For Better or for Worse: The Impact of Workplace Automation on Work Characteristics and Employee Well-Being

Abstract

This article examines the consequences of implementing an automation technology (i.e., Robotic Process Automation; RPA) for work characteristics and employee well-being. Based on the job demands-resources framework we examined to what extent the utilization of RPA was related to job resources (i.e., autonomy and task variety) and a job demand (i.e., information processing), and to what extent these work characteristics were related to work engagement and exhaustion. Data were collected among 420 employees working for a Dutch ministry where RPA was recently introduced. Structural equation modeling revealed that RPA use was negatively related to both autonomy and task variety, which formed a threat to employee work engagement. Contrary to our expectations, RPA use was unrelated to information processing and subsequent exhaustion. These findings put emphasis on the importance of designing new technologies with sufficient job resources to create and maintain a healthy and motivated workforce during and after implementing workplace automation.

1. Introduction

Technological innovations and work process automation play a key role in meeting the increased demands that organizations face in today’s competitive and fast-changing market [1]. To become and remain efficient, automation and robotics offer many solutions for the potential optimization of work processes. Robotic Process Automation (RPA) is currently one of the most used tools in business process automation that stimulates higher organizational productivity [2]. RPA is a software robot that uses the interface of an already present computer system and mimics the actions of a human employee. RPA can automate work processes that are administrative, well-structured, and repetitive of nature [3, 4, 5]. Typical tasks that can be done by RPA entail, for example, processing incoming emails and orders, transferring data from one digital system to another, and searching for and communicating with potential new hires. Organizations greatly benefit from the implementation of RPA, mainly due to increased process speed and production growth, as well as error reduction [6].

However, although there are many advantages to the commissioning of RPA, it is estimated that between 30 and 50% of all RPA implementations fail [7]. This relatively high number of unsuccessful RPA implementations is in line with the general notion that digital transitions often do not result in desired outcomes, such as enhanced productivity and efficiency. A potential explanation for failing technological implementations is the lack of consideration for the employees who have to work with a new technology [8]. With this study we aim to gain more insight in this human-technology interaction [9] by investigating to what extent the implementation of an RPA technology impacts work characteristics and employee well-being.

In-depth knowledge on the consequences of digital transitions for the work characteristics and employee well-being is increasingly important, because both are to a large extent predictive of overall employee performance and organizational productivity [10, 11]. By studying this issue we contribute to the existing body of knowledge on workplace automation and information systems in two important ways. First, this study aims to address the research gap concerning the impact of workplace automation on individual work experiences. Whereas earlier research has mainly focused on the effects of technological innovations on employment and labor market composition, we argue that more understanding of the relationship between the implementation of an automation system and work characteristics, as well as well-being of those who have to work with these new technologies can contribute to the successful implementation of technological innovations. This can help organizations that want to innovate and invest in better designed jobs, as well as lead to better implementations of new technologies such that sustainable employee performance can be consolidated and organizational efficiency can be achieved. Second, we aim to contribute to the growing body of literature on workplace automation and the application of RPA technology [5, 6, 12, 13]. More specifically, we examine to what extent RPA influences the work experiences of employees and whether RPA technology results in the desired outcomes with regard to a decrease in job demands, an increase in job resources, and in turn enhanced employee well-being. In sum, with this research we want to achieve a better understanding of how employees experience working with RPA, so that such an innovation in the future actually contributes to what it is intended for and does not generate negative side effects. In the following sections, we will explain how RPA can influence work characteristics and how these work characteristics are related to employee well-being.

2. RPA and work characteristics

Work process automation changes the way work is performed and perceived by employees. For example, the use of robots that automate heavy manual labor can result in less physically straining job tasks for employees, while at the same time requires them to handle new machinery. Similarly, the use of chatbots can significantly decrease interpersonal interactions at work, but can potentially also result in employees to feel alienated from their original work role. Hence, the introduction of a new technology and workplace automation can have both simultaneously positive and negative influences on employees’ work experiences.

The idea behind RPA technology is that administrative work processes become more streamlined and efficient, so that employees have to spend less time on performing repetitive work tasks [3]. Compared to traditional automation systems (e.g., BPM, CRM) RPA can automate many different work tasks, is easy to implement and use, and does not require modification of existing IT infrastructures [5]. As such, RPA is unique in a sense that it requires minimal human intervention, can be applied to a range of business applications, and is designed such that end users can make changes without the need to possess extensive programming skills. Although RPA does not automate and replace complete jobs, it does substantially change certain tasks and the way jobs are designed [9, 14]. Therefore, the implementation of RPA is likely to change the way employees (perceive their) work and thus can result in better or worse designed jobs, which likely has a profound impact on important outcomes related to employee work experiences and well-being [9].

From a work design perspective, the job demands-resources model states that all work characteristics can be divided into job demands and job resources [15]. Job demands, such as workload, time pressure, and role conflict, refer to all aspects of a job that require continuous cognitive or emotional effort and are related to physiological and/or psychological costs. Job resources are those aspects of a job that help employees cope with high job demands, attain work goals and performance, and stimulate professional growth. Examples of job resources are feedback, task variety, support, and autonomy [15]. In line with Demerouti [16] and Parker and Grote [9], we argue that during and after the implementation of a new technology and workplace automation both job demands and job resources are subject to substantial changes. A recent systematic literature review [17] on the impact of the implementation of technological innovations on core work characteristics, showed that the implementation of a new technology was associated with intensified job demands, including job complexity and workload. Additionally, the relationship between the implementation of a technological innovation and job resources was predominantly positive, especially with regard to autonomy and control. These findings suggest that although job demands tend to increase after the introduction of a new technology, job resources seem, at least to some extent, compensate for these increased demands. However, in finding an answer to the question which work characteristics are susceptible to change after the introduction of an RPA technology, it is important to take a closer look at the defining features of RPA.

Starting with job resources, as mentioned above, one of the main goals of RPA is to take over administrative and repetitive work tasks from employees [5]. This means that the use of RPA frees time that employees otherwise had to spend on monotonous tasks. Considering that RPA generates more time for other aspects of the job, this should allow employees to exert more control over their work structure and tasks. As such, we expect that RPA use relates to more autonomy at work for employees. This notion is also supported by a qualitative interview study of Engberg and Sördal [18], who found that the introduction of RPA enhanced the experienced freedom of employees to independently organize their work schedule and tasks.

Second, although RPA is useful in replacing structured and repetitive tasks, it is less suitable to take over complex work that requires more advanced problem-solving skills and abilities [3]. Taking into account that RPA technology releases employees from carrying out repetitive work duties, it simultaneously leaves more room to perform other and more challenging tasks. In line with this argumentation, several qualitative studies have shown that, overall, employees experienced that RPA enabled them to advance their skill set. In addition the introduction of RPA enabled them to devote time to their professional development and growth while performing new and challenging tasks [18, 19]. These findings suggest that RPA creates space for employees to focus on a variety of challenging work tasks. Therefore, we also expect that the use of RPA positively relates to task variety, which entails the extent to which employees experience variety in their job content and can perform a wide range of tasks that require different skills [20].

Turning to job demands, reference [21] indicates that RPA is able to take over and process up to 300% more information compared to human employees. This increase in productivity is due to the fact that RPA can complete a large range of administrative tasks in a fraction of the time compared to actual employees and can work throughout the night and weekend. Additionally, RPA is relatively easy to implement and configure, meaning that employees can use RPA, as well as make changes in how tasks are performed, without an extensive technical background [22]. This implies that employees are no longer bothered with continuously having to deal with processing and analyzing large amounts of information and are able to easily adjust the system during the implementation process based on the requirements and needs of their job. Information processing refers to the amount of data and information that employees are required to monitor and manage in their job [20, 23]. Considering that RPA takes over data and information processing tasks to a substantial extent, we propose that RPA completes a large range of administrative tasks in a fraction of the time compared to actual employees. Taken together, with regard to changes in job demands and job resources we formulate the following hypotheses:Hypothesis 1:

RPA use is positively related to (a) autonomy and (b) task variety.

Hypothesis 2:

RPA use is negatively related to information processing.

3. Work characteristics and employee well-being

Employee well-being refers to a passive or active work-related affect and individual’s evaluation of the quality of experiences at work [24]. In this study, we included work engagement and exhaustion as indicators of work-related well-being, considering that both are regarded as important factors in the operationalization of employee well-being [25, 26]. In relation to work-related well-being, the JD-R framework proposes two independent underlying processes [15]. First, a health impairment process, in which continued exposure to high job demands results in strain, burnout, and an overall decline in health-related outcomes. Second, a motivational process is proposed, in which access to sufficient job resources protects employees against high job demands and leads to motivation, work engagement, and increased productivity. Following these central assumptions, we argue that higher levels of autonomy and task variety associated with RPA use instigate the motivational process as proposed in the JD-R framework. To clarify, autonomy and task variety are key job resources, which consistently have been found to be predictive of work engagement and performance outcomes (for overviews see [27, 28]. Consequently, we propose that employees who can turn over their repetitive tasks to RPA are likely to experience more autonomy and task variety, and in turn feel more engaged. In addition, consistent exposure to high job demands, including workload and information processing, are linked to increased levels of exhaustion and burnout (e.g. [29]. With regard to information processing, we expect that employees who can transfer their administrative responsibilities to RPA and thus on a daily basis deal with substantially less repetitive job tasks, experience lower levels of information processing (see Hypothesis 2). In turn, lower levels of information processing are likely to relate to lower levels of exhaustion (i.e., a positive relationship). Therefore, we hypothesize the following:Hypothesis 3:

(a) Autonomy and (b) task variety are positively related to work engagement.

Hypothesis 4:

RPA use is indirectly related to more work engagement through (a) autonomy and (b) task variety.

Hypothesis 5:

Information processing is positively related to exhaustion.

Hypothesis 6:

RPA use is indirectly related to less exhaustion through information processing.

4. Method

4.1 Sample and procedure

Data for this study was collected via an online questionnaire. We recruited data from employees within two large departments of a Dutch Ministry (N = 420). The response rate was 37.33%. In 2019 this Ministry introduced and implemented RPA in their organization. Employees working in the two departments typically hold office jobs, such as administrative workers, financial and legal experts, project managers and members, and HR representatives. We invited both employees who could turn over certain aspects of their work to RPA (i.e., RPA users, who make or control work processes and provide input for the RPA robot; N = 140) and employees whose work was not directly impacted by RPA (N = 280). In the questionnaire employees were asked whether their work was somehow impacted by the introduction of RPA and if so, in what way their work has changed and how they interacted with the robot. Based on these answers a distinction could be made between RPA users and non-RPA users. The group of RPA users consisted of employees from whom RPA took over one or several administrative and repetitive work tasks, such as scanning and filing emails and documents, extracting data, and generating (mass) emails. Additionally, this group also consisted of employees that made or controlled RPA work processes and output, as well as provided new input for the RPA robot. All employees received an email from the Ministry with information about the aim of the study, a link to the online questionnaire, and an explanation of the confidentiality was offered to all respondents. This study obtained approval of the Ethics Review Board.

In the total sample, 54.90% was male and the average age was 48.96 years (SD = 10.85). The mean job tenure was 14.61 years (SD = 11.22) and on average respondents worked for 33.48 hours a week (SD = 4.87). Most employees worked in jobs that required a Bachelor’s Degree (48.10%) or an Associate Degree (12.10%). In terms of demographical variables (i.e., gender, age, weekly workhours, and contract type) RPA users did not differ significantly from non-RPA users. Additionally, we tested for differences between RPA users and non-RPA users on the study main variables by conducting independent samples t-tests in SPSS. Table 1 shows the results of these analyses. RPA users reported significantly lower levels of autonomy, task variety, and information processing compared to non-RPA users.

	RPA users		Non-RPA users		t-test
	M	SD	M	SD
Autonomy	3.62	0.74	3.97	0.65	4.86**
Task variety	3.33	0.82	3.68	0.70	4.59**
Inform. Processing	3.83	0.65	4.02	0.67	2.82**
Work engagement	4.71	0.99	4.69	0.90	−0.22
Exhaustion	2.07	0.75	2.01	0.68	−1.15

Table 1.

Results of t-tests comparing RPA users (N = 140) and non-RPA users (N = 280) on the study variables.

Note: n = 420. * p < 0.05, ** p < 0.01.

4.2 Measures

Frist, autonomy was measured with a Dutch translation of three items of the Work Design Questionnaire (WDQ; [20]). An example item of this scale is: “The job allows me to make a lot of decisions on my own”. Cronbach’s α was 0.80. Task variety was measured with two items of the WDQ [20]. One of these items is “The job involves a great deal of task variety”. Information processing was measured with two items of the WDQ . An example of these items is: “The job requires me to monitor a great deal of information”. Work engagement was measured with the three-item version of the Utrecht Work Engagement Scale [30]. An example item of this scale is: “At my work, I feel bursting with energy”. Cronbach’s α for this scale was 0.82. Exhaustion was measured with three items of the Burnout Assessment Tool (BAT [31]). One of the items is: “I feel mentally exhausted at work”. Cronbach’s α for this scale was 0.81.

4.3 Strategy of analysis

First, we evaluated the measurement model using confirmatory factor analysis (CFA). Latent variables (i.e., autonomy, task variety, information processing, work engagement, and exhaustion) were modeled with scale items. The following fit indices were used to evaluate model fit: the Comparative Fit Index (CFI), the Tucker-Lewis Index (TLI), and the root mean square error of approximation (RMSEA). With CFI and TLI values above 0.95, and RMSEA below 0.06, model fit is acceptable [32]. Second, we tested the proposed research model using structural equation (SEM) with the AMOS software package [33]. To assess the specific indirect effects of autonomy and task variety in the relationship between RPA use and work engagement, as well as the specific indirect effect of information processing in the relationship between RPA use and exhaustion, we applied the phantom model approach [34]. In addition, to test the robustness of our proposed research model, we tested an alternative model, that proposed a relationship between RPA use and work engagement, and in turn, autonomy and task variety. Additionally, this alternative model proposed an indirect relationship between RPA use, exhaustion, and information processing.

5. Results

5.1 Descriptive statistics

In Table 2 the descriptive statistics, including the means, standard deviations, and correlations of the variables in this study can be found. Job level was the only demographic variable that correlated significantly with several of the outcome variables (i.e., autonomy, task variety, and information processing). Therefore, we controlled for job level in our further analyses.

	Mean	SD	1	2	3	4	5	6	7	8
1. Age	48.96	10.85	—
2. Job level	2.72	0.68	−0.01	—
3. System use	0.33	0.47	−0.08	−0.25**	—
4. Autonomy	3.85	0.70	0.03	0.34**	−0.24**	—
5. Task variety	3.56	0.76	−0.05	0.48**	−0.23**	0.47**	—
6. Information processing	3.96	0.67	0.07	0.37**	−0.14**	0.36**	0.36**	—
7. Work engagement	4.70	0.92	0.05	0.05	0.01	0.28**	0.32**	−0.14***	—
8. Exhaustion	2.04	0.71	−0.04	−0.01	0.06	−0.14**	−0.13**	0.09	−0.27**	—

Table 2.

Descriptive statistics and inter-correlations of the study variables N = 420.

Note: n = 420. * p < 0.05, ** p < 0.01.

5.2 Testing the hypotheses

The measurement model, including autonomy, task variety, information processing, work engagement, and exhaustion as latent variables, showed a very good fit to the data: χ² = 144.00, df = 55, CFI = 0.96, TLI = 0.94, RMSEA = 0.06. All factor loadings loaded significantly on their respective latent factor and ranged between 0.63 and 0.97.

In Hypothesis 1a and 1b we predicted that RPA use was positively related to (a) autonomy and (b) task variety. Contrary to our expectations, our analysis showed an opposite relationship, namely that RPA use was significantly and negatively related to both autonomy (β = −0.19, p < 0.01) and task variety (β = −0.13, p < 0.05), thereby not supporting Hypotheses 1a and 1b.

In Hypothesis 2, we predicted that RPA use would be negatively associated with information processing. Although this relationship was indeed negative, it was not significant (β = −0.06, p = 0.17), and thereby not in support of Hypothesis 2.

In line with Hypotheses 3a and 3b, we found that autonomy (β = 0.13, p = 0.01) and task variety (β = 0.26, p < 0.05) were as expected indeed positively and significantly related to work engagement, thereby confirming Hypothesis 3.

Turning to Hypotheses 4a and 4b and the indirect relationship between system use and work engagement through both autonomy and task variety, the data showed that this combined indirect effect was negative and significant (estimate = −0.06, p < 0.02 with a bias-corrected confidence interval ranging from −0.10 to −0.02). To assess the specific indirect effects of autonomy and task variety separately in the relationship between system use and work engagement, the phantom model approach was applied [34]. The specific indirect effect of autonomy in the relation between system use and work engagement was indeed negative and significant (estimate = −0.10, p = 0.02), thereby not in support of Hypothesis 4a. The specific indirect effect of task variety in the relation between system use and work engagement was also negative and significant (estimate = −0.06, p = 0.02). As such, Hypothesis 4b was also not confirmed.

In contrast to Hypothesis 5, we found no significant relationship between information processing and exhaustion, (β = 0.08, p = 0.17). As such, Hypothesis 5 was not supported by the data.

Hypothesis 6, in which an indirect effect of information processing in the relationship between RPA use and exhaustion was hypothesized, was not supported by the data (estimate = −0.01, p = 0.16 with a bias-corrected confidence interval ranging from −0.02 to 0.00). As such, Hypothesis 6 was not confirmed.

Overall, the proposed structural model showed a good fit to the data: χ² = 210.235, df = 78, CFI = 0.94, TLI = 0.92, RMSEA = 0.06. Figure 1 shows a schematic representation of all study’s results.

Figure 1.

Overview of results of structural equation modeling. Note: n = 420. * p < 0.05, ** p < 0.01.

Last, we tested a plausible alternative model, in which system use was related to autonomy and task variety, through work engagement. Additionally, in this alternative model we proposed that system use was indirectly related to information processing through exhaustion. The alternative model showed a lower overall fit with the data (χ² = 248.549 df = 80, CFI = 0.92, TLI = 0.90, RMSEA = 0.07). After comparison of the two models, the proposed research model yielded a significantly better fit (Δχ² = 38.314, Δdf = 2, p < 0.01).

6. Discussion

To gain a better understanding of the impact of workplace automation on the quality of work and employee well-being, we examined to what extent the introduction of RPA relates to work characteristics and subsequent work-related employee well-being. We drew on the JD-R framework [15] to argue for a positive relationship between RPA use (vs. non-use) and the job resources autonomy and task variety, and in turn work engagement. Additionally, we proposed a negative association between RPA use and information processing and subsequently exhaustion. Contrary to our expectations, the results showed that RPA use was negatively related to both autonomy and task variety, which in turn were positively related to work engagement. Moreover, the indirect effects of both autonomy and task variety were negative in the relationship between RPA use and work engagement. These results indicate that the introduction of a workplace automation, and more specifically the use of RPA, is at the expense of work engagement for employees who have to deal with this new technology through a decrease in job resources. Furthermore, we found no significant relationship between RPA use and information processing and in turn exhaustion, suggesting that job demands were not significantly affected by the introduction of a new workplace technology.

Unexpectedly and remarkably, the findings of this study demonstrate that working with RPA as a new technology is associated with lower levels of both autonomy and task variety. This is a worrying finding in itself, especially because autonomy and task variety are important predictors of work engagement, as also underlined by the results of this study. Moreover, many studies have demonstrated that a lack of job resources, including autonomy and task variety, is also associated with other negative work outcomes such as burnout, turnover intentions, lower levels of learning behaviors, as well as a decrease in motivation, proactivity, and performance [35, 36, 37]. This further underlines that it is very important to conserve job resources when introducing a new work process automation.

A possible explanation for the negative association between RPA use and autonomy and task variety - and thus the opposite intended effect of RPA - could lie in the ongoing implementation process of a workplace automation [9]. In this case, RPA was relatively recently introduced within the organization, meaning that the system was still in continuous adjustment to the specific demands of the organization and employees. Although RPA takes over well-structured and repetitive work tasks, the provided output still needs to be regularly checked by human employees. Therefore, it could be that employees working with RPA were still spending considerable time to examine and correct potential system mistakes and updating the robot to certain tasks and needs [38]. Considering that employees needed to search, report, and adjust RPA system errors, this could lead to a decreased sense of autonomy and control. Additionally, because this also requires a different set of work skills, room to engage in new and challenging tasks could be limited. Moreover, employees who had to work with RPA could not freely choose whether RPA took over certain work tasks they previously performed themselves. It could be the case that this lack of individual influence on the use of RPA resulted in a decreased feeling of control. As such, because employees were required to use and learn RPA and did not have a final say in whether RPA was implemented in their work or not, their autonomy may be threatened and thus reduced.

Taken together, the present study findings implicate that after the implementation of a work process automation technology, employee engagement, and thus well-being, is at risk due to a significant decrease in core job resources for RPA users. Additionally, the results of this study showed that RPA use is not related at all to information processing. A possible explanation for this finding may be that information processing simultaneously increases and decreases for RPA users, thereby canceling out any effects. More specifically, it could be that while RPA use relates to less administrative information that has to be analyzed and processed, employees do have to process more information associated with learning to work and getting familiar with RPA, as well as controlling and adjusting RPA processes. In that case, it would be useful to distinguish between different types of information processing related to a digital transition. For instance, information processing associated with the eventual effects of a new technology on individuals work content (and thus an expected decrease in simple and monotonous work tasks), and information processing related to the implementation process and learning a new technology.

Alternatively, other job demands than information processing could be taken into account when examining the impact of RPA on work characteristics. For instance, workload and role conflict could be potential interesting factors in light of the implementation of RPA, considering that the job content is likely to change due to the implementation of a new work process automation technology. Additionally, in the initial phase after a digital transition a new system requires new routines and knowledge, which is likely to have an impact on employees’ workload [39].

Last, we did not found the expected positive relationship between information processing and exhaustion. Exhaustion is often regarded as a more distal outcome compared to work engagement, because employee exhaustion only develops after repeated exposure of (high) job demands [40]. Due to the cross-sectional design of this study that also was conducted only two months after the implementation of RPA, it could be that the health impairment processes, as proposed by the JD-R framework, had not been set in motion yet.

6.1 Theoretical contributions

A first important contribution of the current study is that it demonstrates that the relationship between the implementation of technological innovations at work and employee well-being via work characteristics is not straightforward. Our results show that, although RPA is often introduced with the intention to lessen the burden on employees concerning monotonous, repetitive work tasks, this goal is not necessarily achieved. Based on the present study, it seems that job demands do not decrease for those employees that work with the new technology. More importantly and also contrary to our expectations, the use of a new system at work was related to lower instead of higher levels of autonomy and task variety, meaning that job resources of system users seem to decrease after implementation of a work process automation. Thus, our findings demonstrated that the introduction of a new technology did not lower demands. In fact, it even generated less resources in that it created less space for employees to take control over their work and engage in challenging and a wider variety of tasks. Taken together, these findings provide support for the existence of a technology paradox, in that the potential of a new workplace technology does not necessarily results in desired organizational and individual outcomes. Specifically, the implementation of a work process automation should not be at the expense of employee job resources.

Second, the present study contributes to the emerging literature on RPA [5, 6] and offers more insight into workplace automation, and specifically the implementation of RPA, on employee experiences. Whereas earlier studies on the impact of workplace automation and RPA showed overall positive associations with job resources [18, 41, 42], the results of this study present a different picture. Our findings suggest that automation does not always result in a desired reduction of demands, and more importantly, that it poses a potential treat to well-being via a decrease in job resources. To the best of our knowledge this is one of the first papers that focusses on both job resources and a job demand following the implementation of a workplace automation. To gain a deeper understanding of the complex human-technology interaction, future research should place emphasis on changes in both challenging and hindering demands following a digital transition, as well as on job resources that can help employees cope and perform with automation and RPA.

6.2 Limitations and future research

This study has several limitations. First, due to the cross-sectional study design, we cannot draw conclusions about the causal relationships between RPA use, the examined job resources and demand, and work engagement and exhaustion. Although we carefully followed the core premises by the JD-R model [15], it could be the case that some proposed relationships are reciprocal. For instance, it might be that employees who experience high levels of work engagement also perceive more job resources in their work [43]. Earlier studies (e.g. [44, 45] indeed found that job resources and work engagement influence each other in both directions, suggesting a gain cycle in which the presence of job resources and work engagement reinforce each other reciprocally. These findings further underline the importance of more longitudinal research to investigate such bidirectional relationships in a digital transition context. Additionally, when investigating the influence of a technological implementation on employees’ job quality and work experiences, future research could apply a longitudinal study design, in which users and non-users are compared at several measurement points, including pre- and post-implementation. Since employees included in the present study were not randomly assigned to use RPA, the selection of employees to use RPA could be correlated with their perceptions of task variety and job automation. For instance, some employees are more capable and can handle more task variety, so they may be more likely to be selected to use RPA. Thus, it is a challenge to determine the causal effect. Therefore, a within-group pre-post design would have been better and is recommended for future studies.

In addition, the use of self-reports could lead to common method bias [46]. However, additional to the good fit of the measurement model, we conducted Harman’s single-factor test, which demonstrated that variance in the data was not due to a single underlying factor and thus indicating that common method bias was not a problem in this study. Moreover, it can be argued that constructs reflecting individual states, such as work engagement and exhaustion, as well as perceived work characteristics, can best be evaluated by the individual actor, and are not necessarily suitable to cross-validate with other-ratings.

Finally, in this study we focused on (only) three specific work characteristics that were likely to be influenced by the introduction and use of RPA, namely autonomy, task variety, and information processing. The choice of these work characteristics was based on earlier qualitative research findings on the impact of RPA on job resources [18]. Moreover, we reasoned that the implementation of RPA lessens the amount of information that needs to be processed by human employees, as well as generates more control and room for other challenging tasks. Considering that we found no relationship between RPA use and information processing, it would be particularly interesting to uncover if and how other job demands of system users are affected by the implementation of a workplace automation. For instance, future research could further differentiate between challenging and hindering job demands, were job hindrances (e.g., job insecurity, role conflict, and constraints) are associated with exhaustion and job challenges (e.g., workload and cognitive demands) with engagement [47]. It could be that the implementation of new technologies at work simultaneously incite hinderances and challenges for employees. More insight into how workplace automation affects these specific job demands could contribute to a better implementation of new technologies and optimize adaption among employees who have to work with these technologies [9]. Similarly, it would be relevant to examine what and how job resources can help employees to achieve their goals and stay motivated during the introduction of a new technology [16]. In sum, it is of key importance to design jobs in such a way that a digital transition involves both challenging and realizable job demands, as well as sufficient job resources to stimulate employee performance and well-being. A human-centered approach to workplace automation and job design with particular consideration for employee work experiences is of key importance in reducing the technology-paradox and optimizing the full potential of technology [8].

6.3 Practical implications

The present study connects to a broader debate on the quality of work in the rapid-changing contemporary world of work [48, 49]. Both employees and managers benefit greatly from a healthy and motivating work environment, especially in times of digital transition and widespread automation within organizations. The findings of this study provide further insight into how technological innovations relate to employee well-being through work characteristics and underline the importance of stimulating autonomy and task variety in order to safeguard employee motivation after introducing a workplace automation.

Facilitating a work environment in which employees have access to sufficient job resources that help them deal and work with technological advancements, as well as enabling good performance, is one of the most important implications for practice. Based on the results of this study and in line with the recommendations of reference [16], organizations and HR practitioners should take responsibility during and after the implementation of new technologies. One of the focal points should be to (re)design jobs in such a way that technological innovations turn into a resource itself by closely paying attention to the needs and concerns of users. Carefully identifying how job demands may change after a digital transition can also help organizations to offer appropriate job resources for employees to cope with changes in their job demands. For instance, access to sufficient training and education can help employees to become acquainted and more proficient in using a new technology, and thereby also reducing workload, anxiety, and job insecurity [50]. Additionally, providing feedback and support from the organization and managers are two main resources for employees that help them deal with the negative consequences of job demands [51]. Furthermore, another way to successfully gain and maintain well-being and motivation during a digital transition, is by facilitating employee job crafting [52] which refers to an individual proactive strategy to seek out relevant job resources that can help employees during technological change. Managers and organizations play a key role in creating a work environment in which employees feel encouraged to engage in job crafting and proactively seek out resources they need to adjust and perform [53].

In sum, during and following the implementation of a technological innovation, organizations and managers should be aware of changes in job demands and needs from employees, as well as focus on providing adequate resources for employees to cope with these demands and to stimulate optimal performance with a new technology.

7. Conclusion

The current study demonstrates that the introduction of a workplace automation system may have a profound negative impact on employee job resources. More specifically, the findings indicate that use of an automation system relates to lower levels of autonomy and variation in work tasks, forming a serious threat to the work engagement of employees who have to work with the new system. As such, this study shows that organizations should take a close look at and take into account potentially affected job resources due to the implementation of a workplace automation. Importantly, focusing on stimulating relevant job resources, such as autonomy and task variety, during and after digital transitions is necessary to maintain and promote employee well-being and motivation.

Acknowledgments

This research was funded by a grant from A + O fonds Rijk, an independent Dutch Foundation for innovation and research within the Dutch Government.