Transforming professional values in an automated public sector

Automation of the public sector is rapidly increasing worldwide based on dominant discourses of efficiency, equality and standardization (Bannister and Connolly, 2014). Automation can, in its simplest sense, be defined as “the execution by a machine agent (usually a computer) of a function that was previously carried out by a human” (Parasuraman and Riley, 1997, p. 231) and has been introduced in public administration based on ideals of “neoliberal efficiency, cost cutting, accuracy and industriousness” (Sleep, 2024, p. 849). Other values relate to how automation can improve service delivery for citizens (e.g. Reddick, 2005) and how technology can relocate accountability and decision-making to citizens (Petrakaki, 2018; Smith et al., 2010).

The upholding of public values is expected to be done by street-level bureaucrats, or case workers, and it is therefore important to investigate how the rapid increase of automation affects these “upholders,” especially taking into account how their judgements may differ in terms of professional values. Currently, little is known about these changes – not much about value transformation in general and even less so about values in relation to different occupational groups. Several reviews call for empirical investigation (e.g. Zyzak et al., 2024; Charles et al., 2022) since value management in digital transformation “is a relatively unexplored topic” (Zyzak et al., 2024, p. 845)

Looking at the limited evidence at hand, we find a few notable examples. Aschhoff and Vogel (2018) analyze which value tensions arise when public values are co-produced with citizens. Based on three German case studies, they identify three major public value tensions occurring in co-production: the inclusiveness-accountability-tension, the flexibility-accountability-tension and the productivity-diversity-tension. Borry and Getha-Taylor (2019) use data from both state and local government levels in the United States to see how automation impacted the workforce regarding the trade-offs and ethical impacts. They find that professions occupied by female and non-white employees are more likely to be threatened by automation and argue that those making decisions about automation “should be aware of the demographics of those jobs and the potential implications of reductions in force on equity and diversity goals” (pp. 18–19). Swedish studies focusing Robotic Process Automation (RPA) in public sector organizations show how caseworkers often express professional values on effectiveness and legal correctness, but caseworkers also found RPA to threaten their professional autonomy and possibility to make individualized judgments (Andersson et al., 2022; Germundsson and Stranz, 2024; Nauwerck and Cajander, 2019). Caron and Giauque (2006) study challenges to civil servants’ professional identity that arise when new public management principles and tools were introduced in Canada and Switzerland. New professional values such as quality of customer service, flexibility, innovation and efficiency were insufficient “to allow civil servants to develop their own capacity for ethical deliberation” (p. 543). Finally, based on a survey distributed to Swedish municipalities and national agencies, Sundberg (2019) emphasizes a trade-off between different values when public administration is automated. These trade-offs relate to a set of value positions, proposed by Rose et al. (2015), including professionalism, efficiency, service and engagement – value positions that make our starting point for further investigation.

This article sets out to further investigate the trade-off between the value positions proposed by Rose et al. (2015) and studied by Sundberg (2019), from the perspective of professions and professionalism with special emphasis on differences in the experienced trade-offs between occupational groups. Our investigation is based on interviews with caseworkers at two different administrative levels – national and municipal. At the national level, we interviewed case workers at a driving permit licensing unit at the Swedish Transport Agency. At the municipal level, we interviewed social workers in social assistance administration. As shown above, previous research has found that many caseworkers experience the shift to digital administration to have changed their daily practices – especially in relation to their professional values. We want to know in which ways these values have changed and if the changes differ between different occupational groups. We thus pose the following research questions:

By focusing professional values in two different occupational groups, we contribute not only to research on value transformation when public sector is increasingly automated, but also to a more nuanced view on how this value transformation is situated in specific organizational and professional contexts characterized by different values. This view challenges the technology-deterministic view often found in literature on e-government, where technological change is viewed as unidirectional. Another contribution is to connect e-government literature with the sociology of professions. This approach can provide a more nuanced picture of certain dilemmas within the public sector, such as the balance between flexibility and predictability or rule-following and discretion. A final contribution is to elaborate on Rose et al.’s (2015) value positions framework to also account for individuals’ values since this framework was previously mainly used to study entire organizations.

This article addresses value positions by intersecting literature on e-government and the sociology of professions. This intersection is motivated by tensions that are enhanced by the rapidly increasing use of digital tools in the public sector. These tensions concern repeated dilemmas within public authorities to balance flexibility and predictability, but also the expectation on (and of) occupational groups to be equipped with discretion (the power or right to decide or act according to one’s own judgment), based on their professional status. As we will return to below, the literature on e-government has not sufficiently taken the issue of professionalism into account.

The literature on value propositions in e-government (e.g. Rose et al., 2015) equates professionalism with accountable and rule-based Weberian bureaucracy. When tapping into literature on sociology of professions, however, we need a richer vocabulary. Evetts (2013) shows that within organizations, there is a tension between organizational and occupational control. Organizational control is characterized by hierarchical authority, standardization and managerialism. Thus, it overlaps with the understanding of professionalism used by Rose et al. (2015) in the e-government literature. Occupational control, on the other hand, relates to collegial authority, professional ethics and discretion based on professional training. Additional characteristics of this understanding of professionalism include shared ethical standards upheld by professional associations and unions. Taken together, professional power is grounded in the societal expectations that professional groups act in accordance with their ethical standards and make well-informed decisions based on their specific knowledge (Svensson and Evetts, 2010; Brante, 2022).

In previous literature on professionalism, there is a vivid discussion concerning which occupational groups could be considered professions (see, e.g. Brante, 2022). We acknowledge the relevance of discussing the traits of professions while also avoiding a binary conception of professions by taking a more ideal-typical approach, with inspiration from Evetts’ (2013) distinction between managerial and occupational control. From this perspective, professionalism occurs in the conflict between organizations and occupational groups concerning the governing of work. The scope of professionalism is greater when an occupational group has discretion to make decisions based on its professional knowledge. At the other end, the scope of professionalism decreases when an occupational group is given less discretion.

The relevance of tapping into sociology of professions to the discussion of value propositions in an automated public sector concerns the need to consider how occupational control is affected by automation in the public sector and how this, in turn, affects service users. Many public organizations handle complex issues, administered by civil servants with specialized training, that concern citizens with complex needs that do not always follow the streamlined categories and regulations set up by authorities. Further, many public organizations need to moderate goal conflicts inherit in policies. For instance, caseworkers in social services are expected to both support and control long-term unemployed individuals, which creates tensions for individual officers as they must balance these two tasks in relation to individual clients (see, e.g. Torsvik et al., 2022; Lipsky, 2010). From the perspective of occupational control, the argument is that these types of services require discretion to overcome inherent goal conflicts. This argument raises the question of how automation affects public services, civil servants and clients. In our case, we focus on how automation affects caseworkers’ possibilities of acting in accordance with their professional values.

To investigate professional values in relation to automation, we use the value positions framework presented by Rose et al. (2015). There are many frameworks on how to capture public value creation in automation (see Luna-Reyes and Zhang, 2023). Our choice is motivated by the frequent use of the framework in similar settings (e.g. Ranerup and Henriksen, 2019). The framework consists of four value positions: Professionalism, Efficiency, Service and Engagement (Rose et al., 2015):

1. Professionalism is focused on providing independent, robust and consistent administration, governed by a rule system based on law. Automation should support accountability and standardized administrative procedures grounded in the values of equity, legality and accountability.

2. Efficiency concerns the provision of lean and efficient administration that minimizes waste of public resources gathered from taxpayers. Thus, automation is viewed as a productivity tool based on the values of efficiency, value for money and productivity.

3. Service involves maximizing the utility of government to society by providing services to citizens and businesses. Automation ought to improve availability, accessibility and usability of services based on the values of citizen-centricity, service level and quality.

4. Engagement involves engaging with civil society to facilitate policy development in accordance with liberal democratic principles, thus articulating the public good. Automation should support deliberation and networking based on the values of democracy, deliberation and participation.

The framework opens for potential value conflicts, or what Rose et al. (2015) refer to as congruence or divergence between different value positions. A typical conflict is between equal treatment of all and individual treatment based on needs. For instance, Considine et al. (2022) discuss “machine bureaucracies” in welfare and employment services, focusing on tensions between efficiency versus inclusiveness as well as between consistency versus personalization.

Whereas the framework was only used as a sensitizing device in this study, we, however, argue for the need for a bottom-up approach, investigating individual caseworkers’ values related to professionalism. The framework is developed based on a managerial perspective and designed to study entire organizations and needs to be adapted to other stakeholder groups and include these stakeholders’ professional values – values that “lie at the foundation of a profession and serve to shape individual roles and careers of its practitioners” (Holland, 1989, p. 28).

This exploratory study aims to deepen the understanding of the value trade-offs caseworkers experience as their work becomes automated, and whether these trade-offs vary across occupational groups. Accordingly, an interpretive approach was adopted (Walsham, 1995). Qualitative data were collected through semi-structured interviews (Ruslin et al., 2022) and analyzed using reflexive thematic analysis (Braun et al., 2023). These methods enable an in-depth examination of how automation shapes perceived value tensions and influences caseworkers’ ability to act in line with their professional values. Case and informant selection was guided by the aim of capturing diverse perspectives rather than achieving empirical saturation.

The study was conducted in Sweden and investigates two cases of automation: (1) driving permit licensing at the national Swedish Transport Agency and (2) social assistance administration within municipal social services. The cases were chosen to allow us to explore the differences between two occupational groups. Both contexts concerned the use of RPA, an algorithm that automates tasks previously performed by humans. The department at the Transport Agency, employing Driving License Case Workers (DLCWs), is responsible for processing applications for learner’s permits, issuing driving licenses, warnings and revocations of drivers’ licenses, etc. They handle client contacts as well as investigations concerning more complex cases. The RPA ^[1] was implemented in the agency during 2010–2013. The social workers are employed by municipalities, i.e. the lowest governmental tier in Sweden. All social workers worked with the social assistance administration, i.e. processing basic income support. Most worked in middle-sized municipalities in Sweden and in all municipalities, RPA was introduced during 2019–2020.

The two occupational groups differ significantly in terms of occupational control. Swedish social workers have several of the traits associated with occupational control, including shared educational background through the seven-terms-long social worker program. The program is based on the scientific discipline of social work. In terms of unionization, Swedish social workers are either members of the Union for Professionals or Vision. Both associations represent different occupational groups; the former only represents academic occupations, while the latter also represents non-academic occupations. Swedish social workers also have profession-specific ethical guidelines provided by the Union for Professionals (2015). In comparison, DLCW is a less institutionalized occupational group. There is no systematic data on the educational background. According to an informant of the Transport Agency (personal communication, 09/09/2024), there are general requirements regarding academic education authority for most caseworker positions, including DLCWs. However, there are no formal requirements on the specific subjects or scientific disciplines to be covered by this education. Turning to unionization, DLCWs are mostly represented by two organizations: the Union of Civil Servants (which represents civil servants both with and without an academic degree), and the governmental branch of the Swedish Confederation of Professional Associations, which is an umbrella organization that gathers unions representing academic white-collar workers.

Another aspect that divides the two occupational groups is the ethical standards promoted by the unions and intended to guide their work. As acknowledged within sociology of professions (Svensson and Evetts, 2010), such guidelines are central traits of professional groups, manifesting a collective professional self-conception. The Union for Professionals (2015) highlights social workers´ ethical responsibility in supporting their client´s well-being based on respect for clients grounded in solidarity and human rights. In comparison, DLCWs lack designated ethical standards, but are covered by more state-centric guidelines promoted by the Union of Civil Servants (2022). They stress a more rule-of-law-oriented understanding, arguing that DLCWs and other civil servants are part of a legal and objective public administration that works on behalf of citizens.

The choice of informants was guided by the aim of capturing diverse perspectives and deep understanding. We chose caseworkers who had experience working both before and after the RPA implementation and who could reflect on the effects of RPA. The DLCW interviews were conducted in relation to a project on accountability in public sector automation as well as changes in the caseworkers’ work and professional values. The interview guide was guided by theories of public accountability (see, e.g. Behn, 2000) and focused on automated decision-making, how technology was used for case processing and making decisions on driver permits. The interviews with social workers were conducted in relation to a project on social work automation and the interview guide focused on how RPA affected caseworkers’ possibilities of acting in accordance with their professional values and was guided by the theoretical lens of the four value positions (Rose et al., 2015).

All interviews were recorded and transcribed and all statements have been verified through concurrent document analysis in both projects. Table 1 summarizes the time and focus of the interviews in the two cases.

The data were analyzed using thematic analysis, following the six-phase framework proposed by Braun et al. (2023). This approach is well suited for identifying, analyzing and interpreting patterns in qualitative data. The analysis began with familiarization through repeated readings of the interview transcripts. In the initial coding phase, we would draw an arrow from one statement in the transcripts writing down, e.g. “loss of agency and control”. Second, we categorized, or grouped, statements around values that were related to the automation of the informants’ work – e.g. “loss of agency” and “computer made decision” were categorized as the “RPA taking decisions on money to be paid or driving license revoked”. Coding was still primarily inductive here, aiming to capture the case workers’ own expressions and conceptualizations. In the next phase, these values were related to and framed into Rose et al.’s (2015) theoretical framework on value positions where applicable – e.g. “RPA taking decisions” were put in the box “efficiency” when related to “we are quicker now” and when it did not fit as in “loss of agency” we put them in a separate unlabeled box. We went through the text again, questioning and querying our assumptions and themes, and finally labelled the boxes that were not covered by the framework. To address concerns related to subjectivity and enhance reliability, the analysis was conducted collaboratively by the authors. Our coding schema with example quotes can be found in  Appendix.

Once all statements were categorized into the initial framework or in new boxes, we started the final step, which involved synthesizing the findings into a coherent narrative structured around the identified themes. By reading the individual case stories, similarities and difference between the cases were analyzed.

The professional values brought to the fore with the introduction of an automated caseworker were: equity, legality, accountability, efficiency, service level, relational work and discretion. The latter two are not included in the framework proposed by Rose et al. (2015) but added by us. These seven professional values will be described from the view of both occupational groups below.

In the value positions framework, equity is described as “equality before the law” (Rose et al., 2015). Johnson and Svara (2015) describe how “[p]ublic administrators should prevent and reduce inequality, unfairness, and injustice based on significant social characteristics and to promote greater equality in access to services, procedural fairness, quality of services, and social outcomes” (Johnson and Svara, 2015, p. 282). At the Transport Agency, equity through standardization was the explicit motive to automate case processing at the agency:

The quote illustrates that automation of case processing became important for standardization and enhancing rule-of-law. DLCWs emphasized that the automated caseworker pushed the human caseworkers to ensure equal treatment of service users: “We try to use the same rule-system irrespective of if it is an automatic or manual case processing” (DLCW 3).

Among the social workers, equity was also valued but not addressed as directly or frequently:

The introduction of an automated caseworker was mainly described in positive terms by all informants in relation to the professional value of “legality.” Automated case processing reduced manual errors, documents no longer disappeared and it became more difficult for unauthorized people to access the client’s documents and personal information. As described by a social worker:

When decision-making is made by the automated caseworker, the strict rule-following is even further reinforced, as described by a DLCW: The automated decision-making system follows how it is programmed. It only follows the given rules. (DLCW 4)

Whereas the caseworkers in both occupational groups consider the automated caseworker mostly helpful in getting things right, they cautioned for over-relying on it since both clients and the automated caseworker could get things wrong. However, the automated caseworker, both at the Transport Agency and at most social work units, had to provide a notice in their system if there were large deviations in the entered figures, and in most cases, the automated caseworker was only allowed to make positive decisions.

Since a positive decision was almost never appealed, there were no cases where the accountability of the RPA had been tested from such a perspective.

The excerpt illustrates that the legal systems created a structure for automated decision-making, but it could not lead to a formal decision when the response was negative. This illustrates that legality is formed around a positive decision-making process, and even when the Administrative Procedure Act (SFS, 2017:900) recently changed in Sweden to allow for computers to also take negative decisions, many social work units have not implemented this option due to an expressed doubt whether the RPA should make its own decisions regarding payments without control from administrators.

All social worker interviewees believed that it was still the social worker who ought to be responsible for all decisions that the automated caseworker made and that it was the social workers’ responsibility to read through the automated decisions. Thus, in terms of responsibility for decisions, the social workers considered it to be unchanged with RPA. At the Transport Agency, however, the introduction of automated caseworkers re-arranged the accountability with increased uncertainty of who was responsible for the decisions made. When discussing accountability and who or what was responsible for the decisions taken by the automated caseworker, many DLCWs were hesitant:

Efficiency gains and increased productivity were stressed as important values at the Transport Agency, and in a few instances at social services. At the former, being “efficient” was described as highly valued. One of the core ideas of the management was to have a unified manner of working, with a high degree of automation for efficiency gains. This made the automated caseworker popular and DLCWs talked about “him” with a sense of admiration due to his speed:

The social workers similarly talk about the efficiency gains enabled by the RPA, but always in relation to other professional values, typically relational work:

The social workers also mentioned that the case processing was not considered stimulating, but rather a kind of “appendage, a necessary evil for the client to survive until the next month, so to speak” (social worker). The social workers were also more critical, compared to DLCWs, stating that the automated caseworker did not always live up to the promises announced by the management. Some caseworkers felt that the workload had become even higher after the introduction of RPA:

Service concerns accessibility and usability of services online to foster values such as citizen-centricity, service level and quality. During the interviews, availability in terms of more channels for contact was one of the most frequently mentioned value by both occupational groups. For instance:

At the Transport Agency, the DLCWs also valued the new opportunities for contact that the e-services and the automated caseworker have entailed, but they also discussed how information accessibility – especially regarding language – has decreased:

Relational work is a professional value that we did not capture through using Rose et al.’s (2015) framework, but it was mentioned by all social workers – often as a value overriding all other values. When the automated caseworker took over many time-consuming clerical tasks, the caseworkers gained time for relational work with their clients:

The personal meeting was what social workers considered to be the most important part of their profession. Thus, the social assistance administration – that was automated through RPA – was considered less significant. They believed that the tasks associated with their profession were much more important and time-consuming:

Another social worker expressed that some tasks would possibly change due to RPA, but that the basic values would remain. Focus on the client and the personal meeting was still, according to the managers, the most important task:

Whereas relational work was not mentioned explicitly by DLCWs, we find traces of this aspect in the adjacent category of discretion described next.

Discretion is a professional value not captured by Rose et al.´s (2015) framework, but mentioned by all social workers and many DLCWs. Discretion can be described as “the freedom street-level bureaucrats have to make decisions concerning individuals regarding the sort, quality and quantity of sanctions, and rewards during policy implementation including the possibility of no sanction at all” (Busch and Henriksen, 2018, p. 3; Lipsky, 2010). It concerns the power to decide or act according to one’s own judgment; freedom of judgment or choice and is often discussed in terms of making individual assessments for service-users based on professional values. According to the interviewees, there is some human element missing when case processing is made by a computer—“humans are humans, robots are robots” (social worker 8). As described by a DLCW:

Some respondents criticized how the individual manner of working would be affected; that the human perspective based on individual assessments could be neglected if RPA were to have greater influence on the case processing. One social worker considered that certain tasks could be automated, such as payments and the like, while the individual treatment and assessments could not:

Caseworkers in both organizations sometimes tried to override the automated caseworker regarding the professional value of discretion. One DLCW felt sorry for a young man who had his driver license withdrawn after driving too close to the car in front of him, which resulted in him having to re-take his driver license test. The man had just relocated and been offered a new job that required a driver license. He also told the DLCW that his dad had recently died. Based on the information, the DLCW used his/her discretion to speed up the process:

Based on professional discretion, the caseworker minimized the time for decision. The discretion enabled this processing without the DLCW stepping outside the formal legal system. This example shows how the human caseworkers can add dimensions to the decision-making that the automated caseworker cannot take into consideration.

Little is known about value transformation in public sector automation (Zyzak et al., 2024) and more empirical accounts of this transformation have been called for (e.g. Charles et al., 2022). We have heeded this call through our empirical investigations, interviewing caseworkers in the public sector. We found the following professional values to be affected when case processing was automated: equity, legality, accountability, efficiency, service, relational work and discretion. Since the discourses around the automation of the public sector mainly revolves around efficiency and standardization (e.g. Sleep, 2024), it is not surprising that both categories of caseworkers stressed values of equity and efficiency. The automated caseworker in both organizations seemed to push human caseworkers to ensure equal treatment of clients. Further, efficiency gains were reported in both cases. The professional value of legality had in both cases become easier to act upon due to the strict rule-following that comes with RPA, which aligns with previous research (e.g. Nauwerck and Cajander, 2019). Turning to accountability, and especially the responsibility for decisions, we found some discrepancies between the two cases. The value of accountability was not challenged or reinforced by the introduction of RPA for the social workers, while it was made unclear for DLCWs. This difference is in part explained by how automation of the public sector often re-arranges accountability (Smith et al., 2010), but it does not explain why this did not happen within the investigated social services. Delivering good service is also a value that was addressed in both cases. We found that relational work was seen as a very important value by social workers – often as a value overriding all other values. This is not surprising given the importance of human relationships stated in professional guidelines for social workers (e.g. the Union for Professionals, 2015). According to the interviewed social workers, relational work was to be enabled by the efficiency gains made possible by the automation of more administrative and repetitive tasks. Limitations in how automation affects caseworkers’ possibilities of acting in accordance with their professional values are made most visible through the value of discretion. Like the findings by Germundsson and Stranz (2024) and Andersson et al. (2022), the interviewed caseworkers in this article experienced that automation of their work limited their discretion, i.e. their room for autonomy. This included room for making value-based judgments. Prior to this, seemingly simple decisions were problematized both by social workers and DLCWs based on values such as the consequences for single clients. Both examples show that, in practice, public administration is more complex than simply implementing standardized procedures. Rather, public officials use their discretion, including liberation over value conflicts, to overcome – what they consider to be – too rigid interpretation of policies.

We also made a point of how little is known about value transformation in relation to different occupational groups. Whereas similar professional-related dilemmas, investigated by us, have been discussed by, e.g. Caron and Giauque (2006) and Andersson et al. (2022), we contribute by grounding our analysis in sociology of professions (Svensson and Evetts, 2010) to discuss how different occupational groups within the “guild” of caseworkers experience these changes. We found that the scope of professionalism decreased when the occupational groups were given less discretion. Additionally, based on the ethical guidelines provided by the unions, we expected social workers to experience goal conflicts relating to the dilemma between control and support (Torsvik et al., 2022), given that the organizational control (Evetts, 2013) increases. However, the social workers expressed enthusiasm about automation, expecting it to decrease their administrative workload, thereby saving them time to do “actual social work.” This surprised us based on our assumption that social workers would find the introduction of RPA more problematic, compared to DLCWs, given the emphasis on individualized services in social workers’ professional ethical guidelines, as well as in the policy that guides their services. The explanation to this contrary finding is illustrated in the quote where the case processing of unproblematic social assistance applications was not considered “actual social work” by the social workers. Thus, reducing the degree of discretion in that part of their work was not considered problematic among the social workers. Rather the opposite, as described. The lesson to be learned here is the need to consider what tasks are automated. The social assistance RPA only processed approvals, which meant that social workers needed to process rejections, which in turn meant that many goal conflicts were not automated but remained subject to human decision-making. In comparison, DLCWs expressed the need to correct the RPA when it made assessments considered too strict by the caseworkers. These situations seemed to occur because the RPA couldn’t comprehend complex assessments. This illustrates that the type of goal conflicts discussed by Considine et al. (2022) and Rose et al. (2015) also occur in seemingly less conflicting policy areas such as driving license permits. Furthermore, seemingly more conflicting policy areas, such as social assistance assessments, can be automated in a way that retains much of the discretion for value-based assessments undertaken by caseworkers. With regard to practice, public sector organizations need to include caseworkers in the development, or procurement-process, to capture the actual users’ requirements on the system, thereby safeguarding professional control and values. Like Norling (2025) we strongly recommend that policymakers and public sector managers, incorporate a “cultural awareness” of professional groups “to foster a more conducive environment for digital transformation within bureaucratic settings” (Norling, 2025, p. 91).

As with all case study research, the obvious limitation of this study is that it lacks breadth. In a case study, the researcher typically researches one organization or one phenomenon. In our case, we examined two organizations, but the number of informants was limited to the available participants who had experiences also before the automation. Other major limitations are that the two cases are from one country (Sweden) and were researched with some years apart. This limits the potential for generalizing our findings. However, case studies may reveal important information about the object under study through the rich details they provide and allow for generalization, not to other cases, but rather to theoretical constructs. An avenue for further research in this sense relates to our use of Rose et al.’s value positions framework (2015). We argued that a shortcoming of this framework is the top-down approach and its limited use to explain value transformations below the managerial level. By analyzing professional values at the individual caseworker level, we extend the framework to incorporate individual perspectives. Furthermore, we see the need for similar investigations in other contexts, both in terms of countries and sectors, to further expand the understanding of how professionalism taps into questions concerning automatization. Given the rapid growth of more complex algorithms, not least Large Language Models such as ChatGPT, we expect this need to become more pressing when algorithms replace humans in more complex tasks, as the influence of human bureaucrats decreases in favor of black-box algorithms.

This article started out by arguing that too little is known about value transformation when the public sector is automated and that more empirical accounts of this transformation are needed. Based on our empirical investigations and drawing on a framework on value positions for e-government we found several professional values brought to the fore: equal treatment of clients increased; legality was reinforced; accountability of decisions was unchanged in social services but changed at the Transport Agency; efficiency mainly increased; and service-level was mainly increased through the use of more channels. Other professional values that were directly affected by the ongoing automation were: time for relational work increased and the opportunity to make individual assessments (discretion) decreased.

We compared whether there were differences in the professional values that are actualized based on two occupational groups. There were more similarities than differences – a caseworker is a caseworker in many regards. The main implication for public management practice is to include caseworkers in the development, including the procurement process, when their tasks are automated, specifically bearing professional values in mind. The main implications for public management theory are to develop methods and theories on how caseworkers’ values can be integrated with technology and to further adapt Rose et al.’s (2015) framework to better fit the individual level.