 percent in 2025. As above, numbers vary across countries. In Bulgaria, A-winners represented only 8 percent in 2021 and 12 percent in 2025, compared to 52 and 56 percent respectively in the Netherlands.

Considering the above, do the excluded and the greys present a substantial problem for digital welfare states in the EU? It is impossible to answer that question with a simple yes or no. Since numbers differ across EU countries, so too will the answer. Furthermore, in each EU country, the answer will be linked to that country’s opportunities to reduce their levels of greys and excluded. In other words; different answers, corresponding to different problems and solutions, will be given for different countries.

Problems and solutions

One cluster of countries in the north and north-western part of the EU has relatively low levels of excluded and greys, combined with a relatively high level of winners. The countries in question are Denmark, Finland, Sweden, the Netherlands, and Ireland. In these five countries, the excluded made up between 8 and 11 percent of the population in 2025 (having been steadily falling since 2017), while the greys made up between 7 and 20 percent. Meanwhile, the A-winners represent between 38 and 56 percent – far above the six countries with the lowest numbers. These are Romania (11 percent), Bulgaria (12 percent), Cyprus (18 percent), Slovenia (20 percent), Latvia (20 percent), and Slovakia (21 percent).

In these cluster countries, more than 97 percent of citizens have internet access. Consequently, internet access is not a problem. The challenge, the countries face is to establish realistic benchmarks for what the acceptable levels of excluded are, and realistic plans for looking after their interests. Considering that numbers are already low in these countries, and that there will always be elderly citizens, disabled citizens, citizens with significant health issues, and citizens living under relatively poor socio-economic conditions in any society, more research will be needed to establish the necessary benchmarks and strategies related to the excluded. In this context, one question will be of particular importance: how can welfare societies help the group of excluded citizens that already exists and always will exist? In my book, Digital Transformation and Public Welfare Services – the Opportunity, the Challenge, and the Wildcard (2025), I concluded that the best way to do this is to generate a creative cooperation between welfare professionals and data analysts within public welfare institutions’ daily operations. One approach to carrying this out in practice is discussed later under the headline ‘Data-driven provision of e-services and documentation of win-win situations’.

In relation to the greys, the main challenge is twofold. First, citizens defined as greys must be prevented from falling into the excluded group. Second, citizens defined as users with basic digital skills must be prevented from falling into the group of greys. If governments in the aforementioned cluster of countries fail to solve this challenge, the numbers of greys and excluded will increase, and this will create major problems for welfare states; public expenditures will increase, available welfare will decrease, and difficulties in helping the citizens who need help the most will grow.

Why does the cluster of countries discussed above perform so well? Given that socio-economic factors, as mentioned, are a root cause of the digital divide, the intuitive and most likely explanation is that these countries have well-established and well-developed universal welfare states. These welfare states have public welfare institutions designed to improve poor socio-economic living standards; they support broad and high levels of education, and they facilitate skilled jobs and therefore relatively high-level incomes for the majority of citizens. Furthermore, they support and facilitate the integration of unemployed and disabled citizens into the labour market, and promote gender equality within education and employment. More research must be carried out into the links between public welfare institutions, socio-economic factors, the digital transition, and the respective levels of excluded and greys, in order to understand how best to improve poor socio-economic living conditions with the aim of, as far as possible, reducing the numbers of excluded and greys.

However, well-functioning public welfare institutions alone are not enough. Social frameworks also play an important role. Denmark, as one of the cluster countries, provides an illustrative example of this. A survey conducted by Statistics Denmark in 2024 showed that 10 percent of digital e-service users need assistance. Out of those needing assistance, 58 percent received help from family and friends, 50 percent from public institutions, and 13 percent from Google and chat forums (Agency for Digital Government, 2026. Hvem oplever udfordringer ved det digitale?). In general, the following six principles are applied to help and support users of public e-services: 1. Be aware of the consequences of digitalisation, 2. Design solutions for all citizens, 3. Communicate so that everyone understands, 4. Assist citizens with their digital tasks, 5. Help the helpers, and 6. Provide usable alternatives (Agency for Digital Government, 2026. / https://digst.dk/digital-inklusion/principper-for-digital-inklusion/ ). Out of these principles, 5 and 6 are especially designed to help and support the excluded and the greys.

To sum up: public institutions and public policies need to address relatively poor socio-economic living conditions and the social frameworks in which citizens use e-services, if the levels of excluded and greys are to be kept down. A significant challenge to carrying this out in practice is that most data on e-service users are focused mainly on users as individuals and fails to capture their social frameworks. The research underlying the Minimum Digital Living Standard (MDLS), conducted at Loughborough University and the University of Liverpool (2025) in the UK, may be used as a starting point for solving this challenge (About | MDLS Minimum Digital Living Standard for UK Households in 2025: Briefing Paper (June 2025).

The research has a realistic approach to citizens’ use of e-services and may therefore be of particular use in this context. In order to study citizens’ use of services in everyday life, the research defines relevant types of households: those with or without children, those of working or retirement age, those in urban or rural areas, and others. Furthermore, the research specifies the requirements – IT equipment, IT skills etc. – which enable a household to have an acceptable MDLS, and defines how to support a household in meeting these requirements. The research shows that, in the UK, 45 percent of households with children struggle to meet acceptable levels of MDLS, and that the main reason for this is poor socio-economic living conditions (A Minimum Digital Living Standard for UK Household, 2025, page 3).

Given the above, governments must develop flexible public policy ‘cycles’ to prevent the greys from falling into the excluded group, and the users that possess basic digital skills from falling into the group of greys. Furthermore, governments must ensure that welfare state institutions translate the appropriate policy into practice as an integral part of their daily operations. The following five factors are particularly important for achieving these goals in practice.

First, governments must define the necessary minimum requirements regarding citizens’ IT equipment, IT skills, access to IT support, and motivation. This will be a difficult task because the technological advancement of e-services moves at an increasingly rapid speed, and user requirements are therefore constantly raised.

Second, governments must formulate and implement a public policy to ensure that citizens get the support they need to meet the defined minimum requirements. For the same reasons as above, this too will be a difficult task.

Third, governments must develop and regularly conduct real-life tests – similar to the tests related to MDLS – in order to monitor how many greys fall into the excluded group, and how many of the users with basic digital skills fall into the group of greys.

Fourth, governments must, on the basis of the hard data/statistics obtained from the conducted tests, update and, if necessary, redefine the minimum requirements and, subsequently, their public policy.

Fifth, governments must repeat steps 1 to 4 in a continuous cycle.

It is essential that these five steps are supported by research and that European standards for minimum requirements, testing, and feedback are established, in order to compare developments across countries.

If governments fail to develop and implement flexible policy cycles, the levels of excluded and greys are likely to increase, which will create major problems for welfare states, societies, and governments. Governments can, of course, maintain the old welfare states as they are, and as far as possible take care of the excluded group and help the greys. This, however, is not an ideal solution. In this scenario, neither the excluded nor the greys will benefit from the extra welfare that digital welfare states offer other citizens. On top of that, it will be difficult for governments to maintain – and legitimise – their old welfare states within digital societies, as these will become increasingly costly and low-performing compared to digital welfare states.

The cluster countries, who have all performed well during the digital transformation, face two big challenges which they share with most other countries. Namely, a growing dependency on Big Tech’s digital ecosystems and a vulnerability to geopolitical changes, which in combination could result in governments losing control and national sovereignty over the provision of public welfare e-services. Small countries, and therefore all the well-performing cluster countries, are especially at risk. This will be discussed in the last two sections.

When looking at how other EU countries have fared in the digital transformation, Bulgaria and Romania fall far behind. They will need help and support from the EU and other member states to catch up in relation to users’ digital skills, public welfare e-services, and their levels of excluded, greys, and winners.

The positions of the remaining EU countries fluctuate in terms of their levels of excluded, greys, and winners. However, one tendency is clear: the difference between countries is decreasing over time. Generally, this is due to levels being improved from the bottom up. In countries with the highest numbers of greys, the simplest, most straightforward strategy for reinforcing this tendency is to push and pull as many of the greys as possible into the group of users possessing at least basic digital skills – and to keep them in this group for the coming years. To achieve this, governments must improve relatively poor socio-economic living conditions and the social frameworks for citizens’ e-service usage. This means improving education, gender equality, income levels, and the daily support provided to different types of households.

One way for governments to do this in practice is to develop and implement flexible public policy cycles based on hard data/statistics. If, however, governments do not manage to solve the task, they will be faced with a significant challenge: given the expected rapid development of increasingly advanced e-services over the coming years, some users with basic digital skills will be at high risk of falling into the group of greys, and some of the greys will be at high risk of falling into the excluded group. This situation is likely to materialise in, at least, some EU-countries in the near future. If this happens, governments may find it increasingly difficult to legitimise the digital welfare state and its e-services due to the following reasons: the excluded do not benefit from the digital welfare state, and too many greys struggle to receive their fair share of the extra welfare that the digital welfare state offers other citizens.

Following this, a fundamental question arises: how can we know, based on hard data/statistics, whether the digital welfare state, as promised, delivers better, cheaper, and smarter services to citizens (including vulnerable citizens) than the traditional welfare state did? The following two sections will show how to address this issue in practice, based on already existing tools and knowledge.

E-services, win-win situations, and some disadvantages to e-services

An important argument for expanding the use of e-services is that they enable win-win situations for all key stakeholders in public service provision, and that this can be documented by hard data/statistics. Key stakeholders are defined as: governments (which in this context include political-administrative bureaucracies/civil servants), citizens (as both tax-payers and end-users), welfare professionals, data analysts, and Big Tech.

Telemedicine for COPD and diabetes patients may serve as an example of an e-service that enables win-win situations which can be documented with hard data/statistics (Pedersen, 2025, pp. 65-69). By deploying telemedicine for COPD and diabetes patients, governments can both reduce public expenditures and increase COPD and diabetes patients’ objective and self-perceived welfare. This is achieved by avoiding costly admissions to hospitals and clinics, and by the fact that patients themselves perceive being hospitalised, treated, and counselled at home as improved welfare. In addition, telemedicine can make use of objective, individual health data and predefined clinical criteria to provide more individualised and accurate medication. In this sense, telemedicine also contributes to objective, measurable improvements in citizens’ health and welfare.

Welfare professionals in telemedicine such as doctors, nurses, dietitians, social workers and psychologists are pushed to develop a digitalised version – a 2.0 version – of their professional expertise. For the welfare professionals, version 2.0 represents a necessary adaptation to the digital society, and thereby, results in an upskilling of their work.

Data analysts are acknowledged as a key profession in the digital welfare state’s e-service provision. Accordingly, they gain the same recognition as the welfare professionals did in the 1970s.

Big Tech, who facilitate public e-services, are given a lucrative market (this will be discussed in greater detail below).

To sum up: win-win situations are created for all the main stakeholders in e-service provision.

Like many other e-services, telemedicine for COPD and diabetes patients also has downsides. Hospitals, doctors and nurses may, for example, demand that patients provide health information/data before e-consultations by filling in online charts and formulas. They may also demand that patients perform health tests themselves and report the results online. From the patient’s point of view, this is not only time consuming but also requires them to take on a significant amount of responsibility for the public health e-services they receive. Meanwhile, seen from the health sector’s perspective, the involvement of patients increases productivity and, consequently, lowers the costs of the delivered services. It is crucial that a fair balance is reached – in the short and long term – between how much increased welfare patients receive and the time they spend as well as the responsibility they take on when receiving e-services. It must be ensured that governments do not employ e-services only to increase productivity and patients’ responsibility. Furthermore, it is important to note that the welfare professionals, who are integrated in the parliamentary chain of command as civil servants, are the governments’ experts in the daily operations of public welfare institutions. Consequently, they must ensure that e-services deliver as promised, meaning that they result in significantly increased net welfare for citizens as end-users and taxpayers. The significance of this issue is highlighted by cases in Australia, the Netherlands, and the US, where e-services have been deployed to control vulnerable citizens instead of providing help (Yeung, 2022).

The fact that e-services have clear advantages as well as disadvantages makes it even more important to understand how to determine whether or not digital welfare states deliver as promised.[6]

Data-driven provision of e-services and documentation of win-win situations

The digitalisation of the largest welfare institution in Norway[7] – The Norwegian Work and Welfare Administration (NAV) – resulted in win-win situations for citizens, welfare professionals, data analysts, managers, and society(Breit et al., 2019, pp. 149-169). NAV provided and continues to provide public welfare service to vulnerable citizens outside of, or at the margins of, the Norwegian labour market. Unfortunately, the NAV case does not show how these win-win situations were achieved.

However, a case-study of data-driven management and health service provision at a hospital in Denmark, conducted some years back by myself, shows some basic principles needed to create win-win situations for key stakeholders, and to documents these with hard data/statistics (Pedersen, 2025, pp. 57-69). The principles can be summarised as follows:

Hospital management formulated Key Performance Indicators (KPIs) for the hospital’s wards and units that, if achieved, made it possible for the hospital to fulfil its own KPIs formulated by the Regional Council. The hospital’s KPIs focused on 1. Citizen/patient satisfaction, 2. Delivery of health services, 3. Quality of services, 4. Acceptable levels for employees’ and managers’ paid leave due to illness and stress, and 5. Productivity.

At the hospital, several tools[8] were implemented to organise processes in wards and units with the aim of fulfilling KPIs. Based on statistics, data analysts showed in weekly reports whether KPIs were met. If they were not met, the same tools had to be used again until KPIs had reached acceptable levels. Combined, the tools ensured that coordinated and dialogue-based inputs from all professions and hierarchical levels were put forward on a weekly basis, with the aim of improving the provision of services and fulfilling KPIs. The tools also ensured that the inputs used for improvements were formulated as working hypotheses. After these had been tested in daily operations, the data analysts’ weekly reports showed whether the suggested improvement had helped fulfil the KPIs. The reports were available to all employees, used as feedback by all professions, and subsequently, as new inputs for learning how to improve services.

A particularly interesting feature of data-driven management and provision of health services at the case hospital was the ability to foster a creative co-existence between and co-development of all professional groups. This was largely because the tools enabled qualitative and quantitative inputs to service improvements to be combined. As such, the tools helped compensate for important limitations associated with welfare professionals’ practical expertise and data analysts’ digital algorithms.

The limitations related to welfare professionals’ practical expertise concern an overall tendency to overestimate the positive effects of provided services, the uncertainty involved in decision-making processes, and the promotion of self-interests. The limitations of data analysts’ algorithms, on the other hand, are related to a fundamental inability to take an individual’s actual contexts into account, and a tendency to exercise control over users by means of algorithmic authority (for further details, please see Pedersen, 2025, pp. 46-50). It is, of course, essential that KPIs deployed in data-driven management and service provision are realistic. To ensure this, Kahneman suggests that the Reference Class Forecasting (RCF) method/tool is applied (Kahneman, 2012, pp. 251 – 253). This method/tool ensures that KPIs are founded in hard data/statistics.

To sum up: in public welfare service provision, all relevant variables/parameters for stakeholders can be formulated as KPIs at the level of individuals, managers, professions, institutions, municipal regions, and governments. Furthermore, tools and knowledge already exist that can support the continuous achievements and possible improvements of KPIs and document this through hard data/statistics.

The data-driven provision of health services presented above can, without any major problems, be applied to most public welfare institutions as well as private companies that deliver public welfare services to citizens on contracts. Moreover, it can also be applied to e-service provision. See please Pedersen, 2025, pp. 67-69 for an illustrative example regarding telemedicine for COPD and diabetes patients.

In light of this, governments have the opportunity to develop and implement a political-administrative tool which ensures that e-services result in fair, balanced, and documented win-win situations for all stakeholders, including the most vulnerable citizens. By doing so, governments would be able to document the additional welfare that the digital welfare state promises, and as such legitimise public welfare e-services. However, most governments have not seized this opportunity. On the contrary, most governments have implemented a version 2.0 of NPM which is based on public procurement market-oriented purchasing – on shopping e-services. Unfortunately, this strategy makes welfare states’ e-service provision increasingly dependent on Big Tech and this has, at least, two negative consequences. First, governments risk losing control over e-service provision. Second, governments may become increasingly vulnerable to geopolitical changes.

New Public Management 2.0 and dependence on Big Tech

As described earlier, NPM version 2.0 emerged out of the digitalisation of the seven elements/building blocks in public service provision (Pedersen, 2025, pp. 75-78). Version 1.0 and 2.0 of NPM fundamentally serve the same purpose, which is to make public welfare service provision better, cheaper, and faster. While the focus of version 1.0 was to introduce privatisation, markets and private business management concepts into public service provision, version 2.0 is focused on public procurement market-oriented purchasing. Welfare states pay Big Tech to facilitate public e-service provision and, as a result, public e-service provision has become heavily dependent on those same companies.

In Europe, the situation may be described as follows: “European countries are strongly digitally dependent on foreign countries, namely for over 80% of digital products, services, infrastructures, and intellectual property. This dependency is predominantly on the USA and has been growing over the past 10 years (Cerre, 2022, p. 15).”

The ‘digital stack’ is the best indicator of how dependent e-service provision is on Big Tech. There are several definitions of the digital stack. One is given by Dijck (2025a): think of a pyramid with three layers. The bottom layer is digital infrastructures such as personal computers, datacenters, computer chips, cables, etc. The middle layer is gatekeeping platforms such as app-stores, social media, mail-systems, and personal identification systems. The top layer is software interfaces, such as apps designed to enable mobility, ID, fitness, education and health tracking, and many more. All three layers are integrated and together form the digital stack. Big Tech have their own closed and independent digital stacks/digital ecosystems (Dijck, 2025b). For Google, for example, layer 1 contains Google Cloud, Chrome laptops, data centres, satellites, etc., layer 2 contains YouTube, Google Maps, Google Search etc., while layer 3 contains Google Classroom, Google Scholar, Fit Tracker, Navigation etc. In the case of Microsoft, layer 1 contains datacenters, laptops, administrative systems, and tablets, layer 2 contains Windows, Bing, GPT-4, CoPilot, while layer 3 contains ChatGPT (OpenAI), LinkedIn, and administrative systems. The three layers are in both cases vertically integrated to form closed, independent digital services packages controlled by Google or Microsoft respectively.

When it comes to public e-services, Denmark represents a case of successful digital transformation. A recently published Danish report, Digital suverænitet i den offentlige sektor [Digital Sovereignty in the Public Sector] (Digitaliseringsstyrelsen [Agency for Digital Government], 2026), shows the following:

The Danish public sector’s digital stack is defined as: Layer 1. Supply chains, Layer 2. Digital infrastructures, and Layer 3. Digital solutions (Digitaliseringsstyrelsen, 2026, p. 7). The elements in layers 2 and 3 are mapped and visualised, and each visualised element is linked to suppliers. This shows that the Danish public sector is dependent on the American Big Tech to a very high degree (Digitaliseringsstyrelsen, 2026, p. 8). For the Danish public sector, this generates four major challenges: 1. A high degree of dependence upon strong suppliers and a lack of competition among suppliers, 2. A lack of control and transparency regarding the public sector’s citizen data, 3. Limited control and limited management of digital solutions, and 4. Vulnerability to digital supply chains (Digitaliseringsstyrelsen, 2026, p. 12).

According to the report, there do exist alternatives to software provided by American Big Tech. Unfortunately, this is not the case when it comes to infrastructures, and this will make it difficult for Denmark to remedy the four listed challenges in the near future. As a small country of approximately 6 million inhabitants, the only way for Denmark to gain some level of independence from American Big Tech will be by cooperating with other EU countries, and it will take years to achieve. Only one initiative has, at present, been taken to create a European stack – EuroStack. However, with no element having materialised so far, the initiative remains merely a vision for the future (Digitaliseringsstyrelsen, 2026, p. 21; Dijck 2025 c).

During the long process of achieving some level of independence from American Big Tech, it is important for governments to maintain as much control over e-service provision as possible. Considering the liberal ideology behind NPM 2.0, and the libertarian ideology driving Big Tech, governments would do well to apply principal-agent contracts so as to maintain control over public e-service provision. Furthermore, they should include in these contracts the data-driven provision of welfare services, as described earlier, based on documented win-win situations for all stakeholders, to ensure that e-services deliver what they promise to citizens as both end-users and taxpayers.

Although principal-agent contracts – for several reasons, not least in relation to vulnerable citizens – are urgently needed, in most cases these contracts are not deployed, as Yeung (2022) has shown in detail. Big Tech will, inevitably, see principal-agent contracts of this kind as unwanted regulation. Yet, if governments do not regulate, they run the risk of losing control over e-service provision in ways that will ultimately render it dysfunctional. The next paragraph provides an analytical framework for showing why that is.

Three historical models of public service provision

The following analytical framework consists of three historical models of public welfare service provision (Pedersen, 2025, pp. 27-32).

Model A is the original model and had its heyday in the 1970s. In model A, welfare professionals, on the basis of their practical expertise^[9], make all the important decisions related to the seven main elements/building blocks of public service provision. This applies especially to the third element/building block – final decisions about which services citizens should receive. Another key feature of model A is that both citizens and welfare professionals are physically present within welfare institutions such as schools, hospitals, and homes for elderly and disabled citizens. Importantly, in model A digital tools do not play any significant role in public service provision – and Big Tech, of course, did not yet exist.

Model B is the current model and a digitalised version of model A. In this model, welfare professionals’ practical expertise is increasingly framed digitally with the aim of making public service provision cheaper, better, and smarter. In the spirit of Bovens and Zouridis’ (2002) analyses, model B can be characterised as an advanced version of screen-level bureaucracy on the path to becoming a system-level bureaucracy.[10] This has three important consequences: 1. Welfare professionals are increasingly tied to their digital work screens, 2. Their work increasingly comes to resemble digital assembly lines and is on the path to becoming fully automated, and 3. Citizens are to a still higher degree kept physically away from expensive public welfare institutions.^[11] In model B, welfare professionals are integrated into the parliamentary chain of command and are, therefore, accountable to governments, civil servants, and citizens. In other words, in the last instance, it is welfare professionals who have the full and final responsibility for the provision of welfare services. In this way, the impression that services are provided by people (welfare professionals) to people (citizens) is maintained in model B, even though welfare professionals’ influence upon service provision has decreased significantly precisely because of model B. The main reason for this decrease is that data analysts install more and more advanced algorithm-based programs on welfare professionals’ work-computers which means that Big Tech, as developers and providers of these programs, gain more and more control over welfare professionals’ work and, consequently, over public welfare service provision.

Model C is a future model. In model C, data analysts’ algorithms and tools and Big Tech have fully replaced welfare professionals and their practical expertise. Welfare professionals’ work is fully automated and e-services, most of which are DIY e-services, dominate public service provision.

In figure 1 below, the three models are illustrated.

Figure 1. The three historical models of public welfare service provision

In Model A – the original model – the blue arrows illustrate that citizens show up physically at welfare institutions (schools, hospitals, etc.). The physical welfare institutions are illustrated by the circle. In the institutions, citizens meet face to face with welfare professionals (teachers, nurses, doctors, etc.). This is illustrated by the circle’s black colour. Within the institutions, welfare professionals have full autonomy to take final decisions regarding which services citizens should receive based on their practical expertise.

In Model B – the current model – the green arrows and the white part of the circle illustrate the digitalisation of the seven main elements/building blocks in public welfare service provision. The green arrows illustrate that citizens use e-services. The white part of the circle illustrates the digitalisation of the seven main elements/building blocks and consequently, of elements in the welfare professionals’ work. The black part of the circle illustrates that welfare professionals still take most of the final decisions in the welfare service provision.

In Model C – the future model – Big Tech’s data analysts’ digital tools and algorithms have fully replaced welfare professionals’ practical expertise in the service provision. Out of the digital transformation, a national welfare state without welfare professionals has emerged.

In Model B, Big Tech enjoy a privileged position in the form of a lucrative market and limited responsibility for public e-services. If NPM 2.0 and model C are combined, however, these companies will be forced to take on full responsibility for daily e-service operations at many hospitals, schools, and other welfare institutions, as well as for the administrative tasks of public bureaucracies. This would represent a potential problem for several reasons. Big Tech are not private health and education corporations. Rather, they merely facilitate e-services provided by hospitals, schools, and public bureaucracies. Furthermore, these companies have no interest in getting entangled in the tasks and responsibilities of public bureaucracies or in being held accountable for the provision of welfare e-services. Not least because that will result in public and political demands to make their digital ecosystems transparent, which runs the risk of trade secrets such as algorithms and business concepts being made public. Seen in this perspective, Big Tech will likely opt for model B and, via NPM 2.0, try to push it as close as possible to model C. A prospective that will very possibly lead to a dysfunctional provision of e-services.

More than two decades ago, this was made apparent by Bovens and Zouridis (2002) based on their analyses of two cases of full automation. The more welfare professionals’ work resembled digital assembly lines and the closer it moved towards full automation, the less important practical expertise – such as individual experiences, tacit knowledge, professional intuition, expert judgements, and knowledge of individual citizens’ socio-economic everyday life contexts – became. Consequently, the job satisfaction of welfare professionals decreased significantly. This will most likely also be the case if Big Tech push model B as close as possible to model C.

Furthermore, at some point on the path towards model C, it will no longer make sense for welfare professionals to take on the full responsibility for public service provision. This means that a situation may emerge where neither Big Tech nor the welfare professionals will accept full responsibility for e-service provision. This will leave societies and governments in a difficult place. They risk losing control over e-service provision to such a degree that it becomes impossible for welfare states to fulfil their historical task – to help the vulnerable citizens who need help the most. The result will be a dysfunctional version of service provision caused, essentially, by the digital transformation.

At present, US geopolitical changes add a further twist to governments’ potential loss of control over public e-service provision. National digital sovereignty allows governments to provide citizens with e-services without interference from other countries, and since American Big Tech dominate digital stacks in Europe, current US geopolitical changes could potentially interfere with European countries’ national digital sovereignty. The USA has laws that, for security reasons, give the country’s administration access to data retrieved by Big Tech, and as such weakens compliance with important elements of the EU’s GDPR (The Danish Medical Association, 2026, pp. 566-69). Moreover, the US government has threatened the EU with tariffs on certain goods and services, if the EU regulates American Big Tech. In this way, changes in US geopolitics presents a real threat of weakening European countries’ national digital sovereignty.

How can governments solve the challenges related to the loss of control over public e-service provision? One straightforward strategy would be to substitute American Big Tech with national counterparts, where possible. Only a few years ago, such a strategy would have been considered an outdated policy, belonging to a time before NPM 1.0 and completely unrealistic. This year, the EU member states have agreed to start a process aimed at becoming independent from American Big Tech and, thereby, separating themselves from some of the negative consequences of the current changes in US geopolitics. One important step in this process is the European Commission’s European Technological Sovereignty Package (European Union, June 2026).

Society, institutions, and public welfare e-service provision

According to The Institutional Logics Perspective (Thornton et al., 2012, pp. 73, 170), most societies contain the following seven institutions: Religion, Family, Community, State, Market, Profession, and Corporation. Just as with the seven main elements/building blocks in public service provision, these seven institutions have been digitalised, meaning that the individual institutions themselves and the interactions and linkages between them have been structurally reshaped due to digitalisation. The metatheory of the Institutional Logics Perspective explains how the seven institutions can achieve both stability and change; how they can be changed structurally due of digitalisation, and remain important institutions within the digital society. The theory also explains how and why institutional logics from different time periods can coexist within institutions themselves (Thornton et al., 2012, p. 169). For example, how and why the logic of welfare professionals’ practical expertise and the logic of data analysts’ algorithms – and consequently the provision of digital and analogue welfare services – can coexist within welfare states, and how and why the logic of users having basic digital skills and the logic of the excluded group can coexist within the institutions of Family and Community. A new institutional logic may, of course, also generate competition. Currently, the logic of data analysts’ algorithms, for instance, competes with the logic of welfare professionals’ practical expertise in public welfare service provision within model B, which weakens the latter. All three cases of coexistence between institutional logics can, however, all be defined as positive.

In the first case, the two institutional logics’ coexistence makes it possible to offer welfare professionals’ ‘hand-held’ services to the excluded group. In the second case, coexistence results in access to digital help within the institutions of Family (households) and Community (social networks), meaning access to persons who can help e-service non-users and the greys. While the third case has obvious negative consequences for existing welfare professionals, e-service provision generates high demand for ICT experts which means the generation of new jobs. These are jobs that young women especially can access, being an untapped resource in terms of meeting the high demand for ICT experts and thereby helping to fully utilise the potential of e-services (OECD, 2024, p.10). In other words, public welfare e-service provision can, in this case, help address gender inequality not only on the labour market, but also in the context of education and income levels.

The three cases show why, during the digital transformation, it is important for governments to influence the structural changes within institutions, as well as the interactions and linkages between institutions, in order to improve e-service users’ social frameworks and socio-economic living conditions. This will be essential for making the digital transformation successful.

Conclusion

If the EU member states – that is, the majority of countries in Europe – are placed on a continuum, some will experience that their digitalised welfare states cannot fulfil welfare states historical task of providing help to the citizens who need it the most. The main reason for this is a combination of relatively low levels of digital transformation and high levels of excluded and greys. At the opposite end of the continuum, some countries will experience that their digitalised welfare states offer by far the most citizens extra welfare, and that their levels of excluded and greys are manageable. Among these, a cluster of countries can be located that have all performed well during the digital transformation.

Looking ahead to the near future, the first group of countries will experience difficulties when it comes to legitimising their digitalised welfare states. Furthermore, they are at risk of being left behind when it comes to the provision of public welfare e-services and managing the digital transformation. To remedy this problem, help and support from the EU is urgently needed.

The cluster countries are all small countries with well-established and relatively large welfare states. Their main challenges within the near future will be to avoid too much dependency on American Big Tech, to avoid being too vulnerable to US geopolitical changes, and to avoid potential dysfunction in both e-service and traditional service provision. To overcome the first two challenges, help and support from the EU will be crucial. Independence from US geopolitics and Big Tech can only be achieved via joint efforts with the EU. Crucial to solving the third challenge, which the countries share with most other countries, is to establish a creative cooperation between welfare professionals and data analysts as shown in relation to data-driven provision of public welfare services. This is illustrated in figure 2 below.

Figure 2. Creative cooperation between welfare professionals and data analysts in public welfare service.

Figure 2 shows that welfare professionals and data analysts have to coexist within welfare institutions to ensure that the excluded, the greys, and the winners have access to both e-services and traditional services to the necessary extent. This is illustrated with the black and white colour in the circle, and the blue and green arrows in the figure. To achieve that citizens get the right services, societies must ensure that welfare professionals’ practical expertise and data analysts’ digital tool and algorithms are combined within welfare institutions’ daily operations based on dialogues which result in mutual framing of practical expertise and digital tools and algorithms. This is illustrated with the circle within the circle in figure 2. Not only welfare professionals’ practical expertise has to be framed by digital tools and algorithms to achieve cheaper, better, and smarter welfare services. Data analysts’ digital tools and algorithms have also to be framed by welfare professionals’ practical expertise to achieve that the excluded and the greys are provided with welfare services which make it possible for national welfare states to fulfill their historical task to help vulnerable citizens who need help the most. This is the main reason why a well-functioning Model B should be preferred over Model C.

References

Agency for Digital Government, Denmark, [Digitaliseringsstyrelsen] (2026). Hvem oplever udfordringer ved det digitale? [Who needs support with e-services]

Agency for Digital Government, Denmark, [Digitaliseringsstyrelsen] (2026). Principper for digital inclusion [Principals for digital inclusion] / https://digst.dk/digital-inklusion/principper-for-digital-inklusion/

Agency for Digital Government, Denmark, [Digitaliseringsstyrelsen] (2026). Digital suverænitet i den offentlige sektor, Rapport [Digital sovereignity in the public sector, report]

Bovens, M., & Zouridis, S. (2002). From street-level to system-level bureaucracies: How information and communication technology is transforming administrative discretion and constitutional control. Public Administration Review, 62(2), 174-84.

Breit, E., Egeland, C., & Løberg, I. (2019). Cyborg bureaucracy: Frontline work in digitalized labour and welfare services. In Pedersen, J.S. & Wilkinson, A. (Eds), Big Data. Promise, Application, and Pitfalls (pp. 149-69). Edward Elgar Publishing.

Cerre, Centre on Regulation in Europe, Report, Paul Timmers, December (2022). Digital Industrial Policy for Europe.

Danish Medical Association, no. 7, March 30, 2026, pp. 566-69.

Dijck, J.v. (2025, a, b, c). How to achieve digital sovereignty in Europe? Keynote speech at ZeMKI, Bremen University, October 23. YouTube a) time 13 min, b) time 15 min c) time 44 min.

Dijck, J.v., Poell, T., & Wall, M.D. (2018). The Platform Society. Oxford University Press.

European Commission. European Technological Sovereignty Package. European Union, June 2026.

EU, Digital Economy and Society Indicators – DESI – index – https://digital-decade-desi.digital-strategy.ec.europa.eu/datasets/desi/charts

Gomes, A., & Dias, J. G. (2025). Digital Divide in the European Union: A Typology of EU Citizens. Soc Indic Res 176, 2025, pp. 149-172.

Kahneman, D. (2012). Thinking – Fast and slow. Penguin Books.

Minimum Digital Living Standard (MDLS), at Loughborough University and University of Liverpool, UK (2025). About | MDLS Minimum Digital Living Standard for UK Households in 2025: Briefing Paper (June 2025) .

OECD Digital Economy Outlook (Volume 2) (2024). Paris, OECD Publishing.

Pedersen, J.S. (2025). Digital Transformation and Public Welfare Services: The Opportunity, the Challenge, and the Wildcard. Edward Elgar Publishing.

Susskind, R., & Susskind, D. (2017). The Future of the Professions – how technology will transform the work of human experts. Oxford University Press.

Thornton, P., Ocasio, W., & Lounsbury, M. (2012). The Institutional Logics Perspective: A New Approach to Culture, Structure and Process. Oxford University Press

Yan, Z. (2026). Review of Digital Transformation and Public Welfare Services: The Opportunity, the Challenge, and the Wildcard. Social Policy & Administration, Volume 60, issue 4, 0:1-2.

Yeung, K. (2022). The new public analytics as an emerging paradigm in public sector administration. Tilburg Law Review, 27(2), 1-32.

Notes

A short summary of the book is provided in a review by Yan (2026).
Do-It-Yourself e-services. ¨
The Digital Economy and Society Indicators – DESI – index https://digital-decade-desi.digital-strategy.ec.europa.eu/datasets/desi/charts . The data applied in the following are retrieved online from this database.
At least basic digital skills are defined according to the following 5 areas: 1. Information and data literacy, 2. Communication and collaboration, 3. Digital content creation, 4. Safety, 5. Problem solving. The level of skills in each of these categories together determine whether citizens can utilise the majority of e-services in digitalised welfare states. See DESI index/ the indicator: At least basic digital skills.
Data for the greys are uncertain because it is difficult to know for how long citizens having health issues, disabilities etc. can use/cannot use e-services.
In this context it is important to note that the rapid development of advanced, AI-based e-services means that the environment should be considered as a new, additional key stakeholder in e-service provision. This is because AI-based e-services are highly energy-intensive and may therefore challenge governments’ green policies and climate objectives.
Like all Scandinavian countries, Norway has a universal welfare state.
The most important tools were: 1. Identically formatted whiteboards in wards and units, used to keep track of fulfilled KPIs’ for each ward and unit, and initiatives to fulfill KPIs which, according to the weekly reports, had not been met. 2. Short, weekly stand-up meetings with a fixed agenda where representatives from all professions and hierarchical levels participated. 3. The PDSA (Plan-Do-Study-Act) method/tool used to support the organisational processes in order to meet KPIs.
Welfare professionals’ practical expertise is a “(…) complex combination of formal knowledge, know-how, expertise, experience, and skills (…)” (Susskind & Susskind, 2017, p. 41).
An advanced version of a screen-level bureaucracy on the path to becoming a system-level bureaucracy is an increasingly digital automation of welfare professionals’ work – practical expertise – in public welfare service provision.
Citizens are kept away from expensive welfare institutions because the human element – the welfare professionals – in public welfare services is decreasing. See Bovens & Zouridis, 2002, p. 180 – Table 1.