Abstract
Due to digitalisation, automation, and the use of artificial intelligence in policy implementation, public administrations are adapting their workflows and organisational structures. We explore the impact of the digital transformation on the nature of tasks as well as the subsequent (potentially) changing demand for competences and job profiles in public executive agencies, based on a comparative case study and semi-structured interviews with over 60 public servants from three key administrative services in Germany – electronic tax returns, internet-based vehicle registrations, and digital building applications. The findings reveal less complex tasks being partially substituted and core discretionary tasks being standardised. We highlight the emergence of “work for digitalisation” that often remains “invisible”, including “compensating work” (scanning, IT support), “connecting work” (software maintenance) and “cleaning work” (validating data), resulting in an overall increased workload. Additionally, there is a need for role-specific new competences and upskilling of middle-level public servants. Our results indicate a potential substitution of traditional administrative job profiles but a growing demand for highly skilled, academically trained public servants. The findings therefore challenge the notion that digitalisation and automation threaten human labour or effectively counteract staff shortages. Rather, (potential) benefits, burdens, and job perspectives are distributed unevenly across job profiles.
Key points for practitioners
Digitalisation entails many “invisible” tasks necessary for any kind of digitalised administration, like scanning, maintenance, software administration, support and data validation. These can be further differentiated into connecting, compensating and cleaning work. Digital government is rarely met with sufficient organisational adjustment at the execution level. Instead, the work for digitalisation is often attributed to existing staff and conducted “on the side”. Additionally, an increased workload is observed in all administrations. Changes to the task complexity and the discretion of public servants are observed across the different services but are unevenly distributed within the public workforce, placing a particular pressure on the so-called middle-level staff.
Introduction
The digitalisation of public services has become a key issue of public sector modernisation (Coursey & Norris, 2008; Dunleavy et al., 2006; Schwab et al., 2019). The digital transformation of public administration can be considered an important step towards making service delivery more citizen-centred and user-oriented. In many countries, constitutional and legal frames have been adopted to put digital public service delivery, so-called digital government, on a firmer footing.
In this context, digitalisation is understood as the adoption of digital technologies in administrative processes, potentially resulting in changes in organisational structures, processes, and people (Tangi et al., 2021). This includes the provision of public services to citizens following online applications and digital forms of inquiry as well as digitalisation and automation within the organization – ranging from process automation over decision-support systems with varying degrees of human involvement to fully automated systems without any human involvement (Denk et al., 2022). The term “citizen self-service” (Lindgren et al., 2019; Madsen et al., 2022) further points towards a development where citizen-to-government interactions are increasingly, and often mandatorily, conducted through digital channels. Here, citizen self-service is associated with shifting responsibilities and work tasks from case workers to citizens (Madsen et al., 2022). From a normative viewpoint, there are different perspectives on the impacts of digitalisation and automation, ranging from techno-optimist ‘utopian’ visions of more efficient and effective public administration to techno-pessimist ‘dystopian’ images of surveillance and control (Plesner et al., 2018). On the one hand, expectations include improvements, such as greater efficiency and service quality, reduction of error rates, and time savings (Kuhlmann & Bogumil, 2021; Mergel et al., 2019; Vuori et al., 2019), and thus, greater citizen satisfaction (Heuberger & Schwab, 2021). On the other hand, digitalisation might also bring about new problems and unintended negative consequences like work intensification, overload, and stress (Bogumil et al., 2022; Fleischer & Wanckel, 2023; Kuhlmann & Heuberger, 2023). Regarding the street-level bureaucracy, some expect “street-level bureaucrats” to turn into “screen level-” or “system level-bureaucrats” (Bovens & Zouridis, 2002). Notwithstanding, there remains limited empirical research on the internal effects of the introduction of citizen self-services and automation of case processing with regard to public servants’ workplace and tasks (Fischer et al., 2021, p. 17) as well as case workers’ discretion, power asymmetries, re-skilling, and the changing roles at the street-level (Lindgren et al., 2019, p. 433). Research concerning the changing nature of tasks and the employees engaged in these activities in public administration's executive agencies (what Bovens & Zouridis (2002) call “decision-making factories”) is limited, particularly in non-welfare service contexts (for studies on the latter, see e.g. Breit et al., 2021). This is why we pose the research question: How do digitalisation and automation impact tasks and staff composition in executive agencies?
Our research concerns both the micro-level and meso-level effects on employees. Regarding the micro-level, we focus on the immediate changes in task fulfilment. This task dimension revolves around whether public servants perform more or less complex tasks (Autor et al., 2003; Bullock, 2019), have less discretionary power (Bovens & Zouridis, 2002; Buffat, 2015) but higher workload due to additional, “invisible” work for digitalisation (Justesen & Plesner, 2024; Ruiz Ben, 2021). In addition, the staff dimension concerns (possible) meso-level effects on competence requirements and the staff composition. We investigate to what extent the demand for competences and job profiles changes today and in the future (Distel et al., 2019; Mergel et al., 2019) and how this results in potential restratification within public personnel (Noordegraaf, 2016).
Empirically, our study is set in Germany, which is considered a late bloomer in the digitalisation of its public services, falling behind far-reaching hopes and political announcements (Bogumil et al., 2022; Gräfe et al., 2024; Heuberger, 2022; Kuhlmann & Bogumil, 2021; Kuhlmann & Heuberger, 2023). In recent years, however, there has been political pressure to act and adapt both the delivery of public services to citizens as well as internal administrative processes, including process automation and automated data exchange. Our empirical analysis provides new insights on the transformation at the street-level in executive agencies dealing with inspection and taxation. The findings show that the nature of work has indeed already changed, while structural changes in the composition of the workforce have not yet materialised but are anticipated.
The article is structured as follows. First, we review the literature on the impact of digitalisation and automation on work and the workforce from different research strands such as labour economics, information systems, and public administration research (Section 2). Then, we present our methods and data, and briefly describe our cases: internet-based vehicle registration, digital building application, and electronic tax return (Section 3). The subsequent section analyses and compares the findings (Section 4), followed by a discussion (Section 5). Finally, we draw conclusions and provide an outlook on possible avenues for future research (Section 6).
Literature review
Digitalisation and automation 1 are expected to affect and reshape the work of street-level bureaucrats, i.e. those public servants carrying out services for citizens and thus implementing policies (Bovens & Zouridis, 2002; Busch & Henriksen, 2018; Høybye-Mortensen, 2019; Hupe, 2022; Lindgren et al., 2019; Lipsky, 2010; Plesner et al., 2018; Pollitt, 2011). Most authors follow the ‘deterministic approach’ (Høybye-Mortensen, 2019), assuming that digital technologies are taking over tasks previously carried out by public servants, which will lead to a reduction of discretion and ultimately downsizing or redundancy of staff (Buffat, 2015; Hupe, 2022; Snellen, 2002). However, going beyond these “grand narratives” (Plesner et al., 2018, p. 1178) around ‘utopian’ or ‘dystopian’ visions of the public sector’s digital transformation, we concentrate on how “the nature of the tasks” (Pollitt, 2011, p. 387) is changed, leading to a restratification of the personnel composition as competence requirements and the demand for particular job profiles change, thus “changing public officials themselves” (Pollitt, 2011, p. 388). In the following, we built on research from the fields of labour economics, organization studies, information systems and public administration.
Changing nature of the tasks
Previous research has focused on the potential substitution and complementarity of tasks between humans and technology, depending on the complexity of tasks (Autor et al., 2003; Bullock, 2019; Frey & Osborne, 2017). Less complex tasks that “rarely deviate from normal procedures” (Bullock, 2019, p. 756) and “can be accomplished by following explicit rules” (Autor et al., 2003, p. 1280) are thought to be replaced. Their algorithmic nature allows for the application of (partial) automation, “reducing knowledge demands of the task in order to increase the speed” of task fulfilment (Zuboff, 1988, p. 133). In contrast, complex tasks involving communication and problem-solving are considered to be complemented by technological advances (Autor et al., 2003). More recently, however, advances in artificial intelligence have paved the way for substitution of complex cognitive tasks like pattern recognition (Bullock, 2019; Frey & Osborne, 2017). In the context of public service delivery, research finds tendencies of discontinuation and emergence of specific tasks. More mundane tasks not involving specialist knowledge no longer require human interaction (Plesner et al., 2018; Pollitt, 2011), while other, more complex tasks, like reviewing or those requiring technical expertise, emerge (Bovens & Zouridis, 2002).
To consider the broader implications of digitalisation for work and employment, Ruiz Ben (2021) developed a typology of digital work forms. Firstly, she pointed to “digitalised work”, i.e. previously manual, non-digital tasks that are now standardised and partially substituted by digital tools or automation, especially the actual assessment and decision-making process. Secondly, she emphasised “work for digitalisation” which is a prerequisite for any kind of digitalised work but often remains “invisible” (Ruiz Ben, 2021). Justesen & Plesner (2024) further unpack this unnoticed and undervalued “invisible digi-work”, identifying three types: “connecting work”, which involves maintenance and repair activities to ensure connectivity (ibid., p. 10); “compensating work”, which encompasses “activities that compensate for errors, limitations, or other shortcomings related to digital technologies that promise more than they can deliver” (ibid., p. 12), for instance, performing unexpected tasks like supporting citizens and solving IT problems (Pors, 2015); and “cleaning work”, which entails “modifying, purifying and removing [data]” (Justesen & Plesner, 2024, p. 14). However, invisible digi-work is not confined to the “core” of very visible case workers and specialists, but instead often found conducted by more “peripherical” (support) units such as the registry, or even outsourced to citizens or third parties (ibid., p. 3–4).
As a result of the changed tasks, debates have revolved around the workload of employees – with mixed results. On the one hand, digital tools are thought to increase efficiency by simplifying tasks and accelerating case processing (Vuori et al., 2019), and citizen self-services are expected to transfer some tasks to citizens, disburdening public servants. On the other hand, empirical studies found an increased pace of work, additional strain, and overload (Bogumil et al., 2022; Kuhlmann & Heuberger, 2023), related inter alia to techno-overload forcing public servants to work faster (Fleischer & Wanckel, 2023; Tarafdar et al., 2007), frequent interruptions, and information overload (Vuori et al., 2019). Only well-functioning digitalised processes have been associated with reduced workload (Schwab et al., 2019). Yet, the effect of the aforementioned mostly invisible “work for digitalisation“ on workload is not taken into account.
With regard to bureaucratic discretion – the power public servants can exercise to make decisions about (non-)actions in particular cases (Lipsky, 2010) – digitalisation and automation can have different effects: curtailment, enablement, or continuation (Buffat, 2015; Marienfeldt, 2024). According to the curtailment thesis (Buffat, 2015; Snellen, 2002), digitalisation and automation constrain street-level bureaucrats’ discretion. Based on observations of Dutch student grants and traffic regulation in the past decades, Bovens and Zouridis (2002) concluded that digitalisation has transformed large executive agencies into so-called “screen-level bureaucracies” – where technology plays a leading role and assists human decision-making – or even into “system-level bureaucracies” – where technology itself makes the final decision. Consequently, public servants are supposed to take over more complex tasks such as handling exceptional cases, errors, and complaints, assisting and giving advice to citizens, or dealing with the development and maintenance of systems (Bovens & Zouridis, 2002). Indeed, studies show that automation leads to less discretion-as-perceived (de Boer & Raaphorst, 2021). In municipal social services in Sweden that implemented a robot-process automation technology, case workers also lost the opportunity to inform citizens about benefits or to notice other needs, a task previously carried out during personal interaction in the application process (Andersson et al., 2022). Other scholars consider technology to expand the room for manoeuvre (Busch et al., 2018; Meijer et al., 2021; Schuppan, 2016), supporting the enablement thesis, which stresses the potential of technology as action resource (Buffat, 2015). Furthermore, the continuation thesis (Marienfeldt, 2024) claims that – if digitalisation does not facilitate street-level decision-making – public servants establish workarounds to maintain their professional judgement. For example, studying the use of a risk assessment tool in Dutch border control, Dekkers et al. (2019) found that officers either use or ignore the recommendation of the digital systems and make their own final decisions.
Changing public servants themselves
On the meso-level, changes in the task dimension will imply changes for staff as well (Pollitt, 2011). This refers to alterations in competence requirements (Distel et al., 2019; Hofmann & Ogonek, 2018; Hunnius et al., 2015; Mergel et al., 2019), but also a reconfiguration of professional identities (Plesner et al., 2018) and restratification within the professional field (Noordegraaf, 2016). These changes will not necessarily occur immediately but are anticipated in the medium- or long-term as consequences of the changed tasks.
First, digitalised and automated public service delivery requires new competences – i.e. work-related knowledge, skills, abilities as well as personality traits of employees to carry out their tasks (Distel et al., 2019; Hunnius et al., 2015). Previous research concluded that digitalisation requires not only technical competences, for instance, IT skills, technology management and assessment, as well as analytical skills, but also change management, manoeuvring in the political-administrative environment, and social competences (Distel et al., 2019; Ogonek et al., 2020). Mankevich et al. (2023) showed for the Swedish public sector that the demand for digital competences as described in job postings has increased significantly from 2006 to 2020. Thereby, the authors demonstrated alterations of job profiles in the public sector. However, the competence debate often remains very generic. To distinguish between the competence requirements for particular job profiles, it is necessary to consider different roles of public servants, for instance, project managers designing IT in contrast to case workers using IT, as each role provides a distinct set of competence requirements (Ogonek et al., 2020).
Second, accumulating the effects on individual job profiles, restratification within public personnel can be expected. Noordegraaf has argued that the professional field becomes more divided “between those who perform routine tasks and those who perform innovative and risky tasks” (Noordegraaf, 2016, p. 795). On the one hand, digitalisation and automation cause simplification and de-specialisation (Pors, 2015), leading to the disappearance of routine job profiles (Frey & Osborne, 2017). With regard to the substitution potential of occupations, Zanker (2019) calculated for the German public administration that vocational skilled professionals, like middle-level administrative staff, case workers and clerks, are most affected by potential substitution (36.2%). Also affected are unskilled and semiskilled employees, such as supportive-level general administrative staff, who mainly perform supportive or manual tasks (29.6%), as well as highly skilled, often academic specialists with substantial discretion in executing professional tasks (27.4%). A minor impact is expected on highly skilled experts in leadership, management, research, and development (15.5%). Notably, administrative staff, who constitute the backbone of most street-level bureaucracies carrying out standardised procedures of low complexity, are particularly under pressure from digitalisation (Bovens & Zouridis, 2002). On the other hand, further specialisation – either in terms of professional expertise or with regard to the development and maintenance of digital technologies – is often required, which can lead to higher qualification demands on employees (Giritli Nygren et al., 2013; Pollitt, 2011).
In summary, the literature proposes different directions of digitalisation-induced changes in tasks and staff in public administration. However, previous research has mostly focused on the macro-perspective, showcasing tendencies in public administration as a whole but thereby overlooking the particularities that arise for public servants within their respective administrative settings. To investigate how digitalisation and automation impact work and the workforce in executive agencies, we conducted a comparative case study of three selected administrative services in Germany.
Methods and data
Given the broad range of different services and administrations – particularly in a federal context such as Germany – it is necessary to examine the impacts of digitalisation and automation both in greater depth and in multiple settings. A case study design is recommended as it facilitates an in-depth study of single units enabling a comprehensive understanding of a larger class of similar units by investigation of underlying covariate relations (Gerring, 2004).
The study was designed in a hierarchical manner as we examined and compared three administrative domains as well as several agencies within those domains, allowing for more in-depth findings (Gerring, 2004). By purposefully selecting more advanced cases of digitalised public services, we aim for more viable findings on digital transformations. Empirically, we studied the impact of citizen self-services and the digitalisation of internal processes on public servants and their work in the following services: internet-based vehicle registration, digital building application, and electronic tax return.
To get insights into the nature of each digital service and its implications for work and the workforce, we gathered the perspectives from multiple public administrations provided by experts who experience digitalisation and automation firsthand. In total, 55 qualitative semi-structured expert interviews (Bogner et al., 2009) with 63 public servants and stakeholders were conducted from January 2021 to June 2022. The interviews had a length of 30 to 90 min and were conducted via online video tools or phone calls, others in person at the workplace of the interviewees. As such, we chose experts from different hierarchical levels, i.e. management, case workers, staff council members, as well as digitalisation or IT officials. This gathering of multiple perspectives from different organisations in the field, and from multiple sources within the organisations, allows for greater reliability of our data on ongoing changes (Enticott, 2004). It is to note, however, that we were partly reliant on intermediaries to establish access to the organisations as well as on experts volunteering to participate in our study. Assuming a certain bias towards those with expressive convictions, i.e. above average positive or negative stances on digitalisation, we explicitly focused on their expertise on processes and change. Consequently, the guidelines for the interviews were based on the theoretical assumptions and structured into eight blocks of questions. Respondents were first asked to introduce themselves, their professional background, current job position, and relations/experiences with digitalisation. Second, we let them describe the procedures of the services they provide from application to the final notification, and their work routines. Thereby, we asked respondents to contrast the tasks required for the analogue and digitalised/automated service and internal processes. Third, respondents were asked to recapitulate the design and implementation process of the particular digital/automation tool, reflecting on, fourth, general conditions of digitalisation (resources, legal provisions), and fifth, the involvement of the staff in this change process. Then, we focused on the impact of digitalisation on, sixth, working conditions, including further questions on their perceived discretionary power and, seventh, the organisation's performance. In doing so, we asked about both the impacts that have already been experienced and those that are expected in the future. Finally, respondents were invited to tell us anything else that we had not yet addressed in the conversation. The guidelines were adjusted for the different roles of interviewees (see Table 1). In accordance with qualitative “openness”, the guidelines allowed for narrative passages to pick up new aspects relevant to the research subject (Bogner et al., 2009). Some of those new aspects have also been included in subsequent interviews.
Number of interviewees per domain.
Number of interviewees per domain.
The interviews were recorded, transcribed, and then analysed in a qualitative content analysis scheme (Mayring, 2014). We performed a computer-assisted qualitative data analysis using MaxQDA software to categorise the interview data. Concepts derived from the literature hereby served as initial code categories. Starting with a deductive coding strategy, further codes were then inductively added. In accordance with case study methodology, we investigated patterns between the cases for covariate evidence.
For the cases, we examined the internet-based vehicle registration (7 offices), the digital building application (9 authorities), and the electronic tax return (9 agencies). Said services were selected as they constitute fairly advanced examples of citizen self-services in Germany. Particularly the tax administration has been at the forefront of digitalisation for decades, while a few pioneering municipalities began to digitalise vehicle registration and building permit services as early as the mid-2000s. Additionally, with those cases, we cover different types of public services: Whereas tax returns and vehicle registration represent routine procedures with large amounts of formalised transactions (Bovens & Zouridis, 2002), building permits constitute a more individualised service. The digitalisation of tax returns is co-operatively steered and implemented at the Länder (federal states) level, in contrast to the digitalisation of vehicle registration and building application, which is conducted by local authorities, leaving a heterogeneous field of different solutions and implementation levels across the country. Our selection of single cases in each administrative domain reflects this. The findings are presented in the following section.
Internet-based vehicle registration
Vehicle registration, one of the most frequently requested public services in Germany (Bogumil et al., 2019), has been gradually digitalised and automated since 2015. As a citizen self-service, vehicle deregistration, (re-)registration and change of ownership can be applied for online by private individuals, while digital applications for corporate bodies – the most frequent users – have only become available since 2023. However, internet-based vehicle registration only partially automates the registration processes. While applicants can enter their data online and without any postal correspondence or personal audition, the administration still issues physical registration certificates and badges, ships them by post, and requires the applicant to affix the badge on the licence plate. Only deregistering vehicles can be considered fully automated, as the administrative act is issued automatically and in a paperless manner. Yet, applicants must enter the security code from the certificates and badges, thereby devaluating the plate.
The internet-based procedure is not mandatory for citizens and constitutes only an alternative access point besides traditional personal interaction at the counter. The digital application is only marginally used by citizens. Except for Berlin – reaching a seven percent share for internet-based registrations –, the studied offices provide a share far below one percent of all vehicle registrations via the internet-based request, due to a complicated application process and lacking attractiveness for the citizen.
Reviewing, printing, troubleshooting
For the employees, internet-based vehicle registration eliminates data entry and direct interaction with citizens at the counter but adds tasks of reviewing, as well as manual activities and customer support. Once an application is submitted, case workers review it – i.e. compare the data with official records and check for accuracy and plausibility – and then issue the certificates and badges. While the review is for the most part pre-assessed (or even already approved) by the digital algorithm, some offices have established additional routines of human oversight of the algorithm not only on a random but a regular basis. Moreover, the digital service requires new tasks, such as manual activities like printing, re-scanning (for electronic files), and preparation for postal delivery. Consequently, tasks have rather shifted than been eliminated, without any time savings for the administration: And in that respect, for me, the biggest shortcoming of this process is that it's actually an online application, but all of the processing at the end of the employee remains just as time-consuming as before […]. So, there's little relief there. Certainly, […] some things are simplified, but many things are not yet. (Vehicle registration, local manager)
Only ‘simple’ cases are considered less time-consuming than comparable face-to-face services as they are presented with the data and automatically checked for plausibility. In more ‘problematic’ cases, such as missing or mismatched data, the employee is required to contact the applicant – indicating increased workload per case even though task complexity is rather low. For instance, human intervention is necessary if in one database an applicant's middle name is registered, while in another database it is not. The resulting error message must then be solved by an employee. Additionally, employees now exercise support tasks for applicants struggling with the internet-based procedure, either via email or phone. In the eyes of the case workers, the mixture of computerised reviewing, issuing of papers, and communicating with customers via phone or email prevents their work from becoming more monotonous than face-to-face services (“Well, I enjoy it” (Vehicle registration, case worker)). Regarding employees' discretion, we do not find any changes given the already low discretion vehicle registration offers. The idea is the simplification of tasks in such a way […] that the personnel in the administrations are relieved of a great deal of these large-scale routine cases, of these normal processes, and that the work of the employees is improved in a certain way. (Government official)
The low utilisation has so far prevented the establishment of routines in most of the studied offices. In one of the rural offices, internet-based vehicle registration is always assigned to one single case worker, although all should be able to process it. In contrast, Berlin, which has the highest number of internet-based registrations, established new structures to cope with the novel tasks, as special offices are designated to employees who process only online cases for weeks or months. But even here, online registration does not constitute its own organisational entity, and the employees regularly rotate back to the counter.
Paving the way for more specialised case workers?
According to the interviewees, the changed tasks are not associated with a comprehensive change in competence requirements, even though the weight on reviewing tasks demands precision, the ability to grasp information, and familiarity with the computer. In contrast, some social competences necessary at the counter become less important. As working on digital applications is attributed only to a few volunteering employees, there are currently only minor changes to the personnel composition within the vehicle registration offices. Job profiles of case workers as described in job postings remain unaffected. Conversely, the personnel for managing the archives have already been substituted.
The transformation of the workforce in vehicle registrations remains limited for the time being. In the long term, a shift towards more specialised case workers is expected by government officials. We have already observed a rise of new responsibilities in vehicle registration related to, inter alia immission control. As more recent developments, such as automated interfaces for businesses (introduced in 2023), unfold, the more routine case work will be replaced by automated procedures. This entails that the remaining or newly hired case workers will be more concerned with complex cases requiring specialist competences. Because it's no longer just a matter of stamping, issuing badges or taking vehicles out of service, which are just these large-scale routine processes, but that they take care of more complicated cases, for example, they take a look at a classic car and have time to deal with special cases […]. (Government official)
Examples from other European countries show that basic tasks like electronic file maintenance and remote support, or even the whole vehicle registration might be outsourced (as in Austria, see EReg, 2019). To which extent this outlook turns into reality depends on whether Germany will eventually abolish the physical certificates and badges in favour of full automation, which is not yet to be expected.
Digital building application
The digitalisation of building applications is another ambitious project aiming to accelerate construction processes in times of housing shortages and increased demand for renewable energy. The landscape of digital building applications is characterised by numerous local or state-level solutions. By law, every building administration can open a channel for digital applications but the circumstances for such an undertaking differ between the Länder. On most occasions, some extent of interaction on paper between applicant and authorities (e.g. handing in a signed copy of the application form, delivering login data, issuing the building permit) is still legally required. But even with a lot of changes in medium, the omission of postal traffic reduces the duration of the application reviewing significantly. In the studied administrations, the share of digital applications commonly ranges between 17% and 45%. Four Länder (Baden-Württemberg, Hesse, Lower Saxony, Mecklenburg-Western Pomerania) have allowed their municipalities to make digital applications mandatory.
The same tasks but with a faster pace
The assessment of incoming building applications remains the core task of public servants, albeit being increasingly computer-mediated. According to case workers, it is now considered “a different kind of work” (building authority, case worker) due to the predominant screen work. Where previously large paper plans were spread over the desks, there are now two large screens on which PDF files are assessed. But the review of construction plans remains essentially the same. In terms of task complexity and discretion, no changes were observed. Instead, the difference to paper-based work lies in many small new work steps like converting and renaming files, combining PDFs, and sorting them into folder structures. Ticking boxes in the software in order to manage client access has also become necessary, which requires great attention and care each time. Case workers are also increasingly charged with support services for the applicants dealing with their handling issues. Regarding communication, we indeed find a shift towards “screen-level bureaucracies”, as indicated by the increased reliance on electronic communication. One studied administration has even cut down offering consulting to potential builders to reduce in-person interaction.
The change in tasks therefore mainly concerns the introduction of “work for digitalisation.” Remarkably, the various new tasks and process steps were only rarely complemented by organisational adjustments and instead simply attributed to the present employees. The simultaneous handling of both digital and paper-based procedures also causes supplementary workload like scanning and keeping all files up-to-date. Additionally, digitalised case processing is much faster since interactions between case worker and applicant/other offices proceed much quicker, which also drives expectations of immediate response. Consequently, employees stressed increased workload (not necessarily solely attributed to digitalisation but to understaffing) and smaller time frames.
Case workers with few alterations
As the core tasks of reviewing building applications remain the same, the job profile of case workers in building administration, who are technical specialists often with backgrounds in engineering or architecture, hardly changes. Nonetheless, IT-related competences are increasingly essential and transpire into job postings. Additionally, resulting from the intensification of checking application data and plans, permanently increased attention and concentration are required.
This small-scaled change for the case workers is in stark contrast to the outlook for employees with more supportive tasks in the archive and registry. In the medium term, their work is expected to disappear once automated interfaces and resubmissions replace the registry: If we now say we need a digital resubmission system, that would certainly not be a problem to introduce. Then, I would have to be afraid again that the two people I have in the registry would be taken away from me. […] And then such things would again be distributed among my caseworkers, who are already fully loaded [with work] anyway. (Building authority, local manager)
Today, parts of the registries’ tasks are already substituted, indicating that the registries mainly persist by taking on new digitalisation-related tasks such as administering the software, issuing logins, and customer support. These tasks often come with different (IT-related) competence requirements. Occasionally, former registry employees were able to achieve promotion due to this advanced qualified work. Still, such advancements remain short-term given the shrinking need for less qualified individuals, indicating some restratification.
Electronic tax return
Germany's tax administration has been in the process of digitalisation for more than two decades. Since 2007, a joint project of the federal government and all 16 Länder for coordinated new software development of tax authorities has been advancing the uniform digitalisation and automation of taxation procedures, including strict requirements for the deployment of digital tools in local offices. For instance, the implementation of any new digital provision is required within one year after its release. In addition, digital citizen self-services are mandatory for many taxpayers, with the exemption of pensioners and ordinary employees. The few remaining paper-based tax returns are scanned and then processed digitally. Today, about ten percent of tax returns are processed in a fully automated way based on a risk management system (RMS), and tax assessment notices are issued automatically.
Reduced discretion and work intensification
For most returns, final decisions are still made by public servants. This audit procedure is based on audit instructions provided by a computer program. First, this signifies that tax officials are less autonomous in their decisions of what aspects to check more in-depth than previously, as they are no longer required to review tax returns in their entirety. Second, while the public servants are, in theory, allowed to check fields of the tax return not flagged by the program, this is not reflected in practice. Those employed in the routine procedure are subject to severe time constraints inhibiting any assessment beyond the audit notes, as summarised by this manager: Okay, I’ll just check this point, even though I can see there is much more wrong. We are supposed to trust the system in that respect. If we have a lot of time, we can go into depth – we no longer do, so we are limited to the RMS. (Tax office, department manager)
In this sense, an implicit rather than an explicit reduction of discretion is noted. The employees’ perception ranges from the computer program taking on a leading, work-dictating role to a persisting feeling of autonomy. Either way, when boundaries set by technology are reached, the public servants respond with coping mechanisms such as looking for pragmatic workarounds.
A dual development is observed regarding the impact on the tasks’ complexity. A rise in complexity results, first, from the ceased need for simple plausibility checks which enables tax officials to increasingly focus on rather intricate aspects of a tax return. Second, the audit notes are not necessarily straightforward but rather complex. Third, public servants lament overwhelming data volumes and rapid technological changes requiring them to adapt their workflow continuously. However, this aspect is impacted significantly by the complex German tax law with constantly increasing regulations and rules. It is, in fact, reported that the complexity of tax law could no longer be managed without digital tools. The program used for tax audits is perceived as reducing uncertainty and complexity by providing a red string. This is at the expense of partially more monotone work for middle-level personnel who are now charged with data maintenance requiring diligence and meticulousness. Work relief is noted in data entry, as well as the time required for processing cases. However, fragile technological conditions and technical problems lead to increased workload and downtime that must be compensated later. Moreover, the process of building digital databases ties up personnel capacities. Digitalisation and automation bring work relief when technology runs smoothly, but currently this is not the case in the studied tax offices. Temporary work intensification also occurs during pilot phases when new software is tested for errors or weaknesses at a tax office. Testing and reporting on new digital tools increase the workload, and faulty programmes “inhibit” productivity (Tax office, case worker). Therefore, digitalisation and automation cannot yet compensate for staff shortages. Although we have more and more taxpayers, the demand is always, always decreasing – that is, the calculated personnel demand. I always say that digitalisation is already priced in that is not yet running. […] When everything is going well, everything is fine, but that is the case on very few days. (Tax office, local manager)
The respondents anticipated that once a more stable functioning of the digital processes is reached, the work-relieving effects may unfold more fully.
Phasing out of middle-level personnel
As anticipated, simple tasks that do not require tax expertise have increasingly been digitalised or even automated, resulting in a sinking need for lower-qualified and less specialised personnel in the tax offices. Currently, “working for digitalisation” is still needed: Tasks such as scanning and manual registration of tax returns or data reconciliation are performed by middle-level personnel but become increasingly obsolete. Therefore, a phasing out of not specialised personnel is reported. Even though the existing staff is not at risk of dismissal, “if someone new were to apply for such a position, they would have no chance of getting it” (Tax office, staff council), as one staff council comments on colleagues without tax training who carry out support activities in the registry or similar. In contrast, there is a high demand for highly qualified public servants who have completed extensive tax training and are digitally inclined, indicating a growing need for specialised competences and eventual restratification within the personnel.
Discussion
In this paper, we studied how digitalisation and automation impact the nature of work in executive agencies at the micro-level and the composition of the workforce at the meso-level. By comparing three administrative services in Germany (internet-based vehicle registration, digital building application, electronic tax return), we investigated changes in work complexity, workload, discretion, as well as the demand for new competences and consequent substitution of job profiles.
Concerning the task dimension (see Table 2), we identified both “digitalised work” and “work for digitalisation” (Ruiz Ben, 2021) respectively “invisible digi-work” (Justesen & Plesner, 2024). Regarding the former, the core discretionary task of case workers to make decisions in particular cases is (partially) substituted or increasingly standardised. The studied administrations have transformed into “screen-level bureaucracies”, like vehicle registration turning away from direct citizen interaction at the counter, or building administration introducing mandatory digital inquiry. We even observe a turn into “system-level bureaucracies” (Bovens & Zouridis, 2002), like tax administration, where digital citizen self-services are partly mandatory, and automation is most advanced. In this case, the most comprehensive changes can be observed. Applying an RMS, the whole tax audit practice is standardised and structured by, or even subordinated to, the algorithmic advice. Rejecting the recommendations, like Dekkers et al. (2019) or Meijer et al. (2021) identified for Dutch officers, is not feasible within the technological constraints. Here, technology is deployed to reinforce hierarchical bureaucratic control (Meijer et al., 2021), thus curtailing case worker's discretion (Buffat, 2015) in determining which information are prioritized when making final decisions.
Impacts of digitalisation and automation on tasks of public servants.
Impacts of digitalisation and automation on tasks of public servants.
In addition, in all cases, “work for digitalisation” (Ruiz Ben, 2021) or “invisible digi-work” (Justesen & Plesner, 2024), i.e. new tasks required to enable citizen self-services and automated internal processes, have emerged. While this “compensating work” (Justesen & Plesner, 2024) like citizen support, education, or digital upbringing was previously discovered in other contexts (Bovens & Zouridis, 2002; Breit et al., 2021; Marienfeldt, 2024; Pors, 2015), more mundane tasks such as scanning, printing and shipping or IT support emerge as well. In addition, “connecting work” and “cleaning work” (Justesen & Plesner, 2024), such as software maintenance but also entering or validating data, thereby preparing and maintaining the digital service delivery, are rarely recognised in the context of digital government. What the new tasks have in common is that they remain “invisible”, being unnoticed and devalued (Ruiz Ben, 2021; Justesen & Plesner, 2024). These tasks are commonly overlooked by policy-makers, managers, and citizens. It is often assumed that these tasks can already be executed by the employed digital systems, while, in fact, human public servants are performing them, like data matching. This is what Justesen & Plesner (2024) describe as “cleaning work” to ensure the functioning of digitalised organisations, or “compensating work” to uphold the illusion of fully automated systems. Such work is devalued both socially, not acknowledged as real work, and economically, insufficiently allocated with resources or being outsourced. For instance, scanning seems too trivial and is supposed to be only required during the transition phase. In some offices, scanning tasks were outsourced to private providers and thus – although necessary for public service delivery – not considered a relevant task any longer. Given the current set-up of digitalised or automated service delivery in German public administration, we argue that all these support and maintenance tasks are an important necessity that will be lasting and consequently, must be acknowledged in staff planning.
In line with previous research, we observe an increased workload. First, digitalised work results in techno-overload and a faster pace of work, as Fleischer & Wanckel (2023) have shown for government employees. Second, additional maintenance or compensating work (Justesen & Plesner, 2024), for instance, solving frequent technical problems (Bogumil et al., 2022; Kuhlmann & Heuberger, 2023) and increased need for client support (Giritli Nygren et al., 2013; Pors, 2015) causes additional workload. Currently, digitalisation-support and maintenance tasks are carried out by regular employees “on the side,” particularly in the building administration.
Although the literature on citizen self-service expects that some tasks of public administrative work are increasingly burdened on citizens (Breit et al., 2021; Madsen et al., 2022), and this is an underlying feature of our three cases, our findings suggest that citizen self-service only partially alleviates the workload on administration. These gains can only be achieved if “citizens can experience a digital self-service process without regularly turning to authorities for help” (Madsen et al., 2022). Currently, this is the case for annual tax returns, but not for vehicle registration or building applications. Moreover, it must be taken into account that even though in some domains (such as tax offices) the digitalisation has led to simplifications and efficiency gains, at the same time, the scope, legal complexity and regulatory volatility of tasks have significantly increased over the past twenty years (Adam et al., 2019), while the number of employees has remained unchanged or has even shrunk. We identified constant understaffing as a structural issue causing increased workload, especially in tax and building administrations. In these cases, digitalisation benefits are already priced in to compensate for understaffing, essentially leaving no room for relief for the case workers.
Concerning task complexity (Bullock, 2019; Frey & Osborne, 2017), we observe diverging developments. Tasks like checking data or organising client records have become less complex with the integration of decision support systems providing guidance and automated plausibility checks, exemplified by vehicle registration. However, this reduction in complexity is counterbalanced by increasingly complex legislation, rapid technological changes and new workflows, demanding continuous adaptation and precision in execution. Overall, changes in task complexity are better characterised as a redistribution rather than a simplification. Notably, in cases of reduced complexity, discretion is constrained by software-guided procedures, as seen in tax administration, or remains as low as before, as in vehicle registration, while in building authorities the high level of discretion is maintained. Table 2 summarizes how digitalisation and automation impact the tasks of public servants in our cases.
With regard to the staff dimension, our findings support past research showcasing the growing importance of new, digital and IT competences (Distel et al., 2019; Hunnius et al., 2015; Mankevich et al., 2023; Mergel et al., 2019). Moreover, our findings highlight the importance of role-specific differentiation as suggested by Ogonek et al. (2020). Depending on the particular task and role, i.e. public servants performing digitalised work or work for digitalisation, some competences become more relevant: Support or educational tasks require soft skills like communication and conflict management competences; more specialist case work demands analytical thinking and specialised (legal) knowledge. This development entails up-skilling of middle-level public servants in the short term and possibly higher entry barriers for candidates who are less qualified and lack relevant competences in the long term. In contrast to other research, we did not identify new job profiles like, for instance, “citizen guides” in Denmark (Pors, 2015).
In line with Zanker (2019), we identified the biggest substitution potential for traditional administrative job profiles, i.e., vocationally trained, middle-level job profiles such as clerks in archives and registries. We also find moderate substitution potential for case workers handling standardised cases, and supportive-level staff performing manual, compensating tasks (see Figure 1).

Long-term substitution potential of job profiles.
Being under pressure from the increased use of electronic file systems, clerical jobs in archives and registries can only prevail if they change their work content, e.g. involving software administration and rights management. Others, such as archivists in vehicle registration offices, have already been substituted. At present, there is a lesser impact on low-skilled employees, as tasks such as scanning become more important during the current stage of digitalisation. They may also be outsourced or discontinued at a later stage, as anticipated in the tax administration. Only the specialist level with specialised knowledge is not subject to potential substitution. This overall trend of changes in competence requirements and (potential) substitution of certain job profiles supports Noordegraaf's (2016) argument of long-term restratification within public personnel composition and structure. Moreover, it highlights the disproportional impact on workforce “supporting the core” with “invisible” work in contrast to case workers at the “core” (Justesen & Plesner, 2024). Yet, in contrast to the literature presenting technological advancements as substituting, and thus threatening human labour (Autor et al., 2003; Frey & Osborne, 2017), in our cases, digitalisation and automation are understood as a means of counteracting existing staff shortages and recruitment difficulties.
Our findings emphasise the transformation into “digitalised work” in executive agencies, including less complex tasks being partially substituted and core discretionary tasks being standardised. We demonstrate the importance of “work for digitalisation” (Ruiz Ben, 2021) for the public sector, i.e. “connecting work” such as software administration, “compensating work” like scanning and IT support, and “cleaning work” such as validating data, which often remains “invisible”, unrecognised and undervalued (Justesen & Plesner, 2024). In contrast to the ‘utopian’ view of digitalisation and automation as a universal solution for better and more efficient service delivery and internal processing leading to reduced workload, and the ‘dystopian view’ of large-scale public personnel substitution, we show that digitalisation and automation, as it is currently set up and utilised in German executive agencies, is heavily dependent on human labour. Indeed, role-specific new competences and consequent upskilling of public servants are required. More precisely, we showcase a dual development. On the micro-level, traditional administrative, less complex tasks of decision-making in routine cases are progressively phasing out already in the short-term. On the meso-level, this results in up-qualifications of middle-level public servants, but possibly higher entry barriers for mostly less qualified workers who lack the necessary competences in the long-term. Although traditional administrative job profiles are potentially substituted, digitalisation and automation are seen as means to counteract existing staff shortages and recruitment difficulties rather than as threats to human labour. With our findings, we contribute to the literature on the internal effects of digitalisation, in particular on the impacts on public sector work and staff.
Moreover, our findings point to external effects of digitalisation, such as increased administrative burdens and unequal access for citizens due to the digital divide. First, outsourcing tasks to citizens results in higher administrative burdens as citizens must learn how to complete tasks previously conducted by professional case workers and comply with specific data formats (Madsen et al., 2022). Second, as a result of the reliance on the individual citizen's skills, equal access to services can be impaired. It can further deepen the gap between citizens and the state if personal interactions with the administration no longer take place. From a normative point of view, policy-makers should consider the purpose and the type of services to be digitalised and automated. While our analysis focused on largely rule-based services, the “compensating” work as well as unintended side effects are expected to be much more comprehensive for social services.
Yet, as digitalisation and automation in public administration is an ongoing process, the further transformation remains to be seen. As one of several avenues for future research, we propose to analyse the development of work and the workforce in a longitudinal manner, taking into account both previous and upcoming forms of digital technology use in public administration. Given the increasing scholarly debate about the potential of artificial intelligence for public service delivery (Bullock, 2019), the impact of algorithmic decision-making should be empirically investigated, critically examining the nature, scope and pace of changes for work in public administration.
Second and even more importantly, as we have shown with our findings, variations between public administration domains as well as between job profiles occur due to distinct digitalisation and automation approaches and different paces of change. Therefore, we recommend a systematic comparison of different public service domains and respective job profiles or roles, for instance, social services that require more interpersonal interactions or inspection authorities that must control on-site. In this regard, a more profound analysis of framework conditions to explain the variations in digitalisation-induced changes in work is necessary. The extent to which digitalisation and automation will bring about improvements rather than hardship for public employees greatly depends on the organisational, financial, technical, legal and, finally yet importantly, political design of the digital transformation as a whole.
Footnotes
Acknowledgements
The paper draws on findings of two interrelated research projects on “Digitalisation and automation of public administration in Germany” funded by Hans Böckler Foundation, and “Digitalised and automated administrative services and the delivery to citizens” funded by the trade union ver.di. They were carried out jointly by the University of Potsdam and the Ruhr University Bochum under the direction of Prof. Dr. Sabine Kuhlmann and Prof. Dr. Jörg Bogumil.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research presented in this article received funding from the Hans Böckler Foundation and the trade union ver.di.
Conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
