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

ABSTRACT In the last decade, digitalized automated decision-making (ADM) has been implemented in many Swedish municipal social services to achieve values such as legal security, client empowerment and time efficiency. The paper aims to understand how ADM policy is legitimized and performed through time efficiency, by a comparison of ADM policy in two Swedish municipalities’ social assistance agencies. It builds on 17 interviews with managers and professionals in two Swedish municipalities’ social assistance units. Findings show ADM is legitimized through arguments of activation and relationships, and performed by handling more applications or increasing time spent with clients, rather than being perceived as increasing the quality of social assistance services. This highlights the significance of organizational goals regarding how street-level bureaucrats perform tasks within their discretionary powers.


Introduction
Automated decision-making (ADM), where citizens' eligibility to public services and benefits are assessed through computerized algorithms, have spread to different parts of welfare administration.This article centres around how ADM can shape social work by investigating how managers, social workers and job coaches engage with ADM policy in their everyday work.In Sweden, an area that has been subject to ADM is municipal social assistance organizations, where social workers exercise public authority for the poorest citizens as the last safety-net in the Swedish welfare system (Ulmestig and Nordesjö 2019).An individual means-test differentiates the system from socialinsurance but gives the system legitimacy in a mature welfare state (Hvinden and Johansson 2007).Jobcentres are organized for social assistance recipients and other unemployed individuals who meet high thresholds to enter the labour market.
Like all human service organizations, social assistance organizations seek legitimacy and resources from their surroundings by aligning themselves to often inconsistent and contradictory societal norms and values (Hasenfeld 2010).Recent research has shown how ADM is related to several goals and values (Busch and Henriksen 2018;Veale and Brass 2019;Ranerup and Henriksen 2022;Germundsson 2022;Ranerup and Svensson 2022).ADM should safeguard citizens' legal security through transparent, predictable, and fair decision-making by avoiding potential biases like personal moods.Another goal is to empower citizens and make street-level officials more accountable and responsive to individual needs.A third goal, and the topic for this article, is how ADM is intended to make social assistance time efficient for the professional and organization; i.e. to perform social assistance tasks as well as possible with as little actual time waste as possible.Automating time-consuming administrative tasks can reduce caseloads and save resources that social workers can use to direct attention to clients' needs and other core tasks of social work.
Social workers who interact with clients are street-level bureaucrats, with substantial discretionary powers.Although they do not determine explicit policy content, they can be understood as policymakers as they perform and informally construct policy during everyday organizational life (Lipsky 2010;Van Berkel et al. 2017).The goal of time efficiency in ADM policy is thus something that is performed actively in street-level bureaucracies within bureaucrats' discretion.On the one hand, research on 'digital discretion'; i.e. using 'computerized routines and analyses to influence or replace human judgement' (Busch and Henriksen 2018, 4), highlights how ADM may challenge and curtail street-level bureaucrats' discretionary powers (Buffat 2015).On the other hand, street-level bureaucrats can also cope with new digital technology by bending and adjusting them to their own practice (Devlieghere, Roose, and Evans 2020;Breit et al. 2021) and are able to respond to situational factors (Petersen, Christensen, and Hildebrandt 2020) such as through organizational context, automated technology and the specific tasks being replaced (Buffat 2015;Bullock 2019;de Boer 2021).
While previous studies argue that time efficiency is an important value in ADM, knowledge is limited on how ADM policy is legitimized and performed through time efficiency.The aim of this study is to understand how ADM policy is legitimized and performed through time efficiency, by comparing ADM policy in two Swedish municipalities' social assistance agencies, where poor and unemployed clients are subject to means-tests as part of the last safety net of the Swedish welfare state.

Time efficiency and discretion in ADM
Scholars have highlighted how ADM holds promises of greater efficiency, regarding both the quantity (volume) and quality (i.e.fairness) in decision-making (Busch 2019;Busch and Henriksen 2018;Kaun 2021).ADM is often motivated by its potential to increase both productive efficiency and economic efficiency (Germundsson 2022) to get value for (tax) money (Ranerup and Svensson 2022).The time and money gained from slow administrative routines can be spent in more beneficial ways, such as increased client interaction (Scaramuzzino 2019;Constantino et al. 2021).Overall, studies on ADM have focused on efficiency in relation to public values (e.g.Busch and Henriksen 2018), on efficiency as a value-position (e.g.Ranerup and Svensson 2022), and on how automated decision-making affects social workers' discretionary power (Scaramuzzino 2019;Ranerup & Henriksen, 2022;Bernhard and Wihlborg 2022).
Scholars have focused extensively on how the use of digital tools tends to challenge and reduce the discretionary freedom of social workers (e.g Petersen, Christensen, and Hildebrandt 2020;Bernhard and Wihlborg 2022).In line with the 'curtailment' thesis (Buffat 2015), Busch and Henriksen (2018, 18) find that digital discretion 'in general fails in its attempts to strengthen professional and relational values'.Moreover, scholars have highlighted how 'street-level bureaucracies' have turned to 'screen-level bureaucracies' by moving decision-making from the 'streetlevel' (e.g.social workers) to 'system-level' bureaucracies (e.g.programme developers) (cf.Bovens and Zouridis 2002;Petersen, Christensen, and Hildebrandt 2020;Zouridis, van Eck, and Bovens 2020).Still, this shift to system-level bureaucracies is not obvious, because ADM tools do not always make decisions independently.ADM tools may be programmed to provide information to make professionals' decision process more efficient (i.e. more cases, or faster, more transparent and coherent).In fact, Ranerup and Henriksen (2022) describe automated decision-making like that used in Swedish municipalities as a 'human -technology hybrid actor'.Hence, according to previous research, ADM can reduce some aspects of discretion in some areas and increase other aspects of discretion in other areas.
Within the literature, there is also an 'enablement' thesis that finds the curtailment thesis to be simplistic and deterministic as a single explanation.Instead, it insists that digitalization is only one factor shaping discretion, and the context of implementation should be of increasing interest (Buffat 2015;Busch and Henriksen 2018;Bullock 2019;de Boer 2021).We agree with this critique, and as we will show in the next section, the ways in the notion of 'digital discretion' have been used and understood in previous research, is often far from Lipsky's (2010) traditional understanding of street-level discretion.In this paper, we understand ADM as a contextual factor that shapes streetlevel bureaucrat's discretion.Similarly, time efficiency as an organizational goal is another contextual factor that may have significance for how discretion is shaped and consequently for how street-level bureaucrats perform ADM.

Legitimacy, discretion, and time efficiency in street-level bureaucracies
We draw on theoretical propositions related to legitimacy, discretion, and time efficiency.Like all human service organizations, social assistance organizations aim to prove their legitimacy to gather public support and funding (Hasenfeld 2010).This task is difficult due to the complexities of social problems and interventions which are often difficult to define and measure.Thus, social assistance organizations are forced to answer to and balance inconsistent and contradictory societal norms and values.This reflects the contradictory impulses in society that street-level agencies serve (Lipsky 2010, pp. 164-165).In our study, municipalities seek to balance time efficiency with acceptable social assistance values in the performance of ADM.Providing social assistance to vulnerable citizens is complex work, and cannot be completely regulated by law (Van Aerschot 2016).Policymakers 'solve' this problem by giving street-level staff far-reaching discretionary powers (Lipsky 2010;Van Berkel et al. 2017).
On the one hand, discretion, i.e. 'the extent of freedom a worker can exercise in a specific context and the factors that give rise to this freedom in that context' (Lipsky 2010, 2), enables staff to devise individualized solutions to clients.On the other hand, discretion can also be used for personal gain, and it can leave room for decisions which are biased or contrary to the intentions of the law (Lipsky 2010;Van Aerschot 2016).It is discretion that gives street-level practitioners the ability to adapt to changes in policy and management and negotiate service-users' rights.In this way, it delegates contradictory goals to street-level bureaucrats: 'If goals were clearer, workers could direct their energies with less ambivalence' (Lipsky 2010, 199).For example, Van Berkel and Van der Aa (2012) suggest that the contradictory goals of controlling and supporting clients can be separated between different social workers or different units (cf.McDonald and Marston 2006).
One area of street-level work that ADM aims to affect is time, especially time efficiency.Time is always of the essence in street-level bureaucracies who experience chronic resource constraints (Lipsky 2010, 29).Although organizations may find ADM desirable for this reason, aiming for increased time efficiency in a street-level bureaucracy may not have the intended effects and fail to improve social work.Reducing time-consuming administrative tasks, to increase relational services such as client meetings, will not necessarily improve the quality of relational social work.This is because the more a service is improved and is responsive to the needs of clients, the greater the demand on the service becomes.When additional services are made available, demand will increase to consume them.Thus, increased capacity resulting from time efficiency will result in supplying a service at a higher volume (quantity), rather than increasing its quality (Lipsky 2010, 30).
These conditions have at two consequences of interest regarding time efficiency.First, when prioritizing between serving more clients or providing higher-quality service, it is likely that public managers feel pressured to choose the former due to the inability to measure the value of a service (Lipsky 2010, 99).However, street-level bureaucrats try to reduce and ration service demands through routinization.Time-consuming procedures and routines aimed at providing legal security to clients also protect street-level workers by limiting client demands.Moreover, procedures and routines provide a legitimate excuse for street-level workers to be inflexible when necessary, further reducing client claims for service.Thus, if ADM aims to serve more clients, street-level bureaucrats may try to reduce the demand by routinization.Second, streetlevel bureaucrats have little interest in reducing client queues and delays, since faster client processing would only strain available resources (Lipsky 2010, 90).Queues are formed and clients are required to wait for services, fill out forms, are discouraged from complaining in order to make social workers work more efficient (Lipsky 2010, 90).In this sense, increased efficiency through ADM may not take the time clients spend on completing electronic applications into account.

Cases and methods
Our cases, data collection, and data analysis are described below.

Two cases of ADM policy in social assistance and activation
In Sweden there are 290 municipalities with considerable variation in population and size.Municipal activation and social assistance services are characterized by local variation and a lack of evidence as to what works (Vikman and Westerberg 2017;Forslund et al. 2019).Services are instead founded on moral ideas of financial self-sufficiency and activity in exchange for support (Ulmestig and Marston 2015).Many applicants are unemployed, immigrants, or are in poor health, but are not established in universal social security programs like unemployment or disability insurance (Ulmestig and Marston 2015).Activation is driven by governmental economic incentives and the idea that activation reduces costs for social assistance (Ulmestig 2007;Brauer 2022).Municipal activation concerns 94 000 unemployed, employ around 5400 persons in municipal jobcentres and costs 5,2 billion SEK annually (SKR 2021).Although it is not legally required to establish jobcentres, almost all municipalities have them.The handling of social assistance applications is based on individual means-testing and regulated by the Social Services Act (SFS 2001:453).This frame law gives caseworkers few details on how to interpret and deal with individual needs.Caseworkers can thus be considered street-level bureaucrats with high discretion and are also expected to offer support for non-financial matters like addictions and parenting.
Both social assistance caseworkers and job coaches at the municipal jobcentres control and support clients simultaneously (see Van Berkel and Van der Aa 2012;McDonald and Marston 2006).They control by questioning clients about the number of job applications sent or their financial situation.But they also support by helping with their job applications and skills, or low self-esteem.Still, caseworkers and job coaches represent two different logics towards client work.Activation intends to increase participation and give short-term qualifications for work related activities to maintain the work-first approach (see e.g.Ulmestig 2007).It is a contrast to a traditional social work logic -relationships -which refers to the interaction and communication between social worker and client that intends to enable individualized change and support (Nordesjö, Scaramuzzino, and Ulmestig 2022).These logics are ideal types as relationships exist in activation work, and relational social work may aim for work related activities (see Govender 2023).
The social assistance system is a common area to be automated in the Swedish public sector since its main 'product', financial resources, is numeric and its tasks are relatively administrative.We compared two cases of ADM policy in Swedish municipalities' social assistance units to understand how ADM policy is legitimized and performed through time efficiency (instrumental cases, Stake 2005).The cases are two smaller, equally sized, Swedish municipalities (population less than 50'000), fictitiously named Västerby and Sydstad, who handle social assistance in organizations together with jobcentres.Both municipalities have a local labour market policy with emphasis on activation measures.Early conversations with both municipalities, suggested that automation was intended to make social assistance more time efficient.
Before ADM, caseworkers in social assistance generally received and investigated paper applications, met the client for a discussion, gave a decision on social assistance and, if suitable, referred the client to a job coach.Time spent on administration, such as controlling documentation, was high and the client could wait several days for information on the application.Robotic process automation (RPA), the type of ADM that is most prevalent in Swedish social assistance, was intended to make this process more time efficient.RPA applies pre-programmed routines to structured data, such as an application for social assistance, to suggest an outcome such as a decision (Ranerup and Svensson 2022).While the municipalities have similar aims, the main difference is that Västerby had begun their implementation of ADM several years before Sydstad.Although the ambition has been to implement RPA in both municipalities, the empirical findings will show that only Västerby's ADM is considered RPA since it suggests decisions from clients' applications.ADM in Sydstad primarily translates applications to caseworkers' own system for caseworkers to assess, as an automated support system.The similar goal of time efficiency and difference in time spent on implementation of ADM, directed our interest to how ADM policy would play out in the two municipalities.

Data collection
Data from qualitative interviews with 17 managers, quality personnel, and professionals were performed during 2021 (see Table 1).Municipal documents such as guidelines, strategies and PowerPoint presentations were received prior to the interviews.
We approached municipalities by asking their top management about carrying out a multisite case study on the significance of digital tools for the social worker -client relationship.Managers were responsible for and had a long-term perspective on the implementation of ADM.Quality personnel had experience with hands-on assignments and knowledge on the practical ambitions and challenges of ADM implementation.In Västerby, this knowledge was located at the managerial level.Professionals (caseworkers and job coaches) were all working with ADM and meeting clients to various degrees.While caseworkers have social work bachelor's degrees, job coaches' educational backgrounds are more diverse, e.g.behavioural psychology and human resources.
Face-to-face interviews were prepared but were carried out digitally via videocalls due to the COVID-19 pandemic.Interviews lasted 40-90 minutes.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 all interviews, interview guides were semi-structured to encourage participants to reflect while ensuring their answers remained pertinent to the following themes: ADM as it relates to implementation, casework, and relationships with clients.All interviewees received information on project aims and interview procedures beforehand and gave verbal consent.The research project was subject to ethical review and approved by a Swedish Regional Ethical Review Board (Dnr 2020-00114).

Data analysis
Data analysis was rooted in theory as well as earlier research, both of which contributed to the themes in the interview guide.Data was analysed via thematic analysis (Braun and Clarke 2012).
Interviews and documents were read, and tentative codes were formulated beginning from our interest in time efficiency in ADM policy and street-level bureaucracies.This process resulted in two main themes: how time efficiency is legitimized and how time efficiency is performed.Sub themes, such as Time efficiency for activation or relationships, were formulated by comparing data between municipalities.Subthemes thus explored and clarified differences and similarities in legitimation and performance of time efficiency.In the next step, themes were discussed and reformulated in dialogue with theory on street-level bureaucracies and earlier research, before arriving at the final interpretive themes.

The legitimacy and performance of ADM policy through time efficiency
In this section, we present the findings that arose within two main themes: legitimizing ADM through time efficiency and performing time efficiency in ADM policy.

Legitimizing ADM through time efficiency
In the first main theme, time efficiency is presented as the dominant organizational goal through which ADM policy is legitimized.However, municipalities differ in their reasons for achieving it.

Time efficiency as the dominant organizational goal
The municipalities relate their implementation of ADM to powerful interest organizations and present it as significant to the whole of society.For example: E-services and ADM is the future.This work will, according to SKL (The Swedish Association of Local Authorities and Regions), be a prerequisite for social services and the rest of the public sector to meet the demographical challenges ahead.(Document, Sydstad, p. 4) Gaining legitimacy by presenting ambitions that are valued highly by several stakeholders in society (Hasenfeld 2010) is particularly evident in organizations of social assistance and activation, where there is a lack of scientific knowledge on what interventions work for whom, under what circumstances (Vikman and Westerberg 2017;Forslund et al. 2019).
In interviews and documents, municipalities relate ADM policy to several intertwined and potentially conflicting organizational goals.ADM is intended to contribute to legal security by removing feelings and relationships from the application process and thus providing more uniform and unbiased decisions: 'There are no feelings involved' (Manager, Västerby).A professional goal is to enable social workers to do 'real' social work; i.e. work relationally with clients towards self-sufficiency: 'It is about increasing our time with our customer and our work, the social work, so that we won't have to spend time on administration'.(Manager, Sydstad).Moreover, ADM contributes to improving the client's situation.Social assistance is made accessible to clients who ' . . . in a simple and modern way can apply for continued social assistance' (Document, Sydstad, p. 2).Applying electronically empowers clients by allowing them to gain 'greater trust and a more responsible way of applying by not having to produce so much documentation for the ongoing applications'.(Document, Sydstad, p. 2).It reduces stress since 'it's hard to focus on the future when you don't know if you will or won't get any money' (Manager, Västerby).Thus, less time should be spent by caseworkers on controlling documentation when they could instead be working on motivation.This reasoning is recognized in activation research where clients' rights and obligations are reconfigured (Hvinden and Johansson 2007).For example, reducing control of documents may conversely increase control of clients' behaviour through increased responsibilities and motivational work (Denvall, Nordesjö, and Ulmestig 2020).In this sense, ADM in the two municipalities is intended to be a vehicle for activation efforts.
However, the goal that was most often referred to in interviews and documents relates to how ADM makes social assistance and activation services time efficient for social work professionals and for the organization.In Sydstad, ADM should lead to 'faster casework and less administration' (Quality personnel).In Västerby, 'through a digital service, simpler, faster and cheaper casework was made possible, referring to the Swedish Administrative Procedure Act stating that "public administration must be handled easily, quickly and cheaply"' (Document, Västerby, p. 3).By being time efficient, municipalities can not only do better work, but also meet societal expectations of modernity: Most of all we will get faster services, and that's the way it is in society, we can't, in the public sector, we can't stop, we got to be responsive and go with the flow, that's what people are expecting.(Manager, Västerby) In accordance with earlier research (cf.Busch and Henriksen 2018;Ranerup and Svensson 2022), ADM is intended to contribute to several goals in the municipalities that are valued highly in society and social work administration.However, the goal of time efficiency stands out as a dominant organizational goal through which ADM is legitimized, thereby aligning itself with instrumental problem-solving discourse (Germundsson 2022).

Time efficiency for activation or relationships?
Although time efficiency is often presented as an end, it is in essence a means for another goal: to save or reallocate resources.The two municipalities differ in their immediate reasons for being time efficient and in claiming legitimacy for ADM use: time efficiency for activation or relationships.
In Västerby, time efficiency is valued to let caseworkers handle more applications and process clients to a jobcentre.Ideally, clients who apply for the first time do not meet the caseworkers, but with a job coach within a day from their submitted application.This is different from before, according to the manager, who describes how it could take several weeks before you met a job coach or got your application assessed, and how there are fewer caseworkers and more job coaches today.In this way, resources have been relocated to job coaches with a stronger focus on guidance, activation and the future.Some caseworkers who wanted to work more relationally have left their jobs, a caseworker explains, while others have stayed.Some people, including scholars, were angry, the managers says: They think that we don't meet people. . .when in fact we meet people more than other municipalities do/ . . ./they put great emphasis on the fact that it must be social workers who meet clients . . .that it in itself is the most important thing . . .and we say no, it isn't.(Manager, Västerby) Separating the dual social worker role of control and support between social assistance and a jobcentre in this way has made roles 'cleaner' for professionals and clients, according to the manager.Where caseworkers administer the economy through ADM, job coaches support and motivate the client towards self-sufficiency.It is an example of how to balance contradictory goals within street-level bureaucracies (see McDonald and Marston 2006) by separating the individual means-test from the social worker -client relationship.
In Sydstad, the purpose of time efficiency is to promote relational work.Clients who apply for the first time meet a caseworker and may only meet a job coach at a later point in time.Although their goal is also self-sufficiency, the immediate focus is to free up time demanded by administrative tasks to use for caseworkers to work relationally with clients: For us, it has never been about saving money.Maybe for our managers, but not for us.It has been about saving time to work with social work and with employment and everything that is needed.That has been our goal and that is what we have told the staff.(Manager, Sydstad) The managers and quality specialist claim that criticism to ADM has been absent due to caseworkers' participation in the implementation.For example, it was important to identify and share caseworkers' concerns early through workshops and working groups and explain that the purpose of ADM was not to replace personnel but to increase time spent with clients.These ambitions are confirmed by caseworkers who were positive to simplified client administration.According to one of them, employees are included in development: 'I think that also does a lot, motivates, it's fun. . .we get to build it as a team'.(Caseworker, Sydstad).
Gaining time for increased client meetings is important so 'that we do not set up a standardized plan for everyone but one that is adapted to the client's situation and the client's family situation' (Quality specialist, Sydstad).Therefore, Sydstad have goals of customer time, where 30% of each caseworker's time should be dedicated to meetings or communicating with clients.Customer time is thus an indicator for relational work and provides legitimacy for ADM policy.
In sum, municipalities legitimize ADM through time efficiency to achieve different goals.Both municipalities intend to increase the productivity of a certain volume (cases to jobcentre or time spent with clients).

Performing time efficiency in ADM policy
In this theme, the different goals surrounding time efficiency have implications for how street-level bureaucrats perform ADM policy in the two municipalities.

Saving time with ADM
According to all interviewees in Västerby (efficiency for activation), ADM has made casework more time efficient: 'in another municipality, you maybe have 30 households, we have 150-170 households per caseworker/ . . ./it's like night and day' (Caseworker, Västerby).Interviewees attribute time efficiency not only to the technology itself, but to the organization of ADM.The separation between control and support that relocates motivational work to job coaches has made goals and tasks for caseworkers and job coaches clearer, making them able to 'direct their energies with less ambivalence' (Lipsky 2010, 199).Caseworkers take a secluded controlling role, focusing on the ADM.Clients' questions are posed by calling general municipal customer service, and caseworkers return calls within 24 hours.This is an improvement, a caseworker argues, since customer service can deal with the constant flow of questions which would otherwise drain caseworkers' working hours.In this way, queues are organized to make casework more efficient.However, in Lipsky's (2010, 90) perspective, it is not the clients' time that is of value for the organization and there may be little interest in responding to clients' needs since it strains available resource.Moreover, most faceto-face conflicts with clients are removed: Something I think is a good thing is that when you do not meet the client, you do not have this . . .finances are a very sensitive issue.And if you must have a meeting to reject an application, there are often conflicts.Now, as a caseworker, you can make your decisions and think them through in peace and quiet.You are not affected by anything.There were so many different emotions sometimes that could take over that took energy.That's gone today.(Caseworker, Västerby) Thus, the separation of control and support through ADM in Västerby facilitates time efficiency in client processes, while at the same time reducing time, and energy-consuming individual assessments and relational social work potentially involving emotions and conflict.Casework is rationed by being routinized, where technology and citizens are dealt with in a standardized manner.The organization of secluded casework protects caseworkers by limiting client demands (cf.Lipsky 2010, 99).
Caseworkers in Sydstad (efficiency for relationships) have shifted their role towards support.ADM -including the electronic application and the reduction in client documentation -has made roles clearer and contributed to a decrease in time from application to decision, and an increase in caseworkers' time spent with clients (customer time).However, from a caseworker perspective, ADM is restricted to translating electronic client applications to a documentation system, rather than collecting data from other systems and suggesting decisions.'In itself, the robot has not made any difference . . .all the improvements that have reduced administration are basically independent of the robot' (Caseworker, Sydstad).Instead, caseworkers argue that the main contributor to time efficiency is that they stopped routinely asking for all client documentation.This made some caseworkers nervous at first but turned out to change their relationships with clients: It's easier to build a relationship if you don't have to be someone who doubts your documentation . . .this looks weird and so on . . .By letting go of the role of the controller, we have the possibility to instead be a support for our clients and help them in their planning.(Caseworker, Sydstad) In sum, while ADM in Västerby has made caseworkers focus on control and turned support over to job coaches, caseworkers in Sydstad focus on support.Just as in Västerby, caseworkers in Sydstad experience clearer goals and tasks.While this reduction in conflicting and competing caseworker goals seems to be appreciated by caseworkers, it may also reduce the need for caseworker discretion.This issue is explored through the use of 'saved time'.

Using saved time on applications and meetings
The time saved from ADM in Västerby is primarily spent on processing more cases for the activation agency.'We get time for more applications, we get time to handle more cases', a caseworker says.It also leaves time for dealing with appeals, internal meetings and posing followup questions to returning clients who have incomplete applications that cannot be automated.Thus, supporting the client and doing relational social work is not an integral part of caseworkers' work anymore.Instead, job coaches claim to have more client contacts than other equally-sized municipalities.They avoid discussing the client's financial situation: Thus, one the one hand, ADM policy in Västerby has reduced caseworkers' discretion by turning over a large part of the controlling role to ADM and the supporting role to job coaches.In this sense, ADM may be seen as an example of 'digital discretion' where computerized routines curtail streetlevel discretion (Buffat 2015).On the other hand, job coaches may experience increased discretion due to extended 'luxury' time with clients.Overall, ADM may therefore relocate and reconfigure discretion rather than curtailing it.
Managers in Sydstad argue that increased time spent with clients leads to increased cooperation with both clients and other agencies.However, caseworkers think the purpose of 'customer time' is unclear: Well, what's annoying is that it is kind of a hype and pressure to meet clients, it is really important.We must do that now.And that's why we do this, to save time to have more time with clients and customer time . . .but then when we ask what we should fill the time with, what the purpose is, then we're told that we can figure that out as we go along.The important thing is to make meetings possible, instead of coming up with a purpose.It is thought that meeting clients will produce certain advantages, and therefore we should do it.So, it's a good thing to meet clients, we don't know why but we'll figure it out as we go.It's confusing.(Caseworker, Sydstad) Furthermore, caseworkers feel constrained and forced to use their customer time to meet with all clients once every month, preventing them from meeting with different clients a different number of times.
I have a purpose with my meetings.But with some clients I keep in touch with mail and still feel updated and don't have to meet but every third month or so.And that is an assessment I want to make myself.(I: But don't you?)I don't know if you lose it this way, I feel controlled.(Caseworker, Sydstad) In line with Lipsky's argument, the increase in capacity in Sydstad increases demand and results in supplying meetings at a higher volume, rather than necessarily increasing the quality of meetings (cf.Lipsky 2010, 30).This may be due to the problems of measuring the value of a service such as a meeting, and public managers will therefore most likely prioritize a higher volume of meetings rather than a higher quality of meetings (cf.Lipsky 2010, 99).Although caseworkers can hold meetings according to their expertise and do not need to control the documentation, the number of meetings becomes a measurement of success, regardless of the content of meetings.ADM can thus be understood as shaping the form of interaction and thus caseworkers' discretion.
Table 2 summarizes our findings.Overall, ADM is legitimized through time efficiency for activation in one municipality and for relationships in the other and have significance for how municipalities save and use time.ADM's implications for discretion vary accordingly -it is either constraining and relocated to job coaches or shaping the form of discretion.

Discussion
This study has shown how ADM is legitimized through time efficiency for activation or relationships and creates conditions for how ADM policy is performed.The number of cases and client meetings become measures of success.The significance of the organizational goal of time efficiency shows how ADM is primarily performed as a tool for the organization rather than for the profession or the client.This supports research observing a strong focus on efficiency in public sector ADM and ICT, where managers are under considerable pressure to prioritize efficiency (Rose, Persson, and Heeager 2015;Busch and Henriksen 2018;Germundsson 2022;Ranerup and Svensson 2022), implying that managers primarily view ADM as a labour substitution tool (cf.Rose, Persson, and Heeager 2015).This also supports research describing an ambivalence among social workers as to whether ADM may contribute to more time for relational social work (Scaramuzzino).
Additionally, relating time efficiency to different goals (activation and relationships), contributes to understanding how the pursuit of efficiency in ADM policy obtains specific meanings in social assistance organizations working towards clients' self-sufficiency.Here, the separation between caseworkers' roles of control and support seems to facilitate time efficiency by focusing on support by relocating control to ADM.The implications of this observation may warrant further research, since an apparent reduction of control which facilitates relational work could also be argued to open the way for other forms of control, for example control of client, as well as caseworker, behaviour.Also, caseworkers working only with the control-side (Västerby) implies that supporting the client is not an integral part of caseworkers' traditional role with social work with social assistance.While caseworkers are left with controlling clients through ADM on a 'screen-level' (Bovens and Zouridis 2002), resources are relocated to job coaches' activation services.Although we characterize these services as support, they are typically oriented more towards increasing short-term qualifications for, and participation in, work related activities rather than the client's social and economic situation (Salonen and Ulmestig 2018).
Overall, the different ways of separating street-level bureaucrats' contradictory roles of control and support appears to be a consequence of how goals in ADM policy are formulated.While ADM curtails caseworkers' discretion and relocates it to a jobcentre in one municipality (Västerby), it enables more time spent with clients for caseworkers, but simultaneously constrains them in how interaction is structured through meetings (Sydstad).This finding contributes more broadly to the literature on ADM's relationship to discretion.Indeed, ADM's' relationship to discretion is not one-sided in this study, supporting the enablement thesis in that digital tools must be understood as one of several factors affecting discretion in a contextual perspective (Buffat 2015).Rather than succumbing to curtailment of digital discretion, a contextual perspective emphasizes the organizational goals and context, the automated technology and the specific tasks replaced (Bullock 2019;de Boer 2021).Here, a difference in the level of advancement in the municipalities' ADM also seem to have determined which tasks could be automated, although it did not determine the way in which municipalities decided to separate the roles of discretionary control and support.
Of course, this is a small comparative study relying on few interviewees.Interviews with clients could have highlighted if and how they perceive potential increases in efficiency and quality.In particular, it is notable how both municipalities are bothered by clients and struggle with meeting them too often.This also raises the question how street-level bureaucrats cope with demands of time efficiency in ADM and what significance such coping strategies have for service delivery and the social -worker client relationship (Nordesjö, Ulmestig, and Denvall 2020;Breit et al. 2021;Nordesjö, Scaramuzzino, and Ulmestig 2022).
It is difficult to predict what the future holds when it comes to ADM and what wider implications it will have on the knowledge, skills and values in activation and social work.There are currently great advances in AI-technologies, such as ChatGPT.New technologies are constantly being introduced, which can make more and more complex decisions, and it will affect social work practice.
We all have focus on getting this person closer to the labour market.In this way, we have luxury time with the clients if you know what I mean.We don't need to look at the money, what we work with are the motivational and educational interventions.(Jobcoach, Västerby).

Table 1 .
Number of interviewees.

Table 2 .
The legitimacy and performance of ADM.Activation (increasing short-term qualifications for, and participation in, work related activities that are useful according to established societal norms) Increase productivity to process clients for jobcentres Relationships (interaction and communication between social worker and client that intends to enable individualized change and support) Increase productivity to create time spent with clients