The distant,the capable and trustworthy,and the digitally skilled—the production of client roles in municipal activation and social assistance services using automated decision-making

Abstract

While research on the ramifications of automated decision-making (ADM) for welfare organisations is growing rapidly, less attention has been paid to elucidating the role and experiences of clients in the ADM process. This paper draws attention to how the performance of ADM policy in social assistance organisations working from an activation perspective is not merely a passive way of processing clients within the constraints of a given institution but an active interaction between clients and the organisation’s employees that ultimately constructs and produces the individual client in line with the purposes of the organisation. This article aims to explore how the role of the client is produced in the interaction between employees and clients when applying for social assistance in two Swedish social assistance organisations using ADM. The data consist of 28 qualitative interviews with managers, quality personnel, professionals, and clients. Findings depict the emergence of three client roles—the distant, the capable and trustworthy, and the digitally skilled client—the configurations of which are shaped by the interplay between employees and clients, serving to fulfil specific organisational imperatives.

Keywords

automated decision-making social assistance people production welfare encounter social worker-client relationship

Introduction

Digital technologies play an increasingly vital role in contemporary social policy internationally, as governments and welfare agencies aim to enhance service delivery, administrative efficiency, and resource allocation. Automated decision-making (ADM) is becoming integral to social protection and public employment services, shaping both access to and the delivery of welfare benefits (Griffiths, 2024). In Sweden, ADM is implemented in several social assistance organisations, where clients’ eligibility for public services and benefits is assessed. Municipal job centres working with activation services are organised to meet the needs of social assistance recipients and other unemployed individuals whose benefits are conditional and who meet high thresholds to enter the labour market.

Research has discussed how the use of ADM to promote labour market integration affects the work tasks, roles, and professional identities of social workers (e.g. Meilvang and Dahler, 2024; Petersen et al., 2020; Ranerup and Svensson, 2023; Scaramuzzino, 2019). However, less research has focused on the role of ADM in the welfare encounter between clients and welfare employees, as well as its impact on clients. Knowledge on this encounter is crucial to prevent further marginalisation, as clients may lack the knowledge and resources to access digital technology, which can lead to the inclusion or exclusion of specific groups and amplify inequalities (Ragnedda, 2017; Scaramuzzino et al., 2024; Schou and Pors, 2019). Studies on welfare-to-work and labour market integration are not only about “providing (or denying) people access to social services, social support, or forms of income support but also about changing people’s behaviour and attitudes, about passing moral judgement/…/and about the social construction, categorisation, and classification of groups of clients” (Van Berkel, 2017: 16). For example, standardised digital tools can transform binary data into a written storyline which creates “digital clients”, where the border between technique and person is blurred (Martinell Barfoed, 2019). Such accounts demonstrate how digitalisation can contribute to “clientship”, highlighting the dependency and asymmetry of welfare work that generates “appropriate” client behaviour from someone dependent and passive with little agency (Mik-Meyer and Silverman, 2019). Worthiness and neediness are therefore negotiated within a policy framework marked by asymmetry and dependence. As a result, the social worker-client relationship is far from neutral; it is influenced by power dynamics that can evoke feelings of powerlessness, guilt, and shame (Rollins, 2020). Such a power imbalance is a characteristic condition of human service organisations, where non-voluntary clients seek benefits that are not supplied elsewhere, from professionals with substantial discretionary powers (Hasenfeld, 2010).

The article aims to explore how the role of the client is produced in the welfare encounter between employees (managers, quality personnel and professionals) and clients when applying for social assistance in two Swedish social assistance organisations using ADM. To meet this aim, we adopt a social constructivist perspective on technology, emphasising the reciprocal relationship between technological systems and human interactions. We draw on Holstein (1992) to better understand how the performance of ADM policy in social assistance is not merely a passive way of processing clients within the constraints of a given institution, but an active interaction between clients and the organisation which collectively constructs the properties of the individual client to align with the purpose of the organisation. From this perspective, we contribute to the literature on ADM in social work and digital documentation practices by highlighting the clientisation process, in which client roles are constructed through everyday social work interactions, ultimately creating social policy. This analysis highlights the role of ADM in the digital transformation of social services, as it may influence the identity of human service organisations (cf. Wessel et al., 2021).

Previous research

There is extensive research on the use of digital technologies in social work, covering technologies used for different purposes across diverse areas and client groups. This article addresses research on ADM in social work and social assistance, and digital documentation practices in social work. Although they both deal with the significance of digitalisation for social work policy and practice, these strands are rarely combined since they focus on different types of technologies. We argue that there is a gap in understanding how the role of the client is produced when applying for social assistance through ADM systems, as the interaction between social workers and clients changes.

ADM in social work and social assistance

Studies on ADM and social work (Considine et al., 2022; Eubanks, 2018; Gillingham, 2020) focus on the ethical implications of using ADM systems in this field. For instance, Eubanks (2018) demonstrates how automated predictive models raise concerns about the potential for automating inequality and further ‘punishing’ clients with the fewest resources.

In Swedish research, there has been considerable focus on the use of ADM in social assistance (Germundsson, 2022; Kaun, 2021; Nordesjö et al., 2024; Ranerup and Henriksen, 2020; Ranerup and Svensson, 2022), as it was the first sector to undergo automation. Research highlights how ADM aligns with various goals and values (Busch and Henriksen, 2018; Germundsson, 2022, 2024; Ranerup and Henriksen, 2020; Veale and Brass, 2019), including ethical considerations (Ranerup and Svensson, 2023), and shows that ADM systems often prioritise and rank these values in novel ways. Much of the focus has also been on the drive to enhance efficiency in public administration and reduce costs (Nordesjö et al., 2024). ADM systems are also described as mediators of legislation (Wihlborg et al., 2016), specifically in the context of active social policy, where norms and values shape policy. In fact, the professional and ADM system co-create practices and meaning, and the former may choose to form an alliance with ADM or the client (Wihlborg et al., 2016).

Research has also examined shifting professional roles and competencies, the ethical risks of ADM for the social worker-client relationship, and the need for digital ethics (cf. Nordesjö et al., 2021). For example, by digitalising professional discretion, there is a risk of making complex ethical decisions invisible as the focus shifts to what is required to apply (Considine et al., 2022; cf. also Petersen et al., 2020). There is also a risk of creating new moral dilemmas for social workers who must balance policymakers’ demands for increased digitalised services and the needs of individuals (Pors and Schou, 2020).

Moreover, research has addressed which groups of clients are being included and excluded when services are being digitalised, and the social worker-client relationship is changing (Scaramuzzino et al., 2024). In Sweden, the decisions made in social assistance are means-tested and sometimes require more complex decisions than standardised ones delivered by ADM systems (Scaramuzzino et al., 2024). It has therefore been debated whether clients’ social rights and legal certainty can be assured with current ADM systems, as they also tend to generate both incorrect (Carlsson, 2023) and less generous decisions (Germundsson and Stranz, 2023).

Thus, although the goals of ADM systems are to increase efficiency, transparency, and to offer faster welfare delivery to clients, they affect the welfare encounter and risk becoming a threshold for obtaining one’s social rights. Also, they risk expanding the digital divide between groups who have high and low levels of access to digital technologies and skills to use them (Griffiths, 2024; cf. Goedhart et al., 2019; Ragnedda, 2017). For instance, when completing e-applications for social assistance through the ADM systems, clients have experienced barriers such as a lack of language and digital skills and not having the necessary technology (Scaramuzzino et al., 2024), and they may need the help of ‘proxy users’ such as family and friends (see Gallistl et al., 2021). There are often unforeseen negative consequences (Coles-Kemp et al., 2020), such as a greater ‘administrative burden’ for clients to apply (Griffiths, 2024: 11). Hence, these systems have ethical consequences for the possibilities of accessing social rights (cf. Germundsson and Stranz, 2023; Griffiths, 2024), and for the attainment of digital citizenship (Scaramuzzino et al., 2024). Consequently, it is important to design ADM systems that are responsive to the client’s circumstances. This has several potential benefits for clients, such as saving time and money travelling to the office and experiencing less stigmatisation (Considine et al., 2022; Scaramuzzino et al., 2024).

All in all, the literature on ADM systems in social work and public administration highlights the process of constructing meanings of automation, but not the role of ADM in welfare encounters and in constructing and producing the ‘client’ in welfare-to-work.

Digital documentation practices within social work

There is a long tradition of studying how client identities are produced and then reproduced through interactions, institutions, and policies in welfare bureaucracies (see e.g. Juhila and Abrams, 2011). Larsen (2008) shows how there are different popular images of target groups, and these images also influence how clients wish active labour market policy to be developed and performed. Research on digital documentation practices within social work has shown how these practices and systems should not be treated as “neutral carriers of information” but together socially construct “digital clients” (Martinell Barfoed, 2019: 196). Similarly, the asymmetric roles of the professional and client are constructed in a phone-mediated welfare encounter (Schmidt, 2024). Also, Sztandar-Sztanderska and Zielenska (2020) depict how the digital profiling of unemployed persons promoted certain demographic characteristics, individual motivation, and responsibility for finding work. A move towards digital self-services has shown similar results, making unemployed job seekers responsible for their case and situation (Wallinder and Seing, 2022; cf. Henman, 2010). We contribute to this line of literature by studying the construction of clients in ADM in welfare-to-work. We will focus on the production of client roles in the interaction between ADM, the caseworker, and the client.

People production through descriptive practice

Emerson and Messinger (1977) distinguished between the “world of troubles” and the “world of problems.” In the everyday world of troubles, individuals experience ailments that are often vague, anecdotal, and ill-defined, making it difficult to determine the precise cause of their distress. In contrast, the world of problems involves the assessment, classification, and problematisation of these troubles by professionals into problems. This process typically occurs at the frontline of policy implementation, where policies are shaped and performed (Lipsky, 2010). For Hasenfeld (1972: 1), people-processing organisations such as social assistance organisations attempt to “achieve changes in their clients not by altering basic personal attributes but by conferring on them a public status and relocating them in a new set of social circumstances.” This means assessing how clients’ conditions legitimise an action, evaluating clients to determine the different alternatives of an action, deciding on a choice among alternatives, and finally carrying out the alternative. Holstein (1992) expanded on the idea of people-processing, but sees the process as interactionally accomplished, rather than as a passive attribution of client statuses. The individual does not have a fixed identity when applying for social assistance, but one that is formed and ‘produced’ by the everyday actions and interactions taking place between people; for example, between the social worker and the client. How people deal with and describe their social worlds creates and constitutes meaningful phenomena.

Central to Holstein’s (1992) idea of people processing is descriptive practice. Human service workers—like caseworkers in a social assistance organisation—create descriptions by integrating perceptions of clients and their problems with certain organisational actions. These descriptions are always incomplete and potentially biased since they can never capture the full properties of a person and emphasise some aspects over others. However, the descriptions are related to the intentions and practical tasks of the organisation. They can therefore be understood as a goal-oriented practice which does not necessarily describe a reality as much as it assembles and manages it for a current practical purpose. Consequently, descriptions of clients give rise to rhetorically constructed personas to address the practical concerns at hand and legitimise and render certain actions relevant. In this sense, social workers are producers as much as treaters of people.

According to Holstein (1992), analysing a descriptive practice draws attention to the interactional and rhetorical activities that bring together locally situated understandings of the people and problems human service workers encounter. Despite the asymmetrical nature of the people-processing, where expert labelling intersects with client autonomy, this is not an entirely deterministic process, as clients contribute to or may actively resist imposed descriptions. For example, client descriptions such as active and passive client, and worthy and needy, are subject to negotiation and co-construction (Mik-Meyer and Silverman, 2019). Still, as Järvinen (2014) argues, resistance requires both determination and resources. The capacity to contest these descriptions is unevenly distributed, often depending on clients’ socio-economic positions, knowledge of the organisation, and access to alternative sources of support.

Earlier research on digitalisation and clients in social work has used Holstein’s (1992) idea of people production to understand how clients are constructed through forms and questionnaires by transforming binary data into storylines (Martinell Barfoed, 2019). In this article, we use the idea of people production to understand the role of ADM in welfare encounters and how these encounters shape and define client roles relevant to the organisation’s purpose, ultimately co-constructing social policy.

Cases and methods

The case, data collection, and data analysis are described below.

Two cases

In Sweden, municipal activation and social assistance services exhibit local variability and a lack of evidence regarding their efficiency (Forslund et al., 2019). The activation process is primarily incentivised by governmental economic policies, underpinned by the belief that activation initiatives mitigate social assistance costs (Ulmestig, 2007). The administration of social assistance applications is guided by individual means-testing regulations stipulated within the Social Services Act framework. Social assistance caseworkers and labour market secretaries at municipal job centres both control and support clients simultaneously (Van Berkel and Van der Aa, 2012) by monitoring their job applications and finances while also assisting with job searches, skills, and self-esteem.

The study was carried out in two Swedish municipalities (population less than 50’000) that handle social assistance together with job centres and have a local labour market policy with an emphasis on activation measures. Before ADM, social assistance caseworkers manually processed paper applications, held in-person meetings, determined eligibility, and referred clients to job coaching. A significant portion of their time was dedicated to administrative tasks, which often resulted in clients waiting several days for updates on their applications.

At the time of the study, the municipalities used robotic process automation (RPA), which applies pre-programmed routines to structured data, such as an application for social assistance, to suggest an outcome like a decision (Ranerup and Svensson, 2022). This ‘e-application’ must be submitted digitally through a municipal website and requires applicants to provide information about their personal, employment, and financial circumstances. It constitutes the sole point of interaction between clients and ADM. The rationale behind the implementation of ADM in the two municipalities shares a common goal: enhancing time efficiency and fostering better client relationships. This suggests that ADM potentially shapes the dynamics of the interactions between caseworkers and clients. While both municipalities pursued similar objectives, they differ in the timing of ADM adoption, with MunA initiating its implementation several years ahead of MunB. In MunB, ADM primarily functions by translating applications into the caseworkers’ internal system, effectively serving as an automated support mechanism for caseworker assessment. Only MunA’s ADM is considered a fully implemented RPA since it suggests decisions from clients’ applications. This discrepancy in implementation sheds light on the variance in client production across different ADM setups. It is essential to note that this article’s aim is not comparative; rather, it seeks to leverage data and examples from both cases that are instrumental to our comprehension of how ADM contributes to client production (Stake, 2005).

Data collection

Data from 28 qualitative interviews with managers (5), quality personnel (3), professionals (9), and clients (11) in two municipalities were collected during 2021. Two department and three unit managers were included. Department managers had knowledge on the background and rationales of ADM, while unit managers had insights on ADM implementation. Quality personnel supported management and professionals in the day-to-day challenges of ADM and had technical expertise in ADM. While neither managers nor quality personnel met with clients, their perspectives contributed with detailed knowledge on ADM’s purpose in the welfare encounter. Professionals could be either caseworkers or labour market secretaries who worked with ADM and met with clients to varying extents in the two municipalities. While caseworkers have bachelor’s degrees in social work, the educational backgrounds of labour market secretaries are more diverse, e.g., behavioural psychology and human resources. A criterion for inclusion for managers, quality personnel, and professionals was that they had been involved in ADM for at least 6 months. Finally, clients applied for social assistance electronically, had their data processed by ADM, and engaged with a professional in one way or another. Inclusion required Swedish proficiency and application via the e-application, ensuring exposure to ADM. This may have excluded clients with language barriers. Clients in MunB had received social assistance for a longer period than clients in MunA, likely giving them a better understanding of the system. However, all clients were new to the e-application process.

Interviews were carried out digitally via video calls due to the COVID-19 pandemic. An exception was interviews with clients in one of the municipalities, where interviews were performed on-site since it was the most feasible and ethical way to come into contact with clients. Most interviews lasted 40–90 minutes, while client interviews were shorter, lasting 20–40 minutes. Clients, managers, and quality personnel were interviewed individually. Professionals were interviewed in groups (two group interviews in each municipality) to facilitate fuller and more interactive descriptions of casework and ADM. In interviews with managers, quality personnel, and professionals, interview guides were semi-structured to encourage participants to reflect while ensuring their answers remained pertinent to the following themes: how and why ADM was implemented, how ADM is used in social assistance casework, and the significance of ADM for the relationship with the client. Client interviews focused on the experiences of applying online since they did not know anything about ADM. For example, questions were asked about whether the application process was understandable, easy, or empowering, and whether one could contact a caseworker if needed. All interviewees received information on project aims and interview procedures beforehand and gave verbal consent. Clients were given information about the study and asked to participate by professionals before their meeting, after which the interviewing researcher repeated the information, and the interview was performed. The research project was approved by a Swedish Regional Ethical Review Board (Dnr 2020–00114).

Data analysis

Grounded in a people-producing perspective on digitalisation in social work, we approached the data with an awareness of how ADM may construct client roles within the social worker–client relationship. This perspective directed our analytical focus to how interactions between clients, employees, and ADM took place and how clients were described in this process. Consistent with principles for thematic analysis (Braun and Clarke, 2012), descriptions were coded, and subsequently reformulated and regrouped, resulting in the emergence of descriptive subthemes. For example, clients were coded as “emotional”, “distanced”, “trustworthy”, “digitally skilled”, and “independent” and grouped and synthesised into several themes. After further regrouping and in dialogue with previous research and theory, three distinct themes in the form of client roles emerged. These roles represent interviewees’ descriptions of client properties – how a client should act in the welfare encounter in relation to the ADM process and organisation.

In the analysis, different actors’ responses were treated separately and are presented accordingly. Managers’ and quality personnel’s responses provided insights into the organisational purpose of ADM and broader rhetorical framings and expectations. In contrast, professionals’ responses offered concrete accounts of client encounters. Client interviews served as a counterbalance, either reinforcing or challenging professionals’ descriptive practices. As researchers in digital social work, we have observed both rushed and careful implementations of digital solutions, with varying consequences for client relationships. These experiences made us attentive to the diversity of client perspectives in relation to managerial and professional accounts. Together, the accounts gave rise to three client roles, understood as interactionally and rhetorically constructed personas in line with Holstein’s (1992) framework. These roles resemble ideal types, representing tendencies to which empirical cases may align to varying degrees.

Three client roles

The analysis is presented in three themes corresponding to three client roles. Before this, we describe how ADM has changed the welfare encounter.

First, the e-application process was designed as digital self-service, meaning clients were encouraged to seek out digital services previously administered by welfare professionals (Schou and Pors, 2019). New clients confirmed their identity with Bank-ID¹ and attached documents, while recurring clients supplied almost no documentation, only via random controls. In MunA, the e-application was processed by ADM by structuring data, suggesting a decision, and forwarding it to caseworkers and labour market secretaries. New clients could select a meeting time within 24 hours, during which a labour market secretary assessed whether clients were “at the disposal of the labour market.” This often meant applying for ten jobs every week. Meanwhile, caseworkers verified ADM’s calculations and decided on social assistance. In contrast, MunB’s ADM only compiled data from other resources for caseworkers to assess.

Second, caseworkers no longer needed direct client interaction to make decisions. In MunA, caseworkers only met with clients upon request, preferring communication via email or phone. Both municipalities redirected inquiries on applications to a customer service centre handling all municipal matters. When clients submitted questions, customer service forwarded them to caseworkers, who were expected to respond within 24 hours. Caseworkers in MunA described this shift as a relief, as it increased their control over time and interactions. Meeting clients was also described as administratively unnecessary, emphasising it as relevant only when tied to practical tasks in the organisation (Holstein, 1992).

Third, both municipalities experienced increased time efficiency, a key argument for ADM adoption (Nordesjö et al., 2024). Caseworkers (MunB) and labour market secretaries (MunA) described the additional time for client relations as luxury time. For example, unlike caseworkers, labour market secretaries in MunA met clients weekly, focusing on educational and motivational support rather than financial assessments. In MunB, these roles were not divided.

The distant client

The distant client refers to caseworkers’ desires to have less client interaction to carry out a more efficient and impartial decision-making process. Caseworkers in both municipalities described how, before ADM, clients tended to disturb the decision-making process by wanting to explain their situation or negotiate their rejected application. The client’s economy was described as an “emotional subject”, giving rise to arguments, conflicts, and attempts at negotiation:

It was quite common when there is open access to your social worker that perhaps in your desperate situation, you play on emotions. Whether it's threatening or negotiating or crying, it's individual maybe, but it's surely a problem in many municipalities. (Caseworker 1, MunA)

The caseworkers in MunA seemed to find that the decision-making process was more straightforward when using ADM, and that it benefited from being carried out in ‘isolation’. Two caseworkers elaborate:

Caseworker 3, MunA: As a caseworker, you can make your decisions and think them through calmly. You're not influenced by anything. There were so many different emotions sometimes that could take over and drain your energy. That's gone at present.

Interviewer: Hm

Caseworker 2, MunA: Yes, both conflicts, one’s own emotions and values don’t need to be involved. There’s a personal identity number on the screen, and we can make a legally secure decision without negotiation or threats or…

Caseworker 3, MunA: Emotions and tears, yes.

This reasoning illustrates a descriptive practice where clients are seen as potential disturbances to caseworkers’ decision-making process. Instead, the distant client serves a purpose to the organisation by facilitating an impartial and efficient decision-making process free from emotions and negotiation for both caseworkers and clients. Similarly, caseworkers in MunB described how ADM, on one hand, led to more time for client meetings, but on the other hand, caseworkers questioned what such meetings contributed:

But we talk very little about what we should fill that time with. What should we talk about during the meetings? What is it about the client meeting that is so magical, as everyone emphasises? (Caseworker 4, MunB)

This description adds to the distant client by re-evaluating the view of the social worker–client relationship as an integral part of social workers’ professional identity and purpose (Rollins, 2020). It is also consistent with an administrative perspective, where, according to one caseworker, there is no real point to meeting clients:

I mean, administratively, there's no purpose in us meeting or having a relationship. That relationship is managed well by the labour market secretary. We trust them to do their job. However, it doesn't prevent [clients] from having other contacts. For example, at the social services department, if there are other social issues. But those aspects are not managed by us. (Caseworker 1, MunA)

In the quote, the caseworker displaces the responsibility for the welfare encounter onto the labour market secretary. By referring to other contacts, it minimises the caseworker’s role and further distances the client.

Clients’ experiences related to the distant client in MunA describe how they had difficulties contacting caseworkers. Clients either gave up or had to wait in a phone queue for a long time:

The only thing that's a bit more inconvenient online is that if you have to call them, there may be around 20 other people in line, so you end up waiting for two hours before getting through. That's what happened to me yesterday when I called them. Since I'm already registered here, should I attend this meeting or not? It took half an hour to get an answer to that. (Client 1, MunA)

Also, several clients said that they appreciated personal help and would like to explain their application:

I would have liked to develop certain things a bit more. Explain. When you apply in the future, it may flow better. But the first time you want to explain. Like now, “you have only applied for so many jobs…” Yes...but there was a reason, but I haven't been able to write the reasons yet…and she's not really listening so… (smiles) (Client 3, MunA)

The quote illustrates the potentially different perspectives of caseworkers and clients, which gain significance when it comes to explaining and negotiating the application. The ADM system was not someone with whom the client could negotiate or explain their application. In this regard, ADM systems must be designed to be responsive to the unemployed individuals’ circumstances (Carlsson, 2023), but not all clients found the process so. In contrast, although clients in MunB also described the application process as distant and solitary, they experienced that they could more easily contact their caseworker by phone to ask questions:

I’ve reached the age of 57 now… it’s been a couple of years … it’s something new (e-application). How does it work, like this and that… but [the caseworker] was very helpful there, “if any problems arise, just call… come in and…” there were even people in [the municipality] you could meet in person or talk to on the phone when they implemented the system. That way, you could get help. (Client 3, MunB)

ADM plays a role in creating the distant client by minimising interactions with clients, who are perceived as emotionally driven and disruptive to caseworkers’ decision-making. By reducing direct communication, ADM reinforces the portrayal of clients as detached, administrative entities with a focus on application requirements rather than individuals with complex needs (Considine et al., 2022). It also reduces the possibilities for negotiations and resistance in relation to both the application and the client role (see Järvinen, 2014; Mik-Meyer and Silverman, 2019). The distant client emerges not only from reduced interactions but also from a broader institutional process that aligns clients with organisational priorities such as efficient decision-making. This underscores how ADM can contribute to a descriptive practice which not only reflects but actively shapes social realities (Holstein, 1992).

The capable and trustworthy client

The third client role refers to a grown-up and responsible citizen working independently towards self-sufficiency. ADM contributes to this role in two ways, through the closely related labour market and activation perspective, and the reduction of client documentation in recurring applications.

In both municipalities, and especially in the interviews with managers, it was clear that the implementation and use of ADM not only changed the application and decision-making process but also constituted a possibility to contribute to less control and lead to less stigmatisation for applicants (see Carlsson, 2023). “It’s in the walls,” manager 1 in MunA said, alluding to the atmosphere of suspicion, scrutiny, and stigmatisation regarding clients. A central part of changing this atmosphere was to use other words in the workplace and client meetings, thereby challenging old values and practices: from “obstacles,” “client,” “low-educated,” and “weak in language,” to “challenges,” “job seeker,” “short-educated,” and “multilingual”. A client description in MunA was that the clients’ problem was a lack of income, not the presence of social obstacles. A labour market perspective was therefore also much preferable than a more social work perspective in this municipality:

I come from social work, and I've worked quite a bit with emotions, feelings, and ‘curling’, which I maybe thought, well, this is how it should be, and I didn't know any different. So, for me, I think that this change has also contributed to personal development in terms of what one stands for and what one considers important. (Labour market secretary 1, MunA)

Employees’ views of clients had also changed in this process. Clients were described as trustworthy individuals with capabilities, which also had implications for client interactions. Just like in a job interview, a manager said, they don’t ask about health and such, and the client answers in a way they find relevant. This approach prioritised job seeking over the client’s financial situation:

We focus on service instead of control. It makes a difference in the encounter, you know. But also, we have high expectations. We don't tilt our heads and feel sorry for people. It's important that we have high expectations and believe in people, that people can and want to, but also take responsibility for the choices they make. (Manager 1, MunA)

This rhetoric promoted to the production of the capable and trustworthy client. Several clients contributed to this description. One client appreciated the support and ‘push’ to get back to work, and on principle did not appeal:

Client 4, MunA: [Appeal] the total amount...? No, I don’t think like that. I'm satisfied with what I get. I don't call in to complain about getting a hundred SEK more (about 10 euros). I'm satisfied with the help I receive. One should be grateful.

Interviewer: Do you feel that you have had power and control over your application?

Client 4, MunA: Yes, I feel that...you have to fill it out yourself...you have to be honest with everything you fill in, it will eventually come out that something doesn't match. I feel it's 50-50 anyway. It should be correct.

The capable and trustworthy client resonates with literature discussing digitalisation and the role of clients in becoming active welfare subjects (e.g. Henman, 2010) and responsible for their case and situation (Wallinder and Seing, 2022). ADM reinforces this role, which supports the organisation’s purpose by emphasising labour market secretaries and an activation perspective in client interactions.

The capable and trustworthy client’s purpose within the organisation was also important, considering both municipalities had removed the control of documentation for recurring monthly applications for social assistance. In MunB, where caseworkers also acted as supporters, caseworkers argued that although they were not used to it, they tried to trust the client more today, until proven otherwise. While this could lead to disappointment, it could also improve client relationships:

One could say that it's easier to build a relationship if one doesn't have to be someone who doubts, “Have you really submitted the right thing, this looks a bit suspicious”, and all that. By letting go of the requirement that we should be controllers, we have the opportunity to instead be a support for our clients and help them with planning. (Caseworker 4, MunB)

Most clients described the removal of documentation control as beneficial and felt that the occasional random check was manageable:

I don’t need to enter the electricity bills every time, it's only if there’s a random check, right? Then, I should be able to go back in time to make sure I haven’t requested too much. /.../ I got such a random check and requested one month back. Then I called the bank… it was during the pandemic, but… I remember I received it digitally from the bank. I did the same with the papers, I took photos of them and emailed them. It worked. (Client 3, MunB)

Other clients argued that the removal of documentation control deprived them of their ability to negotiate and demonstrate their trustworthiness and eligibility for certain benefits, while also ensuring there were no mistakes in the application.

Yes, absolutely, I have control and all that... but at the same time, it feels very unfamiliar and uncertain. Have I really done it in the right way? So, it feels more uncertain in that way. Also, I don't like the internet in that way either. (Client 3, MunA)

In this sense, the perceived control of applying online also evoked feelings of uncertainty.

The implementation and use of ADM in both municipalities contribute to the capable and trustworthy client through the removal of control for documentation and by strengthening a labour market perspective in social assistance. Drawing on Holstein’s (1992) concept of descriptive practice, client descriptions do not portray individuals as passive recipients of aid, but rather as individuals with the potential to meet expectations and participate in the workforce. While this client persona is generally recognised by clients, the removal of documentation control is not universally appreciated since it deprives clients of a resource for negotiation and proving their worthiness (cf. Mik-Meyer and Silverman, 2019). Still, the portrayal of clients as “capable” aligns with broader institutional narratives, serving a purpose within the social assistance organisation by emphasising client characteristics that are vital for welfare-to-work initiatives.

The digitally skilled client

The digitally skilled client is a client who can interact with the application system, ADM, and therefore the organisation. Both municipalities had a high percentage of clients applying online (80%). This was possible by having employees in the reception helping clients who needed support in applying. Caseworkers described how clients appreciated filling in e-applications more than paper applications. One caseworker argued that the introduction of a simple tool like BankID made clients feel empowered. E-applications were generally described by caseworkers as an efficient and convenient system for both caseworkers and clients:

At least I've perceived that for the applicant, they prefer this system, they do it digitally, and they get the money the day after. They don't have to submit documents. I think of it as a win-win. You can lie at home on the couch and apply for social assistance. (Caseworker 2, MunA)

Most of the clients in both municipalities found the e-application easy to understand and more convenient than the old paper application. They could apply from wherever they were and did not have to keep track of the department’s opening hours or different papers that tended to get lost. Being taught to be digitally skilled was discussed by clients who stated “it’s not complicated once you’ve done it once, you know” (Client 1, MunB), and expressed that it was “a bit cumbersome before getting the hang of it” (Client 2, MunB).

However, not everyone could apply online and meet the expectations of the client role. Clients with protected personal information and those who had a legal guardian still used paper applications, since the client needed to use BankID. Some clients were also afraid of using the technology, as one caseworker argued:

I think it's partly due to a bit of fear of technology. People feel uncertain about how to use it. How can I then support a client who may not know how to use it at all? It requires a bit more training in that regard, and perhaps we need to invest more in developing digital tools so that we feel confident in using them. But it's not everyone. Many people do use it. (Caseworker 4, MunB)

The caseworker thus argued for supporting clients, not only to become self-sufficient, but also to support them in applying for support. Drawing on Holstein (1992), this descriptive practice represents an alteration of the client role, which now needs to have certain skills to access the organisation and its services.

Even though most clients expressed that they preferred to apply for social assistance online, others described that they found filling in the e-application to be both complicated and stressful.

It was really difficult. I've always applied on paper before. So now, I had to log in with BankID, enter personal information, attach the lease agreement, jobs, and bank statements, and then I had to download it to my phone and then attach the bank statements... there were maybe thirty attachments... just for the bank statements. Because I mean, I couldn't see much on my small phone, and I don't have a computer... so I was like... taking screenshots of everything. All jobs. Rent...everything had to be photographed and attached... so I sat there for two hours with just that. I just... oh well... then you have to check the emails where I get all the rents... go to the housing benefits... then back to the application. Three months back. I would have preferred to print it out and send it in on paper. (Client 3, MunA)

This client ultimately managed to apply through guesswork and, like other clients in MunA, with the help of a family member. Getting the help of someone else or letting them apply for you, a form of proxy usage (e.g. Gallistl et al., 2021), seems to help in overcoming most of the obstacles and barriers associated with an e-application. Although speaking to a caseworker was possible through customer service in MunA, it does not seem to be known or used by all clients. In this sense, the clients lacked the necessary knowledge and resources to fulfil the expectation of digital skills associated with the client role (cf. Järvinen, 2014). Instead, the impression was that filling in an e-application was something you did alone at home. In MunB, where caseworkers were described by the clients as ‘easy to get a hold of’, the interviewed clients typically said that it was easy to get the caseworker’s help when they had a question or a problem applying online. However, getting the help of others was common if clients did not have sufficient language skills to understand the e-application. One client in MunA with limited language proficiency relied on her brother for assistance, as she struggled to understand the questions and would have preferred in-person support:

“What is better is coming here and filling out the application. (I: Why?) Because you understand better, you get help, and you understand what they want and so on...” (Client 2, MunA).

Limited language proficiency thus impedes a client's ability to navigate the expectations and purposes set by the organisation. It also creates barriers to understanding their social and legal rights (Scaramuzzino et al., 2024), not least since there are discrepancies between the paper and e-application formats, which may lead to confusion:

It's the language and the words used, and how to fill it in. It's not the same when filling out on paper as it is on e-services. You need a slightly higher level of language when filling out e-services /…/ You have to read the terms to know what rights you have as an applicant. You need to have a higher level of language proficiency to be able to understand what rights you have. (Client 5, MunB)

Thus, the design and operation of the ADM systems created thresholds for clients with lower levels of Swedish language proficiency. Hence, the production of the digitally skilled client risks further excluding those who do not meet expectations due to language barriers (cf. Ragnedda, 2017; Scaramuzzino et al., 2024; Schou and Pors, 2019).

The descriptive practice of caseworkers illustrates that clients must develop digital skills to navigate the e-application process and fulfil organisational expectations. Following Holstein (1992), these descriptions do not merely reflect client attributes but actively construct them through professional-client interactions. This process generates a ‘rhetorical production’ (Martinell Barfoed, 2019) of client roles based on digital competence, requiring clients to acquire skills and support to adapt. Various technical, economic, and social barriers (see e.g., Scaramuzzino et al., 2024) may hinder this transition, making digital proficiency a crucial yet unevenly distributed resource. This lack of resources may limit clients’ abilities to negotiate or resist during the application process (cf. Järvinen, 2014). As a result, clients must not only demonstrate financial need but also navigate a digitalised welfare system, reinforcing an institutional norm that prioritises technological proficiency in the welfare-to-work transition. The ability to use digital tools is not merely a practical requirement to access benefits but becomes a defining criterion in the clientisation process, shaping who is included, excluded, and a digital citizen.

Discussion

This article has explored the role of ADM in the transition between the “world of troubles” and the “world of problems” (Emerson and Messinger, 1977). It has examined how the role of the client is produced in the welfare encounter between employees and clients when applying for social assistance in two Swedish social assistance organisations using ADM. Our findings show that the welfare encounter has been structured as a self-service desk, resulting in, on the one hand, no need for client interaction in the application process, and on the other hand, luxury time for activation for those working with support and motivation. This structure of interaction has contributed to a descriptive practice that promotes desirable client characteristics in line with the intentions of the ADM system, producing three client roles: (1) the distant, (2) the capable and trustworthy, and (3) the digitally skilled. The implications are explored in the following discussion.

The use of ADM has included new ways of talking, thinking about, and interacting with clients that are intended to align them with the purposes of the organisation. The client roles tell us that the ADM era poses new challenges for clients, such as independently understanding and managing a self-service application system, which requires both digital skills and a certain level of proficiency in the Swedish language in order to apply for social assistance. Hence, there are exclusionary mechanisms embedded in the client roles produced in relation to ADM, especially the role of the digitally skilled client, that are important to recognise and address further.

These findings align with research demonstrating how digital tools can raise the threshold for accessing social rights for certain client groups without knowledge and resources, contributing to social exclusion and expanding digital divides (cf. Griffiths, 2024; cf. Goedhart et al., 2019; Ragnedda, 2017; Schou and Pors, 2019; Scaramuzzino et al., 2024). Clients may have to turn to friends and family and rely on proxy agency (e.g. Gallistl et al., 2021).

Moreover, the reduced need for client interaction challenges the ideal of the social worker–client relationship (Rollins, 2020). Digital tools can maintain and change relationships in positive ways and improve accessibility to welfare services on the clients’ terms (Nordesjö et al., 2021). The distant client can reduce stigma by minimising caseworker interactions and documentation requirements while simultaneously improving access to e-application systems in a more flexible and client-centred manner (cf. Considine et al., 2022). However, this increased distance can also restrict communication with caseworkers, making it more difficult for clients to seek assistance, discuss their cases, or negotiate their applications.

This shift aligns capable and trustworthy clients with an active labour market policy and welfare-to-work agenda (cf. Larsen, 2008), emphasising targeted support and motivation for those deemed employable. It reinforces a policy focus on individual agency and the role of personal responsibility and self-sufficiency in securing employment (Henman, 2010; Sztandar-Sztanderska and Zielenska, 2020). Furthermore, this transition is grounded in the assumption that caseworkers’ roles in control and support can be functionally separated (cf. Van Berkel and Van der Aa, 2012), whereby ADM manages clients’ financial situations, while labour market professionals concentrate on fostering motivation and employability.

In this regard, we argue that ADM not only reflects organisational ambitions but also actively shapes the role and direction of social policy. ADM contributes to digital transformation (Wessel et al., 2021) by reshaping the welfare encounter and reframing the identity of human service organisations. Rather than emphasising direct, person-centred social work, ADM shifts the organisation’s focus toward distanced, automated processing of individuals for activation. While this shift may reduce stigma for some, it also reduces opportunities for negotiation and resistance within welfare interactions and the social worker-client relationship.

We highlight three implications for social work policy and practice. First, implementers of ADM systems should consider how to be responsive to clients’ circumstances to improve access to welfare services through ADM. Clients should have contact information for their caseworker and receive support in navigating e-applications, or welfare bureaucracies risk excluding vulnerable groups. Second, while ADM may help reduce stigma through distanced welfare services, social workers should find ways to maintain positive social worker-client relationships for those clients who seek them. Policymakers should consider the professional values and competencies of social work in welfare-to-work activities, as these cannot be fully replaced by job coaches or labour market administrators (cf. Carlsson, 2023; Ranerup and Svensson, 2023). Third, social work management should make it mandatory to involve professionals and clients, or their representatives, in the design and implementation of ADM systems to make them as inclusive and equitable as possible.

These implications should be understood within the limits of the study. This study focuses on two Swedish municipalities with the autonomy to decide whether and how to implement ADM. Together with a limited number of interviews, these findings should therefore be generalised with caution. Second, it is important to consider to what extent ADM has contributed to the production of the three client roles or if these roles had evolved regardless of, or before, the implementation of ADM. Some scholars (e.g. Svensson, 2020) argue that the implementation of ADM is primarily a matter of efficiency rather than a change in work processes in social assistance organisations, since casework was already standardised before ADM. However, our interviews suggest that there are substantial changes in work practices as a direct consequence of the implementation and use of ADM.

Future research could investigate the effects these roles have on labour market integration and how clients are supported in applying for social assistance online. Research on the early stages of the design and implementation process of ADM could deepen our understanding of how the values of the organisation, profession, and client are negotiated and prioritised in digital social work. This would illuminate the values and interests that shape welfare encounters, reveal possible biases and power imbalances, and clarify the conditions needed for ADM to strengthen, rather than weaken, the social worker-client relationship.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Kettil Nordesjö

Gabriella Scaramuzzino

Note

References

Barfoed

(2019) Digital clients: an example of people production in social work. Social Inclusion 7(1): 196–206.

Braun

Clarke

(2012) Thematic analysis. In: Cooper

Camic

Long

, et al. (eds) APA Handbooks in Psychology. APA Handbook of Research Methods in Psychology, Vol. 2. Research Designs: Quantitative, Qualitative, Neuropsychological, and Biological. Washington, DC: American Psychological Association, 57–71.

Busch

Henriksen

(2018) Digital discretion: a systematic literature review of ICT and street-level discretion. Information Polity 23(1): 3–28.

Carlsson

(2023) Legal certainty in automated decision-making in welfare services. Public Policy and Administration 40(2): 302–321.

Coles-Kemp

Ashenden

Morris

, et al. (2020) Digital welfare: designing for more nuanced forms of access. Policy Design and Practice 3(2): 177–188.

Considine

McGann

Ball

, et al. (2022) Can robots understand welfare? Exploring machine bureaucracies in welfare-to-work. Journal of Social Policy 51(3): 519–534.

Emerson

Messinger

(1977) The micro-politics of trouble. Social Problems 25(2): 121–134.

Eubanks

(2018) Automating Inequality: How High-Tech Tools Profile, Police, and Punish the Poor. New York City: St. Martin's Press.

Forslund

Pello-Esso

Ulmestig

, et al. (2019) Kommunal arbetsmarknadspolitik. Vad och för vem? En beskrivning utifrån ett unikt datamaterial. Uppsala: IFAU Rapport, Vol. 5.

10.

Gallistl

Rohner

Hengl

, et al. (2021) Doing digital exclusion – technology practices of older internet non-users. Journal of Aging Studies 59: 100973.

11.

Germundsson

(2022) Promoting the digital future: the construction of digital automation in Swedish policy discourse on social assistance. Critical Policy Studies 16: 478–496.

12.

Germundsson

(2024) Configuring social assistance: conceptualizations and implications of the adoption of robotic process automation in the Swedish personal social services. European Journal of Social Work 1–13.

13.

Germundsson

Stranz

(2023) Automating social assistance: exploring the use of robotic process automation in the Swedish personal social services. International Journal of Social Welfare 33: 647–658.

14.

Gillingham

(2020) Algorithmically based decision support tools: skeptical thinking about the inclusion of previous involvement. Practice: Social Work in Action 33: 37–50.

15.

Goedhart

Broerse

JEW

Kattouw

, et al. (2019) Just having a computer doesn’t make sense’: the digital divide from the perspective of mothers with a low socio-economic position. New Media and Society 21(11/12): 2347–2365.

16.

Griffiths

(2024) Universal credit and automated decision making: a case of the digital tail wagging the policy dog? Social Policy and Society 23(1): 1–18.

17.

Hasenfeld

(1972) People processing organizations: an exchange approach. American Sociological Review 37(3): 256–263.

18.

Hasenfeld

(2010) The attributes of human service organizations. In: Hasenfeld

(ed) Human Services as Complex Organizations. Los Angeles: Sage.

19.

Henman

(2010) Governing Electronically: E-Government and the Reconfiguration of Public Administration, Policy and Power. London, UK: Springer.

20.

Holstein

(1992) Producing people: descriptive practice in human service work. Current Research on Occupations and Professions 7: 23–39.

21.

Järvinen

(2014) Untidy clientization: drug users resisting institutional identities. In: Gubrium

Järvinen

(eds) Turning Troubles into Problems: Clientization in Human Services. London: Routledge, 50–64.

22.

Juhila

Abrams

(2011) Special issue editorial: constructing identities in social work settings. Qualitative Social Work 10(3): 277–292.

23.

Kaun

(2021) Suing the algorithm: the mundanization of automated decision-making in public services through litigation. Information, Communication & Society 25: 2046–2062.

24.

Larsen

(2008) The political logic of labour market reforms and popular images of target groups. Journal of European Social Policy 18(1): 50–63.

25.

Lipsky

(2010) Street-level Bureaucracy: Dilemmas of the Individual in Public Service. New York, NY: Russell Sage Foundation.

26.

Meilvang

Dahler

(2024) Decision support and algorithmic support: the construction of algorithms and professional discretion in social work. European Journal of Social Work 27(1): 30–42.

27.

Mik‐Meyer

Silverman

(2019) Agency and clientship in public encounters: co‐constructing ‘neediness’ and ‘worthiness’ in shelter placement meetings. British Journal of Sociology 70(5): 1640–1660.

28.

Nordesjö

Scaramuzzino

Ulmestig

(2021) The social worker-client relationship in the digital era: a configurative literature review. European Journal of Social Work 25(2): 303–315.

29.

Nordesjö

Ulmestig

Scaramuzzino

(2024) Saving time for activation or relationships? The legitimation and performance of automated decision-making for time efficiency in two street-level bureaucracies serving poor and unemployed clients. Nordic Social Work Research 14(2): 209–221.

30.

Petersen

ACM

Christensen

Hildebrandt

(2020) The role of discretion in the age of automation. Computer Supported Cooperative Work 29: 303–333.

31.

Pors

Schou

(2020) Street-level morality at the digital frontlines: an ethnographic study of moral mediation in welfare work. Administrative Theory and Praxis 43(2): 154–171.

32.

Ragnedda

(2017) The Third Digital Divide: A Weberian Approach to Digital Inequalities. New York: Routledge, Taylor & Francis Group.

33.

Ranerup

Henriksen

(2020) Digital discretion: unpacking human and technological agency in automated decision making in Sweden’s social services. Social Science Computer Review 40(2): 445–461.

34.

Ranerup

Svensson

(2022) Value positions in the implementation of automated decision-making in social assistance. Nordic Social Work Research 14: 471–485.

35.

Ranerup

Svensson

(2023) Automated decision-making, discretion and public values: a case study of two municipalities and their case management of social assistance. European Journal of Social Work 26(5): 948–962.

36.

Rollins

(2020) Social worker–client relationships: social worker perspectives. Australian Social Work 73(4): 395–407.

37.

Scaramuzzino

(2019) Socialarbetare om automatisering i socialt arbete: en webbenkätundersökning. Research Reports in Social Work. Lund, Sweden: Socialhögskolan. Lunds Universitet, 3.

38.

Scaramuzzino

Nordesjö

Ulmestig

(2024) Citizens’ experiences of enablers and barriers to obtaining digital citizenship: E-applications for social assistance. Journal of Comparative Social Work 19(1): 41–67.

39.

Schmidt

(2024) The role of phone mediation: social workers’ and vulnerable clients’ role performances in mediated welfare encounters. Qualitative Social Work.

40.

Schou

Pors

(2019) Digital by default? A qualitative study of exclusion in digitalised welfare. Social Policy and Administration 53(3): 464–477.

41.

Stake

(2005) Qualitative case studies. In: Denzin

Lincoln

(eds) The Sage Handbook of Qualitative Research. Thousand Oaks, Calif: Sage.

42.

Svensson

(2020) Automatisering – till nytta eller fördärv? Socialvetenskaplig tidskrift 26(3-4): 341–362.

43.

Sztandar-Sztanderska

Zielenska

(2020) What makes an ideal unemployed person? Values and norms encapsulated in a computerized profiling tool. Social Work and Society 18(1).

44.

Ulmestig

(2007) På gränsen till fattigvård: en studie om arbetsmarknadspolitik och socialbidrag. Lund: Lunds universitet.

45.

van Berkel

(2017) State of the art in frontline studies of welfare-to-work: a literature review. Frontline Delivery of Welfare-To-Work Policies in Europe. New York: Routledge, 12–35.

46.

Van Berkel

Van der Aa

(2012) Activation work: policy programme administration or professional service provision? Journal of Social Policy 41(3): 493–510.

47.

Veale

Brass

(2019) Administration by Algorithm? Public Management meets Public Sector Machine Learning. In: Yeung

Lodge

(eds). Algorithmic Regulation. Oxford, UK: Oxford University Press, 121–149.

48.

Wallinder

Seing

(2022) When the client becomes her own caseworker: dislocation of responsibility through digital self-support in the Swedish Public Employment Service. Sozialer Fortschritt 71(6-7): 405–423.

49.

Wessel

Baiyere

Ologeanu-Taddei

, et al. (2021) Unpacking the difference between digital transformation and IT-enabled organizational transformation. Journal of the Association for Information Systems 22(1): 102–129.

50.

Wihlborg

Larsson

Hedström

(2016) “The computer says No!”--A case study on automated decision-making in public authorities. In: 2016 49th Hawaii international conference on system sciences (HICSS), 5–8 Jan 2016, Koloa, HI, USA, IEEE, 2903–2912.