Algorithmic management from the perspective of managers: investigating the influence of system, organizational, and task characteristics on managers supported by intelligent systems

Data-driven algorithms enable the automation of management tasks that were previously performed by human managers. This paper examines the use of intelligent algorithmic systems to support managers in complex, human-related planning tasks in the work domains of production and logistics. The presented study aims to uncover how system, task, and organizational characteristics influence managers’ perceived system support for their planning performance. First, two research models are proposed, incorporating system-, task-, and organization-related variables based on the literature. Second, the models are evaluated using structural equation modeling (SEM) based on an empirical online study with 494 managers working in production and logistics. The results indicate that managers perceive greater system support when systems are integrated to a greater extent into the planning processes and are perceived as more intelligent and transparent. Additionally, the results show positive effects of the quality of organizational conditions on perceived planning support and job satisfaction. In the context of planning tasks, the positive effect of system use on planning support seems to diminish when tasks have a high impact on workers. In contrast, the positive effect becomes stronger the more relevant the tasks are considered to be for the company. The results provide insights into how intelligent planning support systems can be utilized effectively to enhance managerial planning performance, fostering not only high-quality outcomes but also fair and objective results for employees affected by the planning process.

Technological developments in Artificial Intelligence (AI), Big Data, and new data infrastructures are changing the way people and organizations make decisions (Brauner et al., 2022; Kinowska and Sienkiewicz, 2023). Data-driven algorithms enable the automation of management tasks that were previously performed exclusively by humans (Jarrahi et al., 2021; Tomprou and Lee, 2022). The use of algorithms for employee-related management is termed algorithmic management, which is by some considered to be one of the most disruptive forms of technological change currently taking place (Parent-Rocheleau and Parker, 2022). Baiocco et al. (2022) define algorithmic management as “the use of computer-programmed procedures for the coordination of labour input in an organisation” (p. 6). As such, it can include the partial or even full automation of managerial tasks that directly affect employees (Bucher et al., 2021; cf. Danaher, 2016). In practice, algorithmic management typically involves ongoing involvement and oversight of a human manager (cf. Jarrahi et al., 2021). Hence, our work focuses on applications, where human managers use algorithmic systems in order to support their managerial tasks (which might contradict definitions such as that of Lee et al. (2015); similar understanding in Lamers et al. (2022)).

While algorithmic management is predominantly used in platform-based work, it is rapidly expanding into other industries and work settings, including production and logistics in traditional organizations (Baiocco et al., 2022; Bucher et al., 2021; Jarrahi, 2018; Rani et al., 2024). This shift is driven by great competitive pressure and dynamic developments in supply chains and external conditions, which demand cost- and time-efficient operating and planning processes in production and logistics (Gabriel et al., 2022; Kreisköther and Herbers, 2024; Nguyen et al., 2025; Pererva et al., 2025; Rani et al., 2024). Consequently, there is a growing need to leverage digital technologies to support planning processes, as they are becoming increasingly challenging for managers (Dillerup et al., 2019; cf. Göbel et al., 2025). Workforce scheduling can be regarded as a typical application of algorithmic management in production and logistics (Baiocco et al., 2022; Brione, 2020; Jarrahi, 2018). Due to a multitude of influencing factors, including personnel-related, temporal, technical, and legal factors, workforce scheduling is highly complex and would therefore benefit from algorithmic support (Burgert et al., 2024; Kreisköther and Herbers, 2024; Latos et al., 2024). At the same time, it directly affects employees, as it involves decisions about working time, place, and content (Bentler et al., 2022; Burgert et al., 2024; Kreisköther and Herbers, 2024).

Empirical studies on algorithmic management have primarily focused on the workers affected, especially in gig work contexts (Cameron et al., 2023; Parent-Rocheleau and Parker, 2022). Existing studies mainly point to adverse effects of algorithmic management on workers, including increased performance pressure, reduced actual and perceived autonomy, and lower job satisfaction (Baiocco et al., 2022; Benlian et al., 2022; Gagné et al., 2022; Jarrahi et al., 2021; Kellogg et al., 2020; Kinowska and Sienkiewicz, 2023; Lamers et al., 2022; Parent-Rocheleau and Parker, 2022; Parker and Grote, 2022; Rani et al., 2024). Conversely, the implications for managers – who are likely to experience a significant transformation in their roles due to algorithmic management – remain understudied (cf. Becker et al., 2023; Cameron et al., 2023; Langer et al., 2021; Noponen, 2019). Proponents argue that it can enhance decision quality, effectiveness, efficiency, and cost savings, thereby helping organizations navigate complex and dynamic work environments (Brione, 2020; Kellogg et al., 2020; Rani et al., 2024; Tomprou and Lee, 2022). In particular, “intelligent” AI-supported systems are posited to improve planning performance by processing vast datasets, analyzing complex relationships, and predicting trends. This facilitates more efficient, dynamic, and accurate decision-making (Brione, 2020; Cheng and Hackett, 2021; Gil et al., 2020; Kellogg et al., 2020; Ribbat et al., 2022; Rossmann and Wald, 2024). While algorithms are often regarded as less susceptible to bias than human decision-makers (Brione, 2020; De Cremer, 2020; Kinowska and Sienkiewicz, 2023; Lee, 2018; Newman et al., 2020), they also carry the risk of perpetuating or amplifying biases when data quality is poor (Kalimeris et al., 2022).

This paper addresses the neglected managerial perspective and investigates how intelligent planning systems can support managers who conduct complex, employee-related planning tasks, such as workforce scheduling or route planning, in the domains of production and logistics by enhancing planning performance and enabling more objective and fair decisions, which may in turn foster positive outcomes for employees. Since such tasks are supported by a variety of different systems and custom solutions in practice (Stich et al., 2013), we do not evaluate a specific algorithm or system, but instead examine managers’ perceptions of the planning support systems that they use in their daily work. The intention is to derive more generalizable insights into the underlying factors that determine the perceived managerial support provided by such systems.

To understand how intelligent planning support systems can support managers in complex employee-related planning tasks in the domains of production and logistics, an examination of the factors that influence planning performance is necessary. Relevant insights can be drawn in particular from information systems (IS) research. Literature in the context of IS (Rzepka and Berger, 2018) and related (sub-)disciplines such as human-machine systems (Stowers et al., 2017), strategic decision-making (Rajagopalan et al., 1993; Shepherd and Rudd, 2014), human-AI interaction (Habbal et al., 2024; Jiang et al., 2024; Papacharalampopoulos et al., 2025; Yan et al., 2025), and management support systems (Kempkes et al., 2024) indicate that the perceived performance can be influenced by system, user, and contextual characteristics. Firstly, the extent of system use is a central antecedent of system impact that needs to be considered since algorithmic management systems cannot contribute to managerial performance unless they are incorporated into the planning process (cf. DeLone and McLean, 2003; Power, 2002). Secondly, the impact of perceived system intelligence is important, as the analytical and predictive capabilities of algorithmic systems are viewed as a core mechanism, enabling data-driven decisions in complex environments (e.g., Baiocco et al., 2022; Benlian et al., 2022; Jarrahi et al., 2021; Kinowska and Sienkiewicz, 2023). Thirdly, transparency is a critical system characteristic, as it enables users to comprehend algorithmic decision-making processes and outcomes (e.g., Mateescu and Nguyen, 2019; Papacharalampopoulos et al., 2025; Rzepka and Berger, 2018; Yan et al., 2025). This seems especially important when planning directly affects workers and may require accountability. Fourthly, the extent to which organizational conditions facilitate effective managerial system use for employee-related planning is important to consider (Kempkes et al., 2024; Sanders and Courtney, 1985; Shepherd and Rudd, 2014). Finally, job satisfaction needs to be considered as a user characteristic as it is presumably impacted by the (in-)effectiveness of the supporting system (Joshi and Rai, 2000; Liere-Netheler et al., 2017).

Our study thus addresses the following research question: RQ1: To what extent is the perceived planning support for complex and worker-related planning processes in production and logistics influenced by the perceived system transparency, the perceived system intelligence, the extent of system use, and the perceived quality of organizational conditions and to what extent is the managers’ job satisfaction affected by the planning support and the quality of organizational conditions?

Furthermore, the effectiveness of algorithmic systems in managerial planning does not depend solely on system-related and organizational factors, but also on task characteristics (Kempkes et al., 2024; Rzepka and Berger, 2018; Shepherd and Rudd, 2014; Stowers et al., 2017). More specifically, research on systems supporting managerial tasks indicates that task characteristics may moderate the effect of system use on performance (Kempkes et al., 2024). Given our focus on complex planning tasks that affect both employees and organizational outcomes, the following research question arises: RQ2: To what extent is the influence of the extent of system use on the perceived planning support for complex and worker-related planning processes in production and logistics moderated by the perceived task characteristics complexity, relevance, and impact on workers?

From a methodological perspective, there is a need for field studies and surveys to examine performance impacts with greater generalizability (Kempkes et al., 2024). Hence, we conducted an online survey with n = 494 managers from production and logistics and investigated the research questions by evaluating two corresponding research models using partial least squares structural equation modeling (PLS-SEM). Our study contributes to algorithmic management research in three ways: First, it provides an empirical investigation from the neglected managerial perspective. Second, it focuses specifically on planning tasks that affect employees, as this is a key characteristic in the context of algorithmic management. Third, the study draws on empirical data from managers who use systems for planning tasks in their daily work, thereby grounding the findings in real-world practice.

The remainder of this paper is organized as follows: In Section 2, the research models to address RQ1 and RQ2 are proposed. In Section 3, the research methodology is described, including the survey procedure, the sample, and the measures used. The results of the structural model evaluation are presented in Section 4. In Section 5, the results are discussed in the context of current research and conceptual implications as well as implications for system design and company practices are derived.

Theoretical background and hypotheses development

Effects of system and organizational characteristics on planning support

The IS literature has long acknowledged the importance of examining how IS use affects performance (Heine et al., 2003; Kuegler et al., 2015). One of the most established frameworks is the DeLone and McLean Model of Information Systems Success (D&M IS Success Model) (DeLone and McLean, 1992; DeLone and McLean, 2003), which links system, system output, and organizational support quality to user satisfaction, system use, and ultimately performance benefits (DeLone and McLean, 2003; cf. Petter et al., 2008). While these relationships have been widely examined and adapted across contexts (see Petter et al., 2008; Urbach and Müller, 2012), evidence suggests that performance impact is shaped not only through usage-related mechanisms but also through more immediate effects of system and organizational characteristics (Petter et al., 2008). Therefore, we examine the effects of system use, intelligence, and transparency as well as the quality of organizational conditions on planning support – defined as the system’s perceived impact to achieve not only fast, accurate, and high-quality decisions, but also decisions that are objective and fair.

The extent of system use captures how strongly a planning support system is involved across the phases of the planning process. From an information-processing perspective, algorithmic systems can augment managerial cognitive capacity and improve decision-making performance, e.g., by enabling the processing of larger and more complex datasets (cf. Brione, 2020; Chan et al., 2015; Jarrahi, 2018; Yan et al., 2025). However, these benefits depend on the system’s effective integration into planning processes (cf. DeLone and McLean, 1992; 2003; Goodhue and Thompson, 1995; Iivari, 2005; Power, 2002; Rai et al., 2002). We thus assume that a greater extent of system use increases the potential to realize stronger system support for complex planning tasks, as deeper integration facilitates contributions to information processing, alternative generation and evaluation, and decision implementation (cf. Parasuraman et al., 2000). In fact, empirical studies in different contexts indicate that system utilization can positively influence performance (e.g., Ikenyei and Haggerty, 2024; Im, 2014; Luarn and Huang, 2009; Petter et al., 2008; Ratna et al., 2020; Sun et al., 2009). Based on this, we propose the following hypothesis (H): H1: The extent of system use affects perceived planning support positively.

Perceived system intelligence refers to managers’ assessment of the extent to which the system performs intelligent, data-driven functions. These functions are relevant because they can extend human information-processing capacity, particularly in complex planning situations. AI can support the collection, interpretation, and evaluation of information, recognize relevant patterns and structures for decision-making, and thus support decision-making performance (El Khatib and Al Falasi, 2021; Gil et al., 2020; Kalimeris et al., 2022; Ribbat et al., 2022; Trunk et al., 2020). In employee-related planning, AI may also lead to a more substantiated, objective, and fair basis for decision-making, as more information can be considered (Gil et al., 2020; Habbal et al., 2024; Ribbat et al., 2022). Hence, we propose the following hypothesis: H2: Perceived system intelligence affects perceived planning support positively.

Perceived system transparency, defined as the extent to which managers can comprehend system functioning, including data, its purpose, and internal algorithms (Balasubramaniam et al., 2023; Peissl et al., 2022; Robert et al., 2020), is a key requirement for responsible and effective AI use (e.g., Habbal et al., 2024; Yan et al., 2025). Prior research shows that system transparency can support informed decision-making, improve decision quality, and increase users’ trust in system advice (Kolbjørnsrud et al., 2017; Papacharalampopoulos et al., 2025; Rzepka and Berger, 2018; Wanner et al., 2022). This seems particularly relevant in complex planning tasks, where managers remain accountable for decisions in light of their consequences for workers and organizations. We therefore propose: H3: Perceived system transparency affects perceived planning support positively.

Moreover, intelligent planning systems may provide more meaningful support when managers operate under organizational conditions that facilitate their effective use (cf. Kempkes et al., 2024; Kozioł-Nadolna and Beyer, 2021; Sanders and Courtney, 1985; Shepherd and Rudd, 2014). The perceived quality of organizational conditions captures whether managers experience the organizational environment as supportive for their planning tasks. The impact of organizational conditions, such as resources, training, infrastructure, and management support, on performance-related measures has been indicated in studies across various contexts (e.g., Costa et al., 2016; Kozioł-Nadolna and Beyer, 2021; Rajan and Baral, 2015), including management support systems (Kempkes et al., 2024) and AI-based decision-making (Cao et al., 2021; Dietzmann and Duan, 2022; Marocco et al., 2024; Sànchez-Marrè, 2022). Overall, the findings suggest that favorable organizational conditions enable more effective use and enhance perceived benefits. Additionally, perceived organizational support theory (Eisenberger et al., 1986) suggests that a supportive organizational environment can enhance job satisfaction (see Eisenberger et al., 2020; Rhoades and Eisenberger, 2002). In fact, studies in management contexts indicate that, for example, training, software support, and management support, not only increase satisfaction with systems but also promote job satisfaction (Bhattacharya et al., 2023; Sykes, 2015; Terek et al., 2018). Thus, we propose the following hypotheses: H4a: Perceived quality of organizational conditions affects perceived planning support positively.

H4b: Perceived quality of organizational conditions affects job satisfaction positively.

Job satisfaction may also be shaped by the perception of planning support. Work design research highlights job satisfaction as an important work-related outcome that can be influenced by technology use (Morgeson et al., 2012; Parker et al., 2017; Parker and Grote, 2022). More specifically, IS-related studies suggest that systems perceived as useful for daily work may improve job satisfaction by reducing unnecessary effort and supporting task accomplishment (Joshi and Rai, 2000; Liere-Netheler et al., 2017). Building on this, we propose: H5: Perceived planning support affects job satisfaction positively.

Based on the hypotheses formulated above, we propose a research model guiding our empirical analysis (Figure 1). It describes the effect of the extent of system use (SYS), system intelligence (INT), system transparency (TRA), and the quality of organizational conditions (ORG) on planning support (SUP) as well as the effects of ORG and SUP on the managers’ job satisfaction (SAT).

Moderating role of task characteristics

The interaction between humans and systems and thus its outcomes are not only dependent on system characteristics and organizational context, but also on the characteristics of the task performed with system support (Kempkes et al., 2024; Rzepka and Berger, 2018; Shepherd and Rudd, 2014; Stowers et al., 2017). This is particularly relevant in algorithmic management, where planning tasks are not merely technical or administrative but directly concern workers while serving organizational objectives. With regard to management tasks, Kempkes et al. (2024) argue that task characteristics may moderate the effect of support systems for management tasks on job performance. This aligns with Task-Technology Fit theory (Goodhue and Thompson, 1995), which posits that systems lead to an improvement in task performance if they meet the requirements of the task. Building on this, we investigate the extent to which the influence of system use on planning support is moderated by three task characteristics that are especially relevant in complex, employee-related planning: task complexity, task relevance, and task impact on workers (RQ2).

Perceived task complexity refers to the extent to which a planning task requires processing extensive and dynamic information, comparing diverse alternatives, and balancing multiple, interdependent, and potentially conflicting criteria (Beach and Mitchell, 1978; Dörner, 1983; Dörner, 2011; Forchhammer et al., 2010; Jungermann et al., 2010; Werner, 1992), which places significant demands on human cognitive abilities (Dörner, 2011; Jungermann et al., 2010). Decision support systems (DSS) are traditionally assumed to expand users’ processing ability, enabling humans to analyze decision problems in more depth and thus enhance decision-making performance (Millot et al., 2011; Todd and Benbasat, 1992). AI-based systems may be particularly useful in this respect (Rzepka and Berger, 2018; Sànchez-Marrè, 2022). Empirical evidence generally indicates that system use improves performance primarily under high task complexity (e.g., Chan et al., 2015; Chu and Spires, 2000; Williams et al., 2007). However, Kempkes et al. (2024) highlight that findings are not fully consistent and call for more field-based research. Similarly, Rzepka and Berger (2018) suggest a moderating effect of task complexity in the context of AI-enabled systems, even though task characteristics remain understudied in AI research. We argue that when task complexity is low, systems may add limited value, whereas complexity increases the benefit of algorithmic assistance: H6: The relationship between the extent of system use and perceived planning support is moderated positively by perceived task complexity.

Perceived task relevance captures the extent to which a planning task is perceived to affect organizational outcomes (Elbanna and Naguib, 2009; Papadakis et al., 1998). Research on strategic decision-making shows that managers allocate their limited resources differently depending on the perceived importance of a decision (Elbanna et al., 2020; Elbanna and Child, 2007). Important decisions are more likely to involve rational and analytical techniques because errors, delays, or poor decisions may have stronger consequences for the organization (AlDhaen, 2017; Elbanna et al., 2020; Papadakis et al., 1998). Elbanna and Child (2007) further propose that decision importance moderates the relationship between rationality and decision-making effectiveness. Thus, we argue that in intelligent planning support, greater system use can strengthen the analytical basis of planning. This should be particularly valuable for tasks with high organizational impact, for which managers are likely to emphasize well-founded, traceable, and effective decisions. Accordingly, we expect: H7: The relationship between the extent of system use and perceived planning support is moderated positively by perceived task relevance.

Since the impact on employees is a key characteristic of algorithmic management (cf. Baiocco et al., 2022), we examine expected task impact on workers as the extent to which managers expect a planning task to affect workers in terms of, for example, work location, work methods, and remuneration. While research on algorithmic management has predominantly focused on implications from the workers’ perspective, often emphasizing adverse consequences (e.g., Gagné et al., 2022; Kellogg et al., 2020; Kinowska and Sienkiewicz, 2023; Parent-Rocheleau and Parker, 2022), the extent to which the impact of management tasks on workers influences managers’ perception of system support remains unclear.

There are reasons to expect that systems may become more valuable when worker impact is high. Algorithmic management research suggests that systems may promote objectivity and fairness, as algorithms can be less biased than humans (Brione, 2020; De Cremer, 2020; Lee, 2018; Newman et al., 2020). For example, algorithmic management may reduce the ability of line managers to exercise personal favoritism (Brione, 2020). Lee et al. (2021) find that managers value scheduling systems when these incorporate worker preferences and thereby support fairer outcomes. Moreover, algorithmic systems might support managers “to make sure that workers are always assigned to the most useful thing they could be doing […] – particularly vital in manufacturing” (Brione, 2020, p. 17). Langer et al. (2021), for example, show that using decision support systems in personnel selection increases decision satisfaction and self-efficacy. However, investigating the influence of task characteristics in more detail was suggested for further research (Langer et al., 2021). From this perspective, greater system use could help managers legitimize planning decisions and reduce concerns about bias or arbitrariness, particularly when decisions strongly affect workers.

At the same time, high worker impact also increases the need for contextual interpretation, interpersonal sensitivity, and consideration of human concerns. Decisions with human implications should respect the welfare and interests of humans affected, which might require contextual interpretation and consideration of diverse perspectives (De Cremer, 2020). Algorithms can be limited in their ability to accommodate exceptions, incorporate factors such as social interactions and personality, or consider human concerns (Lee, 2018). In fact, decisions with human implications can be perceived as less fair when made by algorithms than by human decision-makers (Newman et al., 2020), and research on AI aversion indicates that people may doubt AI’s ability to understand human needs or provide adequate empathy in tasks requiring interpersonal sensitivity (Jiang et al., 2024). From a managerial perspective, this means that extensive system use may be perceived as less supportive when planning has strong consequences for workers because managers may feel that human judgment, accountability, and interpersonal knowledge are indispensable. Consequently, we assume a moderating effect of the perceived task impact on workers on the relationship between the extent of system use and perceived planning support – though the direction of the moderation remains open: H8: The relationship between the extent of system use and perceived planning support is moderated by the expected task impact on workers.

Figure 2 shows a second proposed research model to investigate the impact of the task characteristics perceived complexity (COM), perceived relevance (REL), and expected impact on workers (IMP) potentially moderating the effect of the extent of system use (SYS) on perceived planning support (SUP).

Procedure and sample

To test the proposed research models empirically, a cross-sectional study was carried out with managers working in production and logistics in Germany. An online questionnaire was used for data collection. Participants were contacted through a professional survey panel provider. Participation was voluntary. Anonymity and confidentiality were guaranteed. Respondents received financial compensation for their participation. Screening questions were asked at the beginning of the questionnaire to ensure respondent eligibility. Only respondents whose answers confirmed the following criteria could proceed: (1) the respondents are currently employed in a company within the manufacturing or logistics sector; (2) they hold a managerial position; (3) they regularly perform a planning task that affects the work of employees in production or logistics; and (4) they are supported by a digital system in at least one of these planning tasks. In the following, participants were asked to relate all answers to a specific planning task that fulfills the criteria. During the survey period of 8 days, 624 persons completed the questionnaire. The completion rate after passing screening questions was 90.96%. The average survey completion time was M = 13.5 min (SD = 11.6 min), excluding interrupted survey completions.

To ensure sufficient data quality, a multi-stage screening procedure of the responses was carried out. First, only completed questionnaires were taken into account. Second, 11 data sets of respondents with implausibly fast completion times were excluded based on the relative speed index. The relative speed index expresses individual completion time per questionnaire page relative to the sample’s median page completion time (Leiner, 2019). Third, 70 data sets of respondents with straight-lining response patterns on more than two consecutive questionnaire pages were screened out (Schonlau and Toepoel, 2015). Finally, further 49 data sets were removed if open text answers describing the planning tasks, the systems used, and the respondents’ job titles indicated insincere participation or that the respondent did not meet the target group criteria.

The final sample consists of 494 respondents who stated to work in a managerial position. A total of 311 managers (63.0%) classified their company as belonging to the manufacturing sector, whereas 183 managers (37.0%) assigned their company to the logistics sector. The respondents reported to have been employed in their current profession at their company for M = 10.0 years (SD = 8.2 years). They further stated to typically spend M = 22.9 h per week (SD = 14.1 h per week) on the planning task and have been working with the specific system for M = 6.2 years (SD = 5.5 years). Further characteristics of the sample are shown in Table 1.

Measures

For the measurement of the latent model variables, we derived items based on relevant literature and previously published scales. To operationalize SYS, the participants were asked to what extent the system is used in individual phases of the planning process. The phases were adapted from Trunk et al. (2020) with modifications. This conceptualization builds on the argument by DeLone and McLean (2003) that commonly used measures, such as usage duration, usage frequency, or number of accesses, do not properly reflect the relationship between system use and its effects. Instead, DeLone and McLean (2003) claim that researchers need to consider the quality of system use, as well as its extent and nature. To measure SUP, we developed new items based on potential advantages of system use to support planning from the managers’ perspective, such as perceived planning system support for better, more objective, and fairer decisions, as discussed in the algorithmic management and IS literature (Kellogg et al., 2020; Kinowska and Sienkiewicz, 2023; Möhlmann and Zalmanson, 2017; Sanders and Courtney, 1985; Shamim et al., 2020). Planning support thereby refers to the system’s impact on a user’s planning performance as perceived by the user (based on Iivari, 2005 who follows DeLone and McLean, 1992; Rai et al., 2002). To measure INT, new items were developed which cover AI functions such as monitoring large amounts of data in real time (Castro and New, 2016) as well as methods/concepts such as the use of machine learning (European Commission, 2021). TRA is measured by new items referring to the perception of transparency about system purpose and data selection (Robert et al., 2020), interaction for better results (Hellmann et al., 2022), as well as understandability and explainability (Peissl et al., 2022). The items to measure ORG were developed based on organizational factors shaping managerial decision-making, such as the provision of adequate means and resources, time capacities, or training opportunities (Donelan et al., 2015; Kozioł-Nadolna and Beyer, 2021), supplemented by one item measuring perceived organizational support for the task in case of problems, based on the Perceived Organizational Support Scale by Siebenaler and Fischer (2020). In contrast to prior studies that have mainly examined system satisfaction, decision satisfaction, or usefulness (e.g., Costa et al., 2016; Langer et al., 2021; Lilien et al., 2004; Rai et al., 2002; Ratna et al., 2020), we specifically address job satisfaction as an important work-related outcome (Morgeson et al., 2012; Parker and Grote, 2022). Therefore, the scale for SAT is based on the general job satisfaction scale of the Job Diagnostic Survey (Schmidt et al., 1985). To account for the fact that the planning task might be only one part of the managers’ daily responsibilities, we related all items specifically to the planning task. The items to measure COM were derived from decision science literature, referring to the amount and dynamics of information to be considered, the amount and diversity of alternatives, the amount and interconnectedness of assessment criteria, and the degree of polytely (the need to pursue multiple, potentially conflicting goals simultaneously) (Beach and Mitchell, 1978; Dörner, 1983; Dörner, 2011; Forchhammer et al., 2010; Jungermann et al., 2010; Werner, 1992). The items to measure REL were based on the operationalization of decision importance by Elbanna and Naguib (2009) and magnitude of impact by Papadakis et al. (1998) and cover the perceived impact of the planning task on, for example, the company’s performance and future decisions in the company. To measure IMP, participants were asked to what extent the planning task influences the employees’ working conditions in terms of aspects such as workplace, time, and task sequencing, based on dimensions of worker autonomy (Breaugh, 1999; Spiegelaere et al., 2016; Stegmann et al., 2019), supplemented by the aspect of remuneration as discussed by, for example, Parent-Rocheleau and Parker (2022). For all items, a 6-point Likert scale was used, ranging from “completely disagree” (1) to “completely agree” (6) (all scale points labeled according to Franke et al. (2019)). SUP, SYS, INT, TRA, ORG, COM, REL, and IMP are specified as formatively measured constructs. SAT is specified as a reflective construct. In contrast to reflective constructs, the measurement indicators for formatively measured constructs are considered as influencing factors on the construct (Hair et al., 2022; Huber et al., 2007).

To pretest the questionnaire, n = 11 academic colleagues provided feedback on content, structure, and format of the online survey. After minor revisions of structure and wording, an additional test with n = 4 managers from a German white goods company was conducted to ensure comprehensibility for the target group. The final items used are shown in the Supplementary Material.

Analyses

We used the partial least squares structural equation modeling (PLS-SEM) approach to evaluate our hypothesized models. In contrast to covariance-based methods, variance-based PLS-SEM is particularly suitable for rather exploratory research due to better predictive capabilities (Hair et al., 2022). PLS-SEM is applicable for complex model structures and models containing reflective as well as formatively measured constructs (Hair et al., 2022). Because the study focuses on prediction-oriented model evaluation and includes several formatively specified constructs, PLS-SEM was selected. Although PLS-SEM does not require normally distributed data, normality was checked a priori. An evaluation of skewness and kurtosis for all items reveals no substantial deviations from normality, with all values remaining within the threshold range of −2 to +2 proposed by Hair et al. (2022). The following analyses were performed using SmartPLS4 (Smart PLS GmbH) in a two-step approach: First, the measurement models were analyzed to evaluate their validity and reliability. Second, the structural models were evaluated.

Measurement model evaluation

First, the reflectively specified measurement model of the first research model (SAT) is evaluated. The measurement model was tested in terms of indicator reliability, internal consistency reliability, convergent validity, and discriminant validity (Hair et al., 2022). After removal of two items due to insufficient indicator reliability (Hair et al., 2022), all indicators meet the recommended requirement of significant outer loadings of λ ≥ 0.708 (see Supplementary Material) (Hair et al., 2022). Composite reliability ρ C = 0.887 and Cronbach’s α = 0.809 are within the recommended interval of 0.7–0.9, indicating internal consistency reliability (Hair et al., 2022). The average variance extracted (AVE = 0.723) is greater than 0.5, thus indicating sufficient convergent validity (Fornell and Larcker, 1981; Hair et al., 2022). Regarding discriminant validity, the square root of the construct’s AVE is greater than the correlations with other constructs (Table 2) (Fornell and Larcker, 1981; Hair et al., 2022). Overall, the evaluation indicates that the items measuring SAT are both consistent and distinct.

Second, the formative measurement models (SUP, SYS, INT, TRA, ORG) are evaluated based on convergent validity, collinearity, the significance and relevance of their indicators, and discriminant validity (Hair et al., 2022; Huber et al., 2007). Convergent validity was examined using comparative single-item measures for each formative construct (see Supplementary Material), with a recommended minimum threshold value of 0.7 (Hair et al., 2022). The path coefficients meet this criterion for SUP, INT, and ORG, while SYS and TRA are slightly below it. In view of the fact that the formative measurement of SYS and TRA on the basis of 6 and 5 items, respectively, is opposed to a single-item measurement, we consider the values to be acceptable and retain the constructs. The variance inflation factors (VIF) of all formative indicators are below the critical threshold value of 5, indicating no collinearity issues (Hair et al., 2022). Indicator significance and relevance were assessed at a significance level of p < 0.05, using bootstrapping (percentile bootstrap method) with 10,000 subsamples. As shown in the Supplementary Material, 17 indicators demonstrate relative relevance (significant indicator weights), while 16 indicators show nonsignificant weights but satisfy the criterion for absolute relevance, i.e., significant indicator loadings or indicator loadings greater than or equal to 0.5 (Hair et al., 2022). Accordingly, all formative indicators can be retained (Hair et al., 2022). Discriminant validity can be assumed, as the correlations between each formatively specified construct and all other constructs (see Table 2) are below 0.90 (Huber et al., 2007).

Finally, the measurement models of the second proposed research model (Figure 2) are evaluated. In addition to the measurement models of COM, REL, and IMP, the models of SUP and SYS are (re-)evaluated in terms of the significance and relevance of the indicators and discriminant validity as these measurement properties might change due to their inclusion in a different structural model (Hair et al., 2022). Convergent validity was shown for REL, whereas the path coefficients for COM and IMP with a reflectively specified single-item indicator are slightly below the recommended threshold value of 0.7 (see Supplementary Material) (Hair et al., 2022). Analogous to SYS and TRA, we retain the formative constructs against the background that the formative measurement on the basis of 7 items each is opposed to a single-item measurement. VIF values of all indicators for REL, COM, and IMP (see Supplementary Material) indicate no collinearity issues (Hair et al., 2022).

Regarding indicator significance and relevance, all indicators for SYS, SUP, REL, COM, and IMP are retained as all indicators demonstrate absolute relevance and 14 out of 30 indicators additionally meet the criterion of relative relevance (see Supplementary Material) (Hair et al., 2022). As shown in Table 3, discriminant validity is also confirmed (Huber et al., 2007).

Structural model evaluation – Research Model 1

The path model is evaluated based on construct collinearity, in-sample and out-of-sample prediction capabilities, and the relevance and significance of the path coefficients (Hair et al., 2022). As shown in the Supplementary Material, the VIF values of the exogenous constructs and SUP respectively SAT are below 5, indicating no collinearity (Hair et al., 2022). Out-of-sample prediction capabilities were assessed using the PLS predict procedure (Shmueli et al., 2016; Shmueli et al., 2019) for SUP as the key endogenous construct, following Guenther et al. (2023) and Hair et al. (2022). As shown in the Supplementary Material, all Q 2 predict values are positive, indicating that the path model’s prediction error is smaller than the prediction error given by the most naïve benchmark (Shmueli et al., 2019). Furthermore, 5 out of 6 indicators for SUP have lower root mean square error (RMSE) values compared to the linear regression model (LM) benchmark. This indicates a medium level of predictive power (Shmueli et al., 2019).

Regarding the model’s in-sample prediction capabilities, the coefficient of determination (R 2) for SUP (R 2 = 0.619, adjusted R 2 = 0.616) indicates a high moderate level of explanatory power, while the R 2 value for SAT (R 2 = 0.296, adjusted R 2 = 0.293) lies between moderate and weak levels (Chin, 1998; Hair et al., 2022). Similar to the assessment of formative indicator weights and loadings, the significance assessment of the path coefficients builds on bootstrapping at an accepted Type I error significance level of p < 0.05 (two-tailed), using the percentile bootstrap method with 10,000 subsamples (Hair et al., 2022). Path coefficients and their significance are shown in Figure 3.

The results provide support for the system-related hypotheses. The extent of system use has a significant positive effect on the perceived planning support (t = 9.931, f 2 = 0.450). Also, both system intelligence (t = 2.798, f 2 = 0.021) and system transparency (t = 3.604, f 2 = 0.042) show significant positive effects on the perceived planning support. The effect sizes f 2 indicate small to large effects (Cohen, 1988). Therefore, H1, H2, and H3 are supported. These findings indicate that managers perceive planning systems as more supportive when they are integrated into planning processes to a larger extent and perceived as more intelligent and transparent.

The role of organizational conditions is more differentiated. Although the effect of the perceived quality of organizational conditions on the perceived planning support is positive and statistically significant (t = 1.992, f 2 = 0.014), the negligible effect size indicates that its predictive relevance is not meaningful. Thus, H4a is considered as only partially supported. In contrast, the perceived quality of organizational conditions shows a significant positive effect on job satisfaction (t = 12.697) with a medium effect size (f 2 = 0.317; Chin, 1998), supporting H4b. There is no statistically significant effect of the perceived planning support on job satisfaction (t = 0.055, f 2 = 0.000). Thus, H5 is not supported. Moreover, the results indicate that the perceived planning support does not mediate the relationship between the organizational conditions and job satisfaction, as the indirect effect (ORG → SUP → SAT) is nonsignificant (t = 0.046).

To control for potential confounding effects on the perceived planning support, the model was ex-post controlled for the reported sample characteristics, as these include typical user characteristics relevant to human–computer interaction (e.g., age, gender, tenure; Rzepka and Berger, 2018; Shepherd and Rudd, 2014; Stowers et al., 2017) as well as contextual task and system characteristics. All control variables were added as formatively specified constructs, with categorical variables recoded into binary dummy variables (Hair et al., 2022). As shown in the Supplementary Material, only the task assignment to a production versus logistics context significantly affects SUP (t = 2.341; path coefficient = −0.177; production as reference category). However, explanatory power increases only marginally (ΔR 2 = 0.012 for SUP). Also, the inclusion of control variables did not alter the significance of the path coefficients in the research model, except for the effect of ORG on SUP, which becomes nonsignificant. This effect, however, was already considered practically irrelevant due to its negligible effect size (Cohen, 1988).

Structural model evaluation – Research Model 2

Research Model 2 was evaluated based on the two-step approach (Chin et al., 2003), which is generally recommended by Hair et al. (2022). Initially, the VIF values are below the critical threshold of 5 (see Supplementary Material), indicating no collinearity issues (Hair et al., 2022). Regarding the model’s in-sample prediction capabilities, the R 2 value for SUP (R 2 = 0.597, adjusted R 2 = 0.591) indicates a high moderate level of explanatory power (Chin, 1998). In terms of out-of-sample prediction capabilities, the results of the Q 2 predict procedure (see Supplementary Material) suggest high predictive power of the model (Shmueli et al., 2016; Shmueli et al., 2019).

Bootstrapping with 10,000 subsamples is again used for significance assessment at an accepted Type I error significance level of p < 0.05 (two-tailed) (Hair et al., 2022). Regarding the simple effect, the extent of system use shows a significant positive effect on the perceived planning support (t = 17.056, f 2 = 0.793), confirming the effect already found in the first research model. Regarding the interaction terms, the positive but nonsignificant effect of SYSxCOM on SUP (t = 0.398, f 2 = 0.001) indicates that the perceived task complexity does not substantially alter the relationship between the extent of system use and the perceived planning support. Thus, H6 is not supported. In contrast, the positive significant effect of SYSxREL on SUP (t = 2.533, f 2 = 0.027) shows that task relevance positively moderates the relationship, meaning that for greater perceived relevance, the relationship between SYS and SUP increases. H7 is supported. Lastly, the interaction term SYSxIMP has a significant effect on SUP (t = 2.681, f 2 = 0.028). Thus, the expected impact on workers acts as a moderator. H8 is supported. Based on the threshold values for moderating effects of 0.005, 0.01, and 0.025 for small, medium, and large effects proposed by Kenny (2018), the effect sizes indicate large effects of both REL and IMP. The ex-post inclusion of control variables does not affect the significance of the path coefficients, and none of the control variables show a significant effect on SUP (see Supplementary Material). Overall, the findings indicate that the positive relationship between system use and perceived planning support is stronger for tasks with high organizational relevance, but weaker when planning has stronger implications for workers (see Figure 4).

The present study aimed to investigate how intelligent planning support systems can be leveraged to enhance managerial planning performance. In contrast to previous studies, we explicitly focus on planning tasks in production and logistics that affect workers and conceptualize planning support as encompassing not only traditional performance-related aspects but also fair and objective outcomes for employees affected by the planning process. Accordingly, our study contributes to algorithmic management research by focusing on complex planning tasks that affect workers, a central concern in this research stream, while examining such tasks from a managerial performance perspective traditionally emphasized in IS research (see Hillebrand et al., 2025). From a methodological perspective, we add to the available research evidence by drawing on data from managers who actively use systems to support employee-related planning tasks in their daily work.

Conceptual implications

The first implication can be drawn from the measurement model of the perceived planning support. The indicator loadings show that not only traditional indicators, such as decision-making speed (see, e.g., Rajagopalan et al., 1993; Shepherd and Rudd, 2014; Trunk et al., 2020), shape the perception of planning support, but also the system’s contribution to fairness and objectivity, which are discussed as potential benefits in algorithmic management research (Brione, 2020; De Cremer, 2020; Lee et al., 2021; Newman et al., 2020). Thus, managers appear to evaluate algorithmic planning support also in terms of whether systems help them make decisions that are perceived as legitimate, impartial, and justifiable. This offers a new perspective on support quality in algorithmic management contexts, not only by highlighting the relevance of objectivity and fairness from the managerial perspective, but also by providing implications for how system support can be conceptualized in future studies.

Another key result of our study is that managers experience a system as being more supportive for their planning tasks if they use the system functions more extensively in the various phases of the planning process. This result is consistent with empirical results from IS research, which emphasize system use as a central antecedent of performance-related outcomes (see DeLone and McLean, 2003; Jeyaraj, 2022; Petter et al., 2008; Urbach and Müller, 2012). However, in contrast to other studies, we examine the extent of system use to capture the system’s involvement in the managerial tasks, rather than using metrics such as usage duration or frequency (cf. DeLone and McLean, 2003; Iivari, 2005; Rajan and Baral, 2015). Our findings therefore shift attention from system use as a general behavioral construct to system use as task-embedded involvement in managerial planning processes. Accordingly, the findings reinforce the need in human-AI research to examine more precisely how humans and systems contribute to different phases of decision-making and how this affects performance (cf. Parasuraman et al., 2000; Trunk et al., 2020). It should be noted that established models and theories in IS research (e.g., Davis et al., 1989; Venkatesh et al., 2003) indicate that the operators’ usage behavior can also be shaped by perceived usefulness, or there may even be a reciprocal relationship between system use and usefulness (Rai et al., 2002; Sabherwal et al., 2006).

The results further indicate that managers experience greater support if the system is considered as more intelligent in terms of, for example, real-time monitoring of large volumes of data, pattern recognition, data-driven forecasting, and the use of machine learning, logic-based, knowledge-based, and statistical approaches. Theoretically, this supports the assumption from algorithmic management research that intelligent systems can contribute to managerial decision-making by extending human information-processing capacity in complex tasks (e.g., Baiocco et al., 2022; Benlian et al., 2022; Kinowska and Sienkiewicz, 2023). The finding suggests that the managerial value of such systems depends on whether managers perceive them as capable of handling the complexity of employee-related planning. Since the mostly nonsignificant indicator weights limit conclusions regarding the specific contribution of individual AI functions, future studies should examine which AI characteristics are most relevant for managerial support in algorithmic management contexts. In addition, the positive effect of perceived system transparency on planning support indicates that the (perceived) analytical capabilities of a system need to be accompanied by sufficient transparency. In our context, this may be reinforced by managers’ accountability and by the complexity of the planning tasks themselves (cf. Baiocco et al., 2022). Although these specific mechanisms remain unclear, the findings support the central role attributed to system transparency in algorithmic management and human-AI interaction research (e.g., Baiocco et al., 2022; Habbal et al., 2024; Rzepka and Berger, 2018).

Regarding the organizational conditions, the size of the effect on planning support lacks practical relevance and even becomes nonsignificant after including control variables. In combination with the confirmed effect on job satisfaction, this might be explained by the fact that the construct of organizational conditions reflects broader organizational conditions related to the overall planning task, rather than factors limited to system use. This explanation can also be related to IS success research and the D&M IS Success Model (DeLone and McLean, 1992; DeLone and McLean, 2003), which informed the basic structure of our research model. In the D&M IS Success Model, organizational support is reflected in service quality, which refers more directly to system-related support. However, the effects of system use, intelligence, and transparency in our study are consistent with IS success research, which emphasizes that system use, system characteristics, and output-related characteristics contribute to system benefits (DeLone and McLean, 1992; DeLone and McLean, 2003; Petter et al., 2008; Urbach and Müller, 2012). Accordingly, our study indicates that perceived planning support and job satisfaction are shaped by different mechanisms: While the model’s explanatory power for planning support suggests that system-related factors (i.e., extent of system use, intelligence, and transparency) can be seen as highly relevant predictors, job satisfaction appears to depend more on the organizational environment. Thus, even if systems improve planning performance, this may not be sufficient to affect managers’ overall satisfaction with planning tasks. Instead, job satisfaction is likely shaped by a wider range of job-related factors, including social factors or factors resulting from the integration into a broader work context (Parker et al., 2017; Parker and Grote, 2022; Stegmann et al., 2019). Future research should therefore conceptualize organizational conditions depending on whether the focus lies on system-specific benefits or broader outcomes such as job satisfaction.

Regarding the moderation effects, the results show that there is a tendency for task complexity to positively impact the relationship between the extent of system use and planning support, but it was not confirmed to be statistically significant. This aligns with studies showing that system use can improve performance measures regardless of task complexity and supports the conclusion by Kempkes et al. (2024) that the role of task complexity requires further clarification. One possible explanation might be that algorithmic systems may alter the perceived task complexity, for example, by reducing information-processing demands while simultaneously increasing the number of alternatives considered. Thus, in our survey, it may have been unclear whether managers referred to the inherent complexity of the planning task, as intended, or to the complexity perceived after system support. Future studies should therefore pay attention to this distinction when conceptualizing task complexity.

In contrast to task complexity, task relevance positively moderates the relationship between the extent of system use and the perceived planning support, indicating that a greater extent of system use becomes more beneficial when the planning task is perceived as more relevant for the company (in terms of, e.g., impact on the company’s performance and future decisions). This effect may be due to greater rationality of the planning process, which, in turn, is associated with improved decision-making, particularly for important tasks, as shown by Elbanna and Child (2007). This resonates with the assumption that important decisions require more quantitative and data-driven approaches (Elbanna et al., 2020). This finding indicates that even in employee-related planning contexts, organizational consequences remain central for managers’ evaluation of system support. While algorithmic management research often emphasizes the implications for workers (see, e.g., Hillebrand et al., 2025), our results suggest that managers also assess such systems in light of their contribution to organizational outcomes in real-world settings.

The expected impact on workers affected by the planning task significantly weakens the relationship between the extent of system use and perceived planning support. For planning tasks with high worker impact, managers benefit less from a greater extent of system use than for tasks with low impact on workers. This finding is a central contribution of our study, as it challenges assumptions about the benefits of algorithmic management systems, such as enhanced fairness, objectivity, and efficiency (Brione, 2020; De Cremer, 2020; Lee et al., 2021; Newman et al., 2020). While a greater extent of system use generally supports the realization of these benefits, this tends to apply more to tasks with lower expected impact on workers. As the perceived worker impact increases, managers might feel that the system fails to support the nuanced planning required, such as considering social interactions or tacit knowledge, which algorithms can accommodate only to a limited extent (Lee, 2018). This finding underscores that decisions involving significant worker impact should be made through a symbiotic collaboration between human manager and system support (cf. Kalimeris et al., 2022; Tegtmeier et al., 2022; Trunk et al., 2020). The finding also suggests that worker impact should not only be considered from the perspective of the employees (e.g., Bucher et al., 2021; Gagné et al., 2022; Parent-Rocheleau and Parker, 2022). Instead, algorithmic management research should also consider worker impact as a management task characteristic that shapes how managers interact with and perceive planning support.

Implications for system design

Moreover, our study offers implications for planning support system design. The results confirm a positive influence of system transparency on planning support, as discussed in the context of decision-making, for example, by Balasubramaniam et al. (2023), and as examined for intelligent DSS by Wanner et al. (2022). Managers who perceive the system as more transparent report a higher level of planning support. This underscores the importance of designing systems in such a way that they are perceived as transparent by users, especially for employee-related planning tasks, where managers remain responsible for decisions that affect, for example, working location, methods, and remuneration. Our study reveals that the most important indicator for the perception of transparency (based on indicator weights) is the user’s understanding of how the output is produced by internal algorithms. Practically, this can be achieved through local explanations that show which specific steps were taken for a given system output (Rzepka and Berger, 2018; Wanner et al., 2022). In addition, relevant indicators include basic functioning and the data used. This underscores the need for additional global explanations of the system’s internal functioning (Balasubramaniam et al., 2023; Kreisköther and Herbers, 2024; Wanner et al., 2022). Such explanations can help users assess whether the system is suitable for a given planning problem and may enable them to detect false system behavior (Wanner et al., 2022). In fact, transparency is essential for enabling humans to retain decision-making authority and maintain control over the system (Tegtmeier et al., 2022). Regarding data use, managers should be able to see relevant data sources, missing information, assumptions, and indicators of data quality. This is particularly important because planning in production and logistics is shaped by dynamic conditions and a multitude of influencing factors that need to be considered (cf. Gabriel et al., 2022; Kreisköther and Herbers, 2024). At the same time, poor data quality can reinforce biased algorithmic recommendations (Kalimeris et al., 2022). Moreover, when developing systems, particular attention should be paid to self-descriptiveness and learnability, for example, by taking into account the design recommendations of DIN EN ISO 9241-110 (Deutsches Institut für Normung e.V., 2020). Especially when it comes to AI systems, our results reinforce the importance of research on explainable AI systems (Balasubramaniam et al., 2023; Peissl et al., 2022) to enable informed interaction and mitigate perceived opacity.

The finding that the positive effect of the extent of system use on planning support seems to diminish when tasks have a high impact on workers suggests that system developers have not yet fully incorporated the aspects that managers consider relevant in such planning processes. This may not appear surprising, as worker-related planning processes in production and logistics, such as workforce scheduling, often rely on implicit, non-digitalized experiential knowledge (cf. Gabriel et al., 2022). To uncover and take into account implicit knowledge and informal communication, system development and implementation should follow a participatory approach (Latos et al., 2024; Mütze-Niewöhner et al., 2022) that includes workers’ perspectives (cf. Gabriel et al., 2022). Additionally, qualitative research methods could be used to identify influencing factors that systems should take into account for such tasks. One promising approach is to more extensively incorporate human-oriented factors such as worker preferences, team composition, or physical and cognitive limitations of employees (Burgert et al., 2024; Latos et al., 2024; Lee et al., 2021).

Implications for companies

Our study points toward a number of different approaches for companies to increase the support provided by intelligent planning support systems for complex planning tasks in production and logistics. According to the results, perceived planning support is shaped by the managers’ perception of the system’s contribution to decision quality, speed, adaptability, objectivity, and fairness. One approach companies could take is to evaluate existing or new planning support systems based on these aspects to effectively enhance perceived system support. Because objectivity and fairness are difficult to evaluate (cf. Stich et al., 2013), integrating the workers’ perspectives is essential (cf. Latos et al., 2024; Mütze-Niewöhner et al., 2022).

As the extent of system use shows a significant positive effect on the perceived planning support, companies could consider whether system functions enable more extensive use during the various planning phases. As systems offer different functions and potential benefits for the different planning phases (cf. Onnasch et al., 2014; Trunk et al., 2020), the integration into the individual phases should be examined in a differentiated manner. However, the moderation analysis indicates that the nature of the planning task must also be taken into account. Companies may particularly consider integrating systems more extensively into production and logistics planning when tasks are highly relevant (e.g., capacity planning for bottleneck resources or sequencing of high-value orders). For tasks that imply high worker impact, however, greater system use should be examined particularly carefully in order to avoid negative effects not only on perceived planning support but also on employees affected by the planning task (cf. Gagné et al., 2022; Parent-Rocheleau and Parker, 2022). In practice, companies should avoid fully delegating such high-impact planning tasks to a system. Instead, intelligent systems should be used to generate transparent proposals and highlight constraints, while human managers retain responsibility for the final decision (cf. Kalimeris et al., 2022; Tegtmeier et al., 2022) and actively integrate local knowledge and employee input, for example, through consultation with labor unions or team representatives. From the workers’ perspective, it should always be clear who holds responsibility, and transparency must be ensured (cf. Parent-Rocheleau and Parker, 2022; Tegtmeier et al., 2022), as workers may become frustrated when decisions are not intelligible to them (cf. Kellogg et al., 2020).

The positive contribution of perceived system intelligence emphasizes that companies should critically evaluate the availability and usefulness of intelligent system functions. Based on the indicator weights, real-time monitoring of large data volumes in particular contributes to the perception of system intelligence. However, the benefits of AI functions depend on the amount and quality of available data (Kalimeris et al., 2022; Rossmann and Wald, 2024; Trunk et al., 2020). In production and logistics in particular, multiple systems with discontinuities and missing interfaces often impair data availability and quality (Jacobs et al., 2011; Stich et al., 2013). Before scaling up AI-based functions, organizations should therefore invest in strengthening data management (Bauernhansl et al., 2014; Kozioł-Nadolna and Beyer, 2021; Stich et al., 2013). In addition, ethical considerations and applicable regulations must be taken into account (cf. Balasubramaniam et al., 2023; Peissl et al., 2022; Wood, 2021). Given the impact on workers, companies seeking to use AI functions should first encourage managers to explore AI in a safe, collaborative setting (Kolbjørnsrud et al., 2017).

Due to the influence of perceived system transparency on planning support, companies should not underestimate the importance of the transparency of the systems used. Beyond selecting systems that provide global and local explanations, companies can support transparency through user training initiatives that provide the opportunity to explore intelligent systems in a protected environment (Wanner et al., 2022). In practice, this could include workshops in which managers explore how changes in constraints or input data affect, for example, scheduling recommendations or route suggestions in distribution. Since our results also reveal that understanding the system’s purpose is a relevant indicator for the perception of transparency, organizations should clearly communicate its intended use and limits.

Lastly, managers’ job satisfaction is shaped by the quality of organizational conditions that companies can actively influence. Not only material resources (personnel, technical, financial) are important, but also immaterial resources (advocacy, trust, time). Companies should ensure adequate organizational support, including training opportunities, adequate means and resources, sufficient time, as well as suitable procedures tailored to the planning process. Furthermore, companies should foster a supportive culture. This includes both promoting the use of digital systems, which is the most important factor based on the significant indicator weights, and supporting individual managers. Particularly in the context of tasks that have a high impact on employees and are of great importance to the company, confidence in the manager’s ability to perform the planning task effectively might be essential. This also applies to situations where managers face challenges in the planning process, which was identified as another relevant indicator in our study.

Limitations and future directions

The data used for evaluation are based on self-assessments by participants who were contacted via a survey panel provider. This could have led to distortions, for example, due to dishonest or careless responses or biases resulting from participation in previous studies (Leiner, 2019; Schonlau and Toepoel, 2015). Even though a multi-stage screening procedure was carried out to ensure sufficient data quality, such distortions cannot be completely ruled out. Furthermore, due to the cross-sectional study design, common method bias may occur, referring to a distortion of correlations between exogenous and endogenous variables that is driven by the survey method itself (Kock, 2015; Podsakoff et al., 2003). To mitigate its risk, several measures (Podsakoff et al., 2003) were considered in the survey design, including respondent anonymity, separation of variables across pages, and pretesting with both researchers and practitioners to ensure clarity and unambiguous wording. While the bias cannot be ruled out, VIF values (see Supplementary Material) provide no indication of common method bias in the present study (Kock, 2015). It should also be considered that due to the cross-sectional design, the causal interpretation of the findings is limited. Although SEM is often associated with causal interpretation, such interpretation requires strong theoretical grounding and, ideally, longitudinal data (Hair et al., 2022; Werner et al., 2016). Future research may therefore benefit from longitudinal study designs to examine whether changes in system use actually lead to changes in perceived planning support over time.

A further limitation lies in the heterogeneity of the sample. It covers a fairly wide range of planning tasks and systems and combines production and logistics, which limits the ability to derive implications for specific tasks or systems. In fact, the ex-post analysis of the control variables shows that the perceived planning support is lower for logistics planning tasks than for planning tasks that are assigned to a production context. Although this effect does not substantially alter the relationships examined in this study, future research should focus on an in-depth comparison of relevant differences between the characteristics of planning tasks in production and logistics. It was additionally shown that the examined user characteristics (e.g., age, gender, tenure, management position) did not substantially alter the results. Nevertheless, even though our results indicate a high, moderate level of explanatory power for the planning support, further user characteristics may influence perceived planning support or moderate the examined relationships (Kempkes et al., 2024; Rzepka and Berger, 2018; Shepherd and Rudd, 2014; Stowers et al., 2017). Future research could therefore integrate related factors, such as attitudes toward AI (Cao et al., 2021) or affinity for technology interaction (Franke et al., 2019). With regard to the largely newly developed measurement models used in this study, only two indicators (measurement model for SAT) had to be excluded in the course of the measurement model evaluation. Nevertheless, the assessment of convergent validity of the formatively specified measurement models indicates that the single-item measures used for SYS, TRA, COM, and IMP may not adequately reflect the conceptual scope of the respective constructs and should therefore be further developed in future research.

Regarding algorithmic management research, we adopt a relatively broad understanding of algorithmic management by examining complex planning tasks that affect workers and are supported by systems to a certain extent. Our study therefore does not address fully automated algorithmic management (cf. Baiocco et al., 2022; Wood, 2021) and focuses on planning rather than other algorithmic management functions, such as commanding, evaluation, or controlling (Baiocco et al., 2022; Fayol, 2016). However, this broader perspective allows us to ground our findings in real-world practice, where managers typically remain involved in algorithmic management (cf. Jarrahi et al., 2021), and to examine the underexplored perspective of managers in the application field of production and logistics.

Since our study is based on how managers perceive planning support in employee-related planning tasks, it does not permit conclusions about objective performance metrics. Future studies could therefore combine manager surveys with objective planning performance indicators, such as planning time or error rates, to examine whether perceived planning support corresponds to actual planning outcomes. Similarly, fairness and objectivity as well as the expected impact on workers were assessed from the managers’ perspective, but not by the workers affected. Accordingly, further research is needed on the relationship between managers’ perceptions and the actual implications for employees. Future research would therefore benefit from integrating both managers’ and workers’ perspectives to enable a more comprehensive systemic understanding of these relationships (Hillebrand et al., 2025). For instance, paired manager-worker designs could compare managers’ perceptions of planning performance with workers’ experiences in terms of, for example, fairness, autonomy, and workload (although access to matched samples may be difficult in practice). Further central avenues for future research arise from our study: First, the question arises as to how planning support systems need to be designed for specific planning tasks so that a greater extent of system use contributes to managerial planning performance, specifically for planning tasks with high impact on workers (e.g., in terms of integration of human-oriented planning criteria). Second, while our study shows that a greater extent of system use and perceived system intelligence increase perceived planning support, it remains unclear how system use and intelligent system functions benefit planning across different phases of the planning process. Future research should therefore investigate how system support in specific planning phases affects not only managerial performance but also outcomes for affected employees.

The present study contributes to algorithmic management research by examining how managers perceive planning system support for complex planning tasks that affect workers in production and logistics. The hypothesized research models were empirically evaluated using PLS-SEM based on responses from 494 managers via an online questionnaire. The results indicate that managers’ perception of planning support for worker-related planning tasks is shaped not only by traditional performance indicators, such as contributions to faster planning and better results, but also by the system’s contribution to objectivity and fairness. While the extent of system use, system intelligence, and system transparency emerge as relevant predictors of planning support, managers’ job satisfaction depends on the quality of organizational conditions. Furthermore, the effect of the extent of system use on planning support is moderated by both the relevance of the planning task and its worker impact, whereas the moderating role of task complexity could not be confirmed. Our study thus highlights that the impact on employees should be considered not only from the employees’ perspective, as previous research has focused on, but also as a characteristic of algorithmic management tasks that shapes how managers perceive system support. Future research should capitalize on the insights presented in this study and further explore the human-centered design of algorithmic management systems, delve deeper into the specific contributions of system support to particular phases of the planning process, and incorporate the perspectives of both managers and workers to foster collaboration between managers and systems that can harness the full potential of technology while enhancing support for both managers and employees.