Workarounds as generative mechanisms for bottom‐up process innovation—Insights from a multiple case study

Workarounds are goal‐driven deviations from the standard operating procedures performed to overcome obstacles constraining day‐to‐day work. Despite starting as temporary fixes, they can become established across an organisation and trigger the innovation of processes and IT artefacts that can resolve misfits permanently. Although prior research has elicited antecedents and types of workarounds, it is not known how workarounds diffuse in an organisation and, thereby, innovating co‐workers' activities, IT artefacts, and organisational structures. The results of our multiple two‐year case study provide unique empirical insights into the diffusion of workarounds and how they can act as generative mechanisms for bottom‐up process innovation.

managed as a way of improving processes (Alter, 2014;Laumer et al., 2017). Nevertheless, workarounds have also been much discussed in the context of technological change (Röder et al., 2016), where they are seen as a natural consequence of digitalization, with "new controls inspir [ing] new workarounds" (Pentland et al., 2020, p. 20).
Workarounds can have both negative and positive consequences in an organisation. On the negative side, they can diminish an organisation's ability to control and perform its operations effectively and efficiently, leading to a loss of control and compliance, inferior process quality, loss of revenue, fraud, or penalties (Alter, 2014). On the positive side, workarounds can be seen as a source of improvement, creativity, or innovation (Petrides et al., 2004;Röder et al., 2014), which organisations can use to adapt or innovate their operations (Davison et al., 2019). Indeed, even if such positive deviances in business processes actually constitute an improvement (Beerepoot & Van de Weerd, 2018;Röder et al., 2016), most organisations lack a structured approach to innovate their day-to-day processes through workarounds (Weinzierl et al., 2021). Related research that has explored process drift (i.e., processes change incrementally over time) has focused on endogenous change, that is, change arising within a process based on the variation and reinforcement of actions . However, we prefer to refer this phenomenon as bottom-up process innovation not only because it highlights participants' active role in innovating a process by adapting activities and IT artefacts (March & Smith, 1995), but also because it is complementary to a process being designed from the top down-which is generally preferred in the business process management (BPM) literature Melão & Pidd, 2000).
Although workarounds are commonplace in organisations (Li et al., 2017), their diffusion and impact on organisational structures and IT artefacts are still under-investigated and under-theorised (Davison et al., 2019;Malaurent & Karanasios, 2019;Zamani & Pouloudi, 2021). Previous research stopped at identifying antecedents and types of workarounds, and elicited how and why employees work around perceived constraints on an individual level (Wolf & Beverungen, 2019). Hence, we know little about the diffusion of workarounds in organisations as socio-technical systems (Tucker et al., 2014) and their power to innovate co-workers' activities, IT artefacts, and organisational structures. Seminal work on communities of practice (Wenger & Snyder, 2009), organisational learning (Nonaka & Takeuchi, 2011), socio-materiality (Cecez-Kecmanovic et al., 2014), and the duality of structure and agency in socio-technical systems outlined by structuration theory (Giddens, 1984) strongly suggest the existence of a multilateral relationship between co-workers, IT artefacts, and organisational structures, which can be impacted by the performance of workarounds. Since the elements in an organisation conceptualised as a sociotechnical system are interconnected (Latour, 1984), we argue that workarounds performed to the individual level can cause complex patterns of dynamic change on an organisational level. Hence, to uncover these effects, the diffusion of workarounds must be understood from two supplementary perspectives, that is, from an individual and from a collective level.
While recent work has explored how workarounds can follow users' negative disconfirmation in volitional settings (Zamani & Pouloudi, 2021), current unilateral views on workarounds from either an individual or organisational perspective do not allow to explore the role of workarounds as causal structures that bring about bottom-up process innovation in an organisation. Causal structures that generate empirical events have been termed generative mechanisms (Bygstad et al., 2016;Henfridsson & Bygstad, 2013). We posit that on an individual level, an organisational routines' perspective allows examining micro-dynamic changes in the day-to-day work that employees exert, thereby indirectly innovating co-workers' activities. Organisational routines consist of ostensive and performative aspects that are mutually constitutive (Becker et al., 2003). Ostensive aspects refer to abstract, cognitive regularities and expectations that enable participants to guide, account for, and refer to the specific performance of a routine (Becker & Zirpoli, 2008;Feldman & Pentland, 2003). Performative aspects describe the actual enactments of a routine by specific people at a specific time and under specific conditions (Becker & Zirpoli, 2008;Feldman & Pentland, 2003). Because both ostensive and performative aspects enable and constrain each other, organisational routines can be employed to observe the dynamic actions of "(re)production, over time and space, through the ongoing effort of actants (people and things)" (Feldman et al., 2016, p. 505). We posit that on an organisational level, a business process perspective provides an appropriate level of abstraction to understand how workarounds impact day-to-day work for several actors in socio-technical systems. A process is a "collection of inter-related events, activities and decision points that involve a number of actors and objects, and that collectively lead to an outcome that is of value to at least one customer" (Dumas et al., 2018, p. 5). As most organisations represent socio-technical systems, in which humans employ IT artefacts to perform their day-to-day work, we posit that we need to integrate both strands of research to investigate the diffusion and consequences of workarounds in organisations more thoroughly. Consequently, we phrase our research question as follows: "How do workarounds diffuse in an organization, acting as generative mechanisms for bottom-up process innovation?" We conducted a multiple case study over a period of two years which enabled us to obtain rich insights into the diffusion of workarounds and explore their role as a generative mechanism for bottom-up process innovation in organisations. Investigating nine workarounds from three cases, we find that if workarounds are observed or communicated, they may have complex consequences beyond their individual initiators, innovating the enactment of routines by co-workers, the form and function of IT artefacts, and organisational structures. We provide empirical evidence of how workarounds act as generative mechanisms for bottom-up process innovation, which occasionally takes place without being noticed or steered at a managerial level-an effect sometimes referred to as drift . Simultaneously, we observe how different types of misfits and their individual perceptions impact the occurrence of workarounds.
The remainder of this paper is organized as follows. In Section 2, we review related research on workarounds and present business processes and organisational routines as complementary theoretical lenses for investigating the diffusion of workarounds in organisations. In Section 3, we describe and justify our multiple case study approach. In Section 4, we analyse nine workarounds identified in three case organisations and present six theoretical propositions. The discussion in Section 5 elicits the diffusion of workarounds in organisations from a theoretical and managerial perspective.
Section 6 concludes the paper, discusses its limitations, and provides a brief outlook on future research.

| Workarounds as individual performances in organisations
Organisations are socio-technical systems, consisting of two mutually dependent sub-systems, that is, a social system and a technical system (Bostrom & Heinen, 1977). The technical system refers to the technology and to processes that are used to support human needs and achieve organisational goals (Fischer & Herrmann, 2011), while the social system is concerned with dynamic organisational relationships and actors' capabilities, using IT artefacts in a social context (Sarker et al., 2019). Social structures result from continuous transformation with emergent behavioural patterns (Fischer & Herrmann, 2011), that is, actors using technology to perform tasks.
Technology innovation cycles accelerate at increasing speed (Fichman et al., 2014), creating challenges for employees at all organisational levels. Organisations often struggle to keep up with the speed of changes in their environment (Gimpel & Röglinger, 2015). In an unstable and complex environment, organisations need to adapt more often (Lee et al., 2011), which, on the one hand, constitutes an opportunity for organisations to grow their business, while, on the other, adaptations entail high uncertainty and risk (Lee et al., 2011). In particular, standardised business processes are not easily adaptable to new, especially disruptive technology, or they may become outdated when new ways of working emerge (Breu et al., 2002). This can create misfits between technology and the organisational structures, which is a common cause for the occurrence of workarounds (Alraddadi et al., 2018). Building on tasktechnology fit theory (Goodhue & Thompson, 1995), Strong and Volkoff (2010) describe misfits as a lack of fit between organisations as collectives and IT artefacts as complex software tools. Extending this view, we posit that misfits can also occur as a lack of fit between individuals' performances of tasks and IT artefacts.
In socio-technical systems, misfits can emerge from either technical or social sub-systems. More specifically, in the context of an enterprise system, six domains of misfits-functionality, data, usability, role, control, and (Koopman & Hoffman, 2003). However, since workarounds are embedded in a social system, they are subject to the system's social rules and norms, making them more than a mere improvisation by individual actors in response to a misfit (Malaurent & Karanasios, 2019). In line with structuration theory (Giddens, 1984), we posit that workarounds-like all human actions-can also manifest as organisational structures. Previous research, however, is inconclusive about how this manifestation takes place. On the one hand, workarounds are mostly treated as emergent phenomena and as a black box, and seen as outcomes rather than as dynamic change processes (Alraddadi et al., 2018). On the other hand, workarounds have also been called a source of emergent "change when they illuminate a path toward greater efficiency or effectiveness" (Alter, 2014(Alter, , p. 1049. Recent research proposed that workarounds can become embedded as the new "business-as-usual" (Crick & Chew, 2020), affecting co-workers' subsequent activities (Wolf & Beverungen, 2019), triggering the adaptation of business processes and technology (Röder et al., 2015), and, over time, becoming institutionalised in an organisation's structure (Alraddadi et al., 2018). However, we currently lack an exploration of these effects and the conditions of their manifestation in organisations.

| Business processes as a theoretical lens for studying workarounds in organisations
Business processes represent the structure of the operations in an organisation, providing a control flow of activities that guide employees (Lindsay et al., 2003) on how to perform their part (Becker et al., 2003). Processes are defined end-to-end (Dumas et al., 2018) to span all the business functions in an organisation, re-integrating knowledge that has been disaggregated because of a division of labour that manifests in a "segmentation of the institutional order" (Berger & Luckmann, 1966, p.82). They are often engineered purposefully to prescribe a standard operating procedure that puts the effectiveness and efficiency of the process ahead of any one particular activity (Dumas et al., 2018). Thus, on an organisational level, a business process can be assumed to describe the most appropriate way of conducting work activities (Alter, 2015).
Organisations are socio-technical systems, in which the design and management of business processes often follow the pursuit of increasing the efficiency and effectiveness of an organisation's value creation (Dumas et al., 2018).
Although IT artefacts (March & Smith, 1995) provide valuable means to this end, they can never be conceptually complete since the number of requirements that an IT artefact would have to implement is infinite (Alexander, 1964). Thus, even mass transaction processes are subject to the competencies of employees to perform their specific tasks , including behavioural changes performed along the way (Boudreau et al., 2016;Schmitz et al., 2016). To enable employees to perform a business process as originally defined, organisations can provide formal training by appointing experts who explain how a process should be conducted or how an IT artefact should be used, whereas informal training relies on observing co-workers performing their day-to-day work (Dunford & Perrigino, 2018).
Dependencies among the activities constituting a process exhibit three coordination patterns: flow, share, and fit (Malone et al., 1999). Flow describes how the output of an employee's activity becomes the input of another. Share refers to multiple actors accessing the same resources when performing an activity, whereas fit refers to multiple actors producing the same output (Malone et al., 1999). In contrast to mass transaction processes, which often rely on flow as a coordination pattern, other types of processes, including knowledge-intensive processes, may not exhibit the same level of coordination and structure (Hall & Johnson, 2009) and cannot or should not be modelled beforehand (Hall & Johnson, 2009;Seidel et al., 2010). The requirements of how to handle these processes "can be viewed as […] result[ing] from the dynamism of the organisational and [the] business environment (Valença et al., 2013, p. 14). If anticipated early, process variability (Frei et al., 1999) can be addressed at the design time by implementing different control flows into process models to equip process participants with ready-to-use variants to deal with specific situations and customer needs (La Rosa et al., 2017).
Beyond the variability built into process models, process participants enact their work within socio-technical systems that are dynamic, constantly changing, and self-organised (Arrow & McGrath, 1995). Consequently, to realise specific business objectives, participants aiming to fulfil their day-to-day work often end up improving processes (Becker et al., 2009) by implementing "workarounds [which constitute] knowledge about the information systems but in the context of work needs" (Safadi & Faraj, 2010, p. 4). Workarounds, in turn, are an inherent part of business processes (Röder et al., 2015), as they help process participants to solve misfits (Maurer et al., 2012). In this regard, workarounds can be considered "non-standard processes in organizations to accomplish work-based tasks" (Wibisono et al., 2020, p. 1).
Although workarounds are often the first-order solution to problems (Röder et al., 2015), the diffusion and consequences of workarounds in organisations can be described as a secondary design process in which "functions and content emerge during interaction, modification, and embodiment of the system in use" (Germonprez et al., 2011, p. 662). Organisations that become aware of workarounds and analyse how process participants work around pre-designed process models can integrate this knowledge into the next iteration of a process' re-design (Boudreau et al., 2016;Cresswell et al., 2017).
However, much of the literature on business processes still views deviations from predefined process models as undesirable and non-compliant behaviour (Hadasch et al., 2016), despite acknowledging that deviations are often needed to fulfil practical requirements that were not specified in the original process model (Brander et al., 2011).
Recent calls for action have claimed that a broader view on business processes "should consider actual work practices and actual performance results and not just process models" (Alter & Recker, 2017, p. 66), acknowledging that workarounds not only exist but also diffuse in organisations where they function as a "stimulus for change and not only [to] capture the reactive part of flexibility" (Valença et al., 2013, p.28), resulting in process drift . Hence, exploring how workarounds diffuse in an organisation seems a logical next step towards further understanding drift and bottom-up process innovation in organisations.
Building on previous attempts to provide a conceptual link between ostensive aspects, performative aspects, and IT artefacts by Pentland and Feldman (2008) and Beverungen (2014) (cf. Figure 1), we expect workarounds to be generative mechanisms for bottom-up process innovation that influences the relations between ostensive aspects, performative aspects, and IT artefacts.
Organisational routines account for the flexibility required by employees to perform workarounds (Crick & Chew, 2020). Employees can either follow the ostensive aspects of a routine by focusing on what they should do (Alter, 2014), or choose to "do otherwise" (Giddens, 1984, p. 14), including to perform a temporary workaround that might become persistent over time (Alter, 2014;Feldman & Pentland, 2003). Employees evaluate the perceived ease of use and usefulness of an organisational routine, increasing their cognitive load to execute a specific task, which may lead to either resistant behaviour (Ferneley & Sobreperez, 2006;Laumer et al., 2016) or serve as an inspiration to perform a workaround (Choudrie & Zamani, 2016). If a workaround is performed frequently enough, it can gradually manifest as a new ostensive aspect of routine (Alter, 2014;Rerup & Feldman, 2011). Hence, by enacting a workaround as part of the performative aspects of a routine, employees-willingly or unwillingly-alter the potential repertoire of activities that create and recreate the ostensive aspects of the routine (Feldman & Pentland, 2003).
Employees may experiment with various options to work around a perceived misfit (Ferneley & Sobreperez, 2006;Safadi & Faraj, 2010), for example, by consulting an expert, looking online, or simply following a trial-and-error approach (Alojairi, 2017;Zamani et al., 2020). If employees experience with a specific system or business process, they may soon discover new ways of performing their day-to-day work more efficiently (Safadi & Faraj, 2010). Individuals do not always perform workarounds on their own, but are influenced by organisational norms, IT artefacts, and the activities of coworkers, and workarounds can also be socially constructed and thus be based on mutual consent (Dunford & Perrigino, 2018). Socially constructed workarounds can be explained by common stressors, such as job demands or environmental stressors, to which multiple employees are exposed (Dunford & Perrigino, 2018). Hence, stressors can lead to a higher degree of innovation, such as novel ways of using an IT artefact to perform work in new ways (Maier, Laumer, Tarafdar, et al., 2021). Once a workaround is established in an employee's routine, it might become accepted by coworkers as an agreed-upon work practice (Dunford & Perrigino, 2018).
Although the literature on IT adoption states that employees interpret, appropriate, and use IT artefacts in different ways (Malaurent & Karanasios, 2019;Zamani et al., 2020), the extant research has remained silent on how workarounds in organisational routines can initiate an IT artefact's adaptation to re-establish task-technology fit. The first result indicates that the regular enactment of workarounds might have the power to transform co-workers' activities (Wolf & Beverungen, 2019), the form and function of IT artefacts (Röder et al., 2015), and organisational structures (Alraddadi et al., 2018;Azad & King, 2012). In this case, workarounds may diffuse through communication (Dittrich et al., 2016) and observation (Safadi & Faraj, 2010). Viewing organisations as socio-technical systems allows the detection of a recursive relationship between information systems and the routines performed by social actors (Beerepoot & Van de Weerd, 2018). Although organisations try to re-establish task-technology fit through a gradual process in which the system and the organisational context are "mutually adapted" (Maurer et al., 2012;Woltjer, 2017), workarounds might also yield a cascading effect that induces other workarounds (Alraddadi et al., 2018) or misfits (Maurer et al., 2012;Woltjer, 2017). Despite the first indications of these effects, an in-depth examination of workarounds as generative mechanisms for bottom-up process innovation that includes the adaptation of IT artefacts and organisational structures on a more permanent basis are still missing.

| A critical realist case study on workarounds
The phenomenon of workarounds is complex and inextricably linked to the socio-technical context in which they occur (Malaurent & Karanasios, 2019 Feldman, 2008 andBeverungen, 2014).
separated from their real-world context and researchers have little or no control over them, case studies are a suitable method for researching operational links guided by how and why questions (Yin, 2017). Conducting case study research allows us to explore how workarounds diffuse in an organisation, innovate co-workers' activities, IT artefacts, and organisational structures.
Critical realism emphasises the existence of enduring natural, social, and conceptual entities that, regardless of their effectiveness, have the potential to constrain and enable the occurrence of empirical events. The generative mechanisms that force specific events to occur exist independently of the observer and can only be understood by analysing all three domains of critical realism: the real, the actual, and the empirical (Mingers, 2004). The real encompasses all mechanisms that can cause events, while the actual comprises all possible events that might be instantiated by the underlying mechanisms, and the empirical is a small subset of the actual world that we can observe and experience (Mingers et al., 2013). Consequently, critical realism enables us to observe change (Volkoff & Strong, 2013) that is induced by workarounds as a generative mechanism for bottom-up process innovation in an organisation.
In our research approach, workarounds play a dialectical role. Similar to a related study by Zamani and Pouloudi (2021), who used critical realism to identify the generative mechanisms that give rise to different user practicesincluding workarounds-following negative disconfirmation with consumerised IT artefacts, our starting point is to view workarounds as observable events in organisations (empirical domain). However, at the same time, we treat workarounds as generative mechanisms (real domain) that produce events (actual domain) for us to observe and experience in an organisation (empirical domain). We refer to these empirical events as the consequences of workarounds.
This dialectic consideration of workarounds as both cause and effect in an organisation is in line with the conceptualization of organisational routines, which are also viewed as generative systems (Pentland & Feldman, 2008).
Building on structuration theory (Giddens, 1984), the routines literature states that the performances of routinesincluding workarounds-are enabled and constrained by social structure as ostensive aspects, whereas performances also constitute and re-constitute the ostensive aspects, forming a mutually constitutive duality.
Drawing on the observable event (i.e., workarounds), we can build transcendental arguments about causal mechanisms (Mingers et al., 2013), and explore the diffusion of workarounds triggered by misfits and their impact on different elements in socio-technical systems. As competing explanations might exist, it is important to check their selfconsistency (Mingers, 2001) and assess them against other explanations (Bygstad et al., 2016). In our multiple case study, we collected data over a period of 2 years, to trace employees' experiences in their day-to-day work in three organisations over time. Thus, we were able to understand the observed events and relate them to their underlying mechanisms (Wynn & Williams, 2012). The three cases are summarised in Table 1. The business processes we analysed differed in terms of scope, governance, level of IT support, maturity, variability, interdependence, degree of creativity, degree of formalisation, frequency, and stability (c.f. Table A2 in the Appendix). We purposefully selected these cases based on a theoretical replication logic, assuming that the diversity of the processes and their contextual embedding will potentially lead to the identification of different workarounds and their consequences, "for anticipatable reasons" (Yin, 2017, p. 57). Thus, we achieved a high robustness of our results (Yin, 2011). In each case, we investigated how employees established workarounds on an individual level, and how they subsequently innovated co-workers activities, IT artefacts, and organisational structures.

| Data collection and analysis
Following Yin (2017), we identified and analysed multiple sources of evidence for each of the three cases. Primary sources of evidence included organisational documents, direct observations, and semi-structured interviews. Documents included process models, work descriptions, standard operating procedures, emails, spreadsheets, and lists that employees used in their day-to-day work. Interviews were conducted with several informants who were involved in the same business process. This approach was intended to reveal the different perspectives that stakeholders adopt in a process and to substantiate the significance of the collected data. Saturation was achieved when we no new information was obtained regarding the phenomenon under investigation. We conducted 27 semistructured interviews, each lasting between 30 and 77 minutes, in addition to numerous informal conversations. The interviews were audio-recorded and transcribed to extract rich insights from all the notes, audio recordings, and documents. Different types of evidence were carefully triangulated (Yin, 2017) to identify the consequences of workarounds more accurately from multiple perspectives.
The processes we investigated were purposefully selected to study workarounds in three different settings and involving different types of IT artefacts. From the initial interviews, we identified that these processes had drifted without top-down initiatives on the re-design of the processest involved. The object of the investigation in Case 1 is the hire-to-retire process performed by the IT division of a large media company. The process was recently changed by introducing a new workflow engine designed to digitalize the process and make it more efficient. In Case 2, an auditing process was analysed, which is performed by one of the big four international auditing companies. The process is performed for external clients and also bridges the company's front-stage (consultants working at the clients' facilities) and back-stage (shared service centre). This process was being established at the time of our investigation and was still in considerable flux. The object of investigation in Case 3 is the process of resource planning and allocation of courses in a public university. However, this process had remained unchanged for a couple of years. The process bridges several organisational units in the university, but while there is no single IT artefact to support it, various collaboration tools are used by the process participants.
For each case, we conducted qualitative content analyses of the retrieved data (Mayring, 2004). Content analysis of existing documentation helped us to gain meaningful insights into the workarounds of organisational routines established by individuals. For this purpose, electronic coding in a team was performed with the support of MAXQDA as a computer-assisted qualitative data analysis software.
Following state-of-the-art advice on analysing qualitative data (Mayring, 2004), we organised data analysis in three basic steps, consisting of data reduction, explication, and structuring. For data reduction, we first paraphrased T A B L E 1 Introduction and replication of the three cases. the statements from the interviews by cutting text components that are not content bearing. We then generalised the statements to an appropriate level of abstraction by extracting the sentence kernels and re-testing them to ensure that they are representative (Mayring, 2004). Subsequently, we summarised the data to form higher-level statements, based on theoretical insights about the occurrence and diffusion of workarounds outlined in Section 3.3 (see Table 2), the data were coded by abstraction and categorization (Mayring, 2004). We began by performing an initial coding step to identify workarounds in one case, along with their properties and diffusion mechanisms. The first phase of our data analysis led us to make small adjustments to the coding scheme. We then analysed the data obtained from the first case using the revised coding scheme, as recommended by Dubé and Paré (2003). Thus, we carefully aligned the data and theoretical constructs, a common approach to ensure high internal validity (Hartley, 2004). After performing a within-case analysis for each case, the resulting insights were compared across cases.
After outlining and analysing the individual cases in detail, we performed a cross-case analysis that followed the guidelines prescribed by Miles and Huberman (1994), which allowed us to identify recurrent patterns emerging from our data. First, we inspected cases and partitioned the data into smaller categories. We then clustered the data to identify broader categories. We applied diverse tactics to identify insights rooted in our case data, including drawing comparisons, noting patterns, and relationships between constructs (Miles & Huberman, 1994). By cross-comparing data from the cases, we identified commonalities and differences regarding their organisational setting, the occurrence of workarounds, and their consequences, which we aggregated in a conceptually-ordered matrix (Miles & Huberman, 1994) (cf. Table 4). Combining within-and cross-case analysis allows solving the paradox of recognising the uniqueness of each individual case and gaining more general insights across the cases (Miles & Huberman, 1994). Analysing and comparing data on a particular phenomenon from multiple cases not only minimises the risk that "findings are not merely the result of idiosyncrasies of the research setting" (Cavaye, 1996, p. 237), but also allows to reconcile the uniqueness of an individual case with the need to derive a more general understanding of generic processes that occur across cases (Miles & Huberman, 1994;Silverstein, 1988).

| Theoretical preconception for coding
Inspired by Yin (2011), who postulates that "for case studies, some theory development as part of the design phase is highly desired" (Yin, 2011, p. 37), we developed a coding scheme to guide our qualitative data analysis (Dubé & Paré, 2003). Our coding scheme includes concepts related to business processes and organisational routines as well as prior conceptualizations of misfits developed in an IS context. It comprises three analytical dimensions (Laumer et al., 2017): the settings (I), occurrence (II), and consequences (III) of workarounds. These dimensions constitute the conceptual frame of the coding table (cf. Table 2) and comprise nine categories. Departing from related research, we made small adjustments to the coding schema during data analysis to increase its conceptual clarity and precision (Hartley, 2004).
The three dimensions constituting our coding schema are as follows (see Table 2): Dimension 1: Setting. To examine the setting in which workarounds occur, we analyse whether employees received formal training, informal training, or no training at all on how to perform their tasks and how to use IT artefacts to conduct their work activities (1) (Dunford & Perrigino, 2018). Describing the degree of formalisation of training allows us to identify if the training facilitates employees' familiarisation with the intended structures and processes in an organisation or whether they are expected to draw their own subjective conclusions about the structures and processes that govern their work. Furthermore, core aspects of organisational routines (i.e., ostensive and performative aspects) and their business process equivalents (i.e., process type and process instances) are included in coding p. 17). "The ostensive aspect is the ideal or schematic form of a routine" (Feldman & Pentland, 2003, 101). "A process type represents a particular business process" (Rinderle et al., 2004, p. 274). "Flow dependencies arise whenever one activity produces a resource that is used by another activity" (Malone et al., 1999, p. 429).
"Workarounds emerge either from obstacles to getting work done or from goal misalignment of stakeholders" (Röder et al., 2014, p. 3 "Generativity arises in the small, from the bottom up" (Mendling et al., 2020, p. 210).
"Sharing dependencies occur whenever multiple activities all use the same resource" (Malone et al., 1999, p. 21).
"Workarounds occur because technology that is used does not fit realities and contingencies of day-to-day work" (Alter, 2014(Alter, , p. 1049 Rinderle et al., 2004). It is important to conceptually pin down the exact source of an emerging change to distinguish workarounds from related yet distinct phenomena (e.g., a process was changed on the type level). In addition, we added a construct that specifies the type of IT artefact used by employees to perform their work. This distinction is of particular importance because different types of IT artefacts (e.g., instantiations, such as enterprise systems or collaboration tools) can lead to different misfits that result in different workarounds. As indicated by Wolf and Beverungen (2019), the type of collaboration among co-workers can impact the diffusion of workarounds. In line with coordination theory, three types of dependencies can be conceptualised, including flow, share, and fit (3) (Malone et al., 1999).
Dimension 2: Workaround occurrence. The consolidation of different research streams on how workarounds occur (Alojairi, 2017;Alter, 2014;Ignatiadis & Nandhakumar, 2009;Laumer et al., 2017;Pinto et al., 2018;Röder et al., 2014) can be summarised by identifying three categories of misfits that may cause workarounds (4). A Category A misfit points to inconsistencies between the ostensive and performative aspects of a routine, that is, the extent to which expectations on an organisational level about how a task should be performed deviate from the interpretation and enactment of that task by a single actor (Zamani & Pouloudi, 2021). A Category B misfit refers to artefacts that are not appropriately aligned with the ostensive aspects of a routine, whereas a Category C misfit refers to a lack of alignment between IT artefacts and the tasks performed by employees.
In addition to identifying three categories of misfit, we added a label that identifies whether a workaround is socially constructed (Dunford & Perrigino, 2018) or individually constructed (Azad & King, 2012;Zamani et al., 2020) to explore the constitutive mechanisms that trigger workarounds (5). Regarding the type of workaround (6), we distinguish between artefact-centred workarounds-where either an alternative IT artefact is used, or an IT artefact is used differently to its intended design-and process-centred workarounds, performed by employees to bypass certain activities of a business process (Choudrie & Zamani, 2016;de Vargas Pinto et al., 2018;Outmazgin & Soffer, 2016;Weinzierl et al., 2020).
Dimension 3: Consequences. Third, we introduced three concepts to elucidate the consequences of workarounds. First, we specify an organisation's response (7), distinguishing bwtween process adaptation, IT artefact adaptation, the prevention of workarounds, and ignorance of workarounds (Beerepoot & Van de Weerd, 2018;Choudrie & Zamani, 2016;Dunford & Perrigino, 2018;Mendling et al., 2020). In addition, we encompass communication and observation as categories to analyse the diffusion mechanisms of workarounds (8) (Dittrich et al., 2016;Safadi & Faraj, 2010). Finally, in line with socio-technical theory and the findings by Barrett (2018), we added co-workers' activities, IT artefacts, and organisational structure as three concepts that can be affected by a workaround over time (9)

| Case description and within-case analysis
An analysis of the three cases in our dataset found nine workarounds (see Table 3). For each workaround, we performed an in-depth analysis to provide a thick description (Yin, 2017) of the setting, its occurrence, and its consequences for co-workers' activities, IT artefacts, and organisational structures, outlining the peculiarities of each of the three cases.
Case 1. The first case represents a multi-national media, services, and education company. The IT department implemented the process under analysis to support human resources (HR) performing their daily business, supplying them with a workflow implemented in a workflow management systemreferred to as starter-changer-leaver (SCL) system. The purpose of the process is to help HR better coordinate their administrative processes for the end-to-end support of employees (hire-toretire), ranging from on-boarding activities (starter), and HR-related changes (changer), to offboarding activities for employees leaving the company (leaver). The frequency of the process is irregular, triggered by employees starting, changing, or quitting their work in the organisation.
Multiple business units are involved in the process, including HR, IT, property management, and employees' respective departments. Before spring 2019, the organisation used PDF checklists, spreadsheets, emails, and phone calls to coordinate SCL activities. As part of a wider improvement project-for which we were able to access the project documentation-the old PDF checklists were replaced with a proprietary software tool-specifically, a workflow engine-to enact a prescribed workflow for SCL activities. To comply with the European General Data Protection Regulation (GDPR) on protecting personnel data, the system grants very specific access rights and graphical user interfaces (GUIs) for each of the roles involved in the process. This means that an employee can only see the information and tasks that are relevant to their own job role and responsibilities. After implementing the SCL system in the organisation, all employees received an official introductory presentation by IV1-the process owner responsible for implementing the SCL system. The HR department instantiates the process by looking up a staff ID number in the Enterprise Resource Planning (ERP) system and subsequently enters this number into the SCL system. This triggers a workflow that can take different paths through the organisation, depending upon the respective employee's start date, any changes to their T A B L E 3 Overview of the identified workarounds. Since ostensive and performative aspects of cut-off testing were not aligned, employees perform a manual consistency check in advance and convert documents to PDF format.
As SharePoint is perceived as not user-friendly for course scheduling, employees create their own overview in Excel.
Workaround 2 Due to the ill-defined definition of a "changer", users are not permanently deleted but de-activated to allow a later re-activation.
Not all languages can be detected by the software, requiring a manual document matching process.
Since the room planning tool is convoluted and allows double-booking of rooms, employees use a physical room management board.
Workaround 3 Since the process is not designed to capture all types of employees, HR uses the former PDF list as fall-back option.
Erroneous reports cannot be modified in the software, requiring a subsequent report manipulation in Excel.
Due to a lack of conformance between the designed process and day-to-day work necessities, a request for room change is performed via email. employment, or whether they are about to leave the company. In any subsequent activity, the employee responsible for performing a task is informed about the required administrative tasks via an automated email notification. On completing these tasks, an employee marks them as 'done'. Then, the next employee receives a notification about their task, and so on, until all the activities in the process have been completed (a detailed process model is supplied in the appendix, Figure B1: Process Model of starter-changer-leaver).
We identified three substantial workarounds in the SCL process. The first was introduced by HR to overcome a Category C misfit regarding a missing data field to specify the type of change requested (cf. Table B1 in the Appendix). For example, employees in HR can start a new instantiation of the SCL process to update the personal data of an employee, for example, a name change caused by marriage or an address change. However, the implemented workflow system was unable to distinguish between different types of change request.
Thus, all the activities of the change workflow had to be executed, even though some were redundant in certain circumstances, for example, a name change workflow did not require a technical equipment check. In response to this inefficiency, IV2 from the HR department and her co-workers agreed to perform an artefactcentred workaround and started using an existing free text field to add a comment on the particular reason for initiating a change process.
I have to enter it [the type of change request] manually in a comment field but the colleagues still cannot see immediately which change is actually requested. (IV2) Over time, the implemented workaround spread via communication to co-workers who adopted it and started specifying the changer process in the free text field. Subsequently, the diffusion of the workaround in the organisation affected the IT artefact, because the central IT department adapted the SCL system to better support employees in their daily work. The type of change request can now be specified when initializing the process, resulting in efficiency gains.
We have created the option to specify the changer exactly, e.g., name change, or cost center change.
The respective processes are then significantly reduced and therefore also much clearer for the user.
This is a change that has simply matured over time.
[…] It has been brought to our attention by users and relevant HR departments. (IV9) The second workaround was performed by all employees in the central IT department to overcome a Category A misfit (cf. Table B2 in the Appendix). It became apparent that the different business units involved were interpreting the roles of changers and leavers quite differently. In line with a legal perspective, the HR department ascribed the label of 'leaver' to employees that had left a subsidiary to join another entity of the organisation. However, when the central IT department received the notification of an employee leaving the organisation, they permanently deleted their access rights and the corresponding users in their system. Only much later did they realise that the employee had not actually left the organisation, but only moved to another corporate unit, with its own legal entity.
In that instance, the central IT department had to create a new user ID for that employee, although changing the access rights would have sufficed. Hence, this workaround resulted from a lack of conformance between the ostensive and the performative aspects of a routine, and is not directly related to the IT artefact.
To overcome this Category A misfit, IV3 came up with the idea of implementing an artefact-centred workaround whereby users are only deactivate instead deleted.
A leaver could be someone who actually leaves the company or someone who moves to a different legal entity of our organization. We do have colleagues who moved between business units within the company, so that the mailbox can often be adapted and not deleted. Even though we get a leaver notification, my colleagues and I already know that "we have to be careful not to delete everything immediately," to allow to re-activate the account later. (IV3) Since the underlying misfit was critical on an organisational level, because it affected senior management representatives that were not able to continue working due to deleted users, the misfit escalated quickly by communication with the IT department. The IT artefact was affected because new process variants were implemented that better reflected the different reasons for instantiating a leaver process. Hence, from then on, the workflow only triggered the activities that were required for implementing the requested changes.
In the case of inter-company changers-where user [records] were deleted because the staff member had been treated as leaver-the next day, he or she no longer had any access rights and could not continue working. All these cases escalated-up to a senior level-and we were told to make changes to the process.
[…] We have now implemented different leaver processes, e.g., when someone goes on long-term or maternity leave for three months or perhaps half a year. Then, the user is protected so that [their record] is not automatically deactivated due to inactivity. (IV10) The third workaround in this case occurred on a regular basis due to a Category A misfit (cf. Table B3 in the Appendix). Although the SCL system was designed to administrate full-time employees, it did not cover part-time staff, such as interns and consultants. Excluding part-time staff from the process was an intentional decision embedded in the process, but it did not reflect an HR employees' day-to-day-work requirements. We trace this misfit back to a lack of conformance between the ostensive and the performative aspects of a routine. However, to be able to create logins for part-time employees, the HR department decided to initialize the process without using the SCL system at all. Correspondingly, the organisation did not create a user ID for those employees. A user ID, however, was mandatory for instantiating a process in the SCL system.
Thus, IV3 and his colleagues decided to initialize the process by using a PDF form (an artefact that was used before the SCL system was implemented), comprising a range of questions about the work-space equipment, authorization, and technological equipment (e.g., mobile phone, tablets, or laptop computers). This artefact-centred workaround was socially constructed, constituting a viable way of handling starters outside the SCL system on which they had "decided at some point over the course of time" (IV3).
Sometimes, we simply re-use the former questionnaire, or we request the information via email. (IV3) The workaround implemented here diffused via observation and revealed that proper organisational structures required to update the process and resolve the underlying misfit hab been missing. Although the company discussed how to proceed, the workaround remained a temporary fix. However, observing the workaround has led the IT department to develop a new GUI that enables creating new users ad-hoc, whereas problems regarding their specific roles and responsibilities (e.g., concerning the question of who is responsible for interns) remained unresolved, awaiting proper governance structures to be implemented first.
This is not a system issue; it is ultimately a governance issue, which the departments also discuss cur- Case 2. The second case study was conducted at a professional services network organisation with multiple subsidiaries across the globe. The process we analysed supports auditing, that is, conducting a financial examination regarding the correctness and completeness of invoices, revenue data, and financial statements for B2B customers. The organisation compares revenue data with transaction data by reviewing the corresponding documents (e.g., purchase orders, delivery notes, invoices, etc.). This process is called cut-off testing and was outsourced and off-shored from the headquarter to shared service centres in Germany and Poland a few years ago, in order to save costs. Although the auditing consultants are in direct contact with the company's customers, collecting the necessary documents (either digitally or as paper-based files that are then digitised) and doing the actual audit, the employees in the shared service centre were responsible for performing the cut-off testing process, that is, the necessary pre-processing of data that was performed in advance of the actual audit. Cut-off testing was conducted manually daily, but was often delayed or required consultants to perform additional document checks. In 2019, the organisation decided to implement a proprietary software application that is based on machine learning algorithms to support the semi-automated document checking of PDF files. We refer to this well-documented application-for which the organisation provided handbooks, manuals, and best-practice examples that we could access-as Software A. The goal of implementing Software A was to support the shared service centre with cut-off testing to realise efficiency and quality gains and, thus, to improve the accuracy and speed of the cut-off testing process (see a detailed process model in Figure C1 of the Appendix). After finalising cut-off testing, the employees in the shared service centre generated a report that was imported into a platform-called here Software B-from which the consultants retrieved the data they needed to perform the auditing process.
The first workaround was implemented by the shared service centre right at the start of the cut-off testing process to overcome a Category A misfit (cf. Table C1 in the Appendix). However, it became apparent that the shared service centre often received other document types than the PDF files they expected, for example, Word documents, screenshots, or handwritten notes that could not be processed by Software A. Consequently, any other incoming document type needed to be converted into a PDF format. In addition, an upload of revenue data required a tabular structure, as is implemented in SAP ERP. Thus, any revenue data that were exported from other ERP systems needed to be transformed into the required input format. Consequently, IV2 from the shared service centre reported how and why he performed a process-centred workaround by doing a manual consistency check, that is, checking the format of the documents and converting them before proceeding with their upload into Software A for cut-off testing. The misfit occurred because of a misalignment between the ostensive and the performative aspects. The customers and consultants simply did not know that the shared service centre expected them to send PDF files.
What they actually do is check the document delivery, control for Word documents, and then, transform the formats that the tool cannot process. That's actually a workaround. (IV2) The workaround performed by shared service centre employees diffused via communication and led the organisation to initiate user training which affects the organisational structure. Customers are now trained to only hand in documents in a PDF format, whereas shared service centre employees are trained to detect and transform document types that the tool cannot process. In addition, the tool will be further developed to adapt the IT artefact, but this is not a priority, because much time is saved already by using the tool with its current functionality.
We have implemented user training-I hope that helps for now. […] We are working on the technical implementation, but other things currently take a higher priority […] a 85% recognition rate is already achieved and [this is] a huge step that saves us a lot of time. (IV7) The second workaround was caused by a Category C misfit, with Software A unable to detect distinct font types (like handwritten notes) or foreign alphabets (like Cyrillic letters) (cf. Table C2 in the Appendix), which meant that the semi-automated document-checking and matching process of the uploaded documents with the revenue data did not work properly, solely displaying the matching rate (ranging from 0%-100%). In response to this misfit, the employees in the shared service centre implemented an artefact-centred workaround by checking all documents manually and identifying those that remained unmatched.
The engagement teams need to add some additional information. With any [document] hat is not extracted by Software A, we still need to go back to the original documentation, look up this information, and enter it manually. (IV1) The workaround spread via communication, and affected the IT artefact. On the one hand, the tool was updated to automatically recognise and process additional languages. On the other, a mouseover function was added to help employees with adding missing data more efficiently. However, an enhancement of the tool's functionality is planned, and the adaptation of the IT artefacts is still pending.
We have improved the performance of the extraction, added more languages, and implemented more functionalities. Now, unrecognized data can be selected by mouseover and the OCR recognition process, which then transfers them into the data field.
[…] Bilingual documents cannot be processed automatically so far, because there is no tool for it available on the market, yet. (IV7) Since Software A is based on machine learning algorithms and was still trained in day-to-day work, the matching rate stagnated at 80%--90%. As employees in the shared service centre detected incorrectly matched documents and other errors, they had no option but to correct these data directly in Software A, which indicates a Category B misfit. According to IV2, in order to provide correct results for the consultants, the employees of the shared service centre manipulated the spreadsheet that was generated at the end of the cut-off-testing process. This meant they performed an artefact-centred workaround by typing in missing data, inserting columns to display additional data, and adding comments about incidents in the cut-off-testing process (cf. Table C3 in the   Appendix).
Afterwards, you'll get a changeable Excel report. If software A has extracted and matched 80%, which would be great, you'll have to do the remaining 20% manually by typing it in the report. (IV3) The misfit and the implemented workaround were observed on different levels in the organisation. However, a solution to overcome the workaround had not yet been implemented. Thus, the workaround is currently accepted as a temporary fix to compensate for the Category B misfit. Meanwhile, improving reporting tools is on the organisation's agenda as part of an overarching project to review all reporting systems that will also help to improve reporting with software A.
It is better now, but it is still not really cool. That's also the feedback we got from our leadership team.
Up until extraction, it looks cool, but then it grates [sic]. That's really the only way to describe it. […] We are working specifically on this as part of an independent documentation tool that Software A will also be able to use. (IV7) Case 3. The third case study was performed at a public institution for higher education, where, twice a year, employees had to perform activities as part of planning the teaching schedules across different professorial chairs. As each chair-each representing an independent unit within a faculty's department-implemented decentralised organisational structures, managed their own budgets, and did not directly interrelate with each other, the course scheduling process was, in effect, an interorganisational process. This conceptualization follows the notion that processes are deemed interorganisational if "each department has its own organizational structure" (Debois et al., 2018, p. 401).
The teaching planning process was established gradually over a period of 6 years, after the CEO's decision to introduce Microsoft SharePoint as a collaboration tool, with no clear guidance on how to use it in the course scheduling process. Twice a year, SharePoint was primarily used by the faculty's central administrative office to collect data about the upcoming courses offered by the professorial chairs. The process was initiated by the institution's central administrative office, which also set deadlines for the planning process. These deadlines were then scheduled by the business faculty's office and forwarded to the professorial chairs via email, informing them about the deadlines they needed to comply with when entering their course information in SharePoint. After the deadline had expired, the administrative office copied the data from SharePoint into a spreadsheet and sent it to the institution's central room management office, which assigned adequate rooms for the classes, using a paper-based timetable and a physical planning board. Once the rooms had been assigned, the room management staff entered the room numbers for each course in the spreadsheet and sent it back to the faculty's administrative office. The overall process included all three coordination patterns and was subject to substantial variations due to the independence of each business unit, that is, the professorial chairs (see a detailed process model in Appendix Figure D1: Process model of module planning). Based on triangulating the data from this case, we identified three workarounds.
The first workaround was implemented by the professorial chairs' assistants to overcome a Category C misfit (cf. Table D1 in the Appendix), which was rooted in the perception that SharePoint did not provide sufficiently detailed documentation with all the course-related information. Hence, IV1, IV2, and IV3, who were responsible for planning the courses for their respective professorial chairs, individually set up their own spreadsheets in which they entered all relevant course data (artefact-centred workaround).
For this purpose, I designed an Excel list. This list includes all semesters and I can copy information from the previous semesters to the new one. But I can also add other information like preferred dates for specific modules. (IV2) The implemented workaround diffused through communication in the organisation. After its recognition by the faculty's central administrative office, they decided to prevent the workaround. To do so, the underlying misfit was eliminated, which affected the IT artefact. First, additional fields were added to SharePoint to collect data for midterm and final exam dates simultaneously. Second, additional information about the requested data was added to the tool.
We now map this in SharePoint so that it will be possible to select both the intermediate and the final exam dates directly. This way you can bypass the multiple adding of information. Then, you can enter all the information in one place, instead of three.
[…] In addition, we have now added notes in the fields to be filled in that inform the user which information is requested. (IV10) A second workaround was implemented by the central room management office to overcome a Category B misfit (cf. Table D2 in the Appendix). Once the business faculty's administrative office had collected all the course information from the professorial chairs in SharePoint, they forwarded a list to central room management with the course planning data via email. Not only did the room and course booking tool-called Software C-allow double-booking, but it also failed to provide a clear, easily readable time table. This prompted IV4 from room management to use a manual planning board in her office, implementing an artefact-centred workaround. She perceived this manual planning process to be the more efficient and effective for the room allocation task.
At the moment we're doing this manually. It's a lot of work, but you get a good overview.
[…] We print out all the lists and then we work through them. (IV4) Even though some courses in the last semesters were solely held online because of the Covid-19 pandemic, and no room allocation was required, the manual process for room scheduling was maintained. Hence, the workaround did not affect the work performed by co-workers even though it diffused through communication in the organisation.
The processes have remained the same and will not change. (IV10) As the administrators received neither formal nor informal training about the process, misperceptions of the activities in the course scheduling process led to implementing a third workaround to overcome a Category A misfit (cf. Table D3 in the Appendix) concerning requests for a change of room allocation. This misfit originates in the fact that additional requests, such as a change of room, had not been considered in the process design and thus resulted in non-conformance between ostensive and performative aspects of a routine. To remedy this misfit, some assistants implemented a process-centred workaround by simply calling or emailing central room management to ask for specific rooms and to influence room allocations in their favour, as IV3 revealed.
I write an email anyway because it supports me with following up on the process. Further, I get a read confirmation that lets me know that they have received (and read) the message. (IV3) This workaround, which spread through communication, was still in use, at the time of writing, to track requested changes to the teaching schedule, to overcome the Category A misfit that arose from the fact that additional inquiries or interactions had not been included in the planning process, and, hence, had not been supported by the form and function of the IT artefact.
What also happens frequently are short phone calls, but depending on what is asked and how complicated it is […] in 99% of the cases, I ask for an email. That way, I can also understand afterwards what has been done, which makes my work much easier, of course. (IV10)

| Cross-case analysis
Analysing the data across cases (cf. Table 4) reveals that workarounds diffuse through an organisation in an infinite loop, until the underlying misfits are resolved. We visualise this loop with the metaphor of a Gordian knot. We visualise this loop in Figure 2, which depicts the (two-directional) diffusion pathways of workarounds that they undergo an become a generative mechanism for bottom-up process innovation, thereby, adapting IT artefacts and organisational structures, and impacting co-workers activities. Some workarounds can also induce organisations to initiate topdown driven adaptation projects (cf. Figure 2). In the following, we elaborate on the results by discussing six propositions summarised in Table 5. On their way, workarounds change performances, structures, and IT artefacts, as generative mechanisms for bottom-up process innovation. Some workarounds can also induce organisations to initiate top-down driven adaptation projects (cf. Figure 2). In the following, we elaborate on the results by discussing six propositions summarised in Table 5.
T A B L E 4 Cross-case analysis. Participants perform workarounds to resolve misfits that they perceive to profoundly constrain their day-to-day work. Misfits represent three categories, pointing to ostensive aspects, performative aspects, and IT artefacts that are misaligned. Misfits between ostensive aspects and IT artefacts remain latent and manifest indirectly as one of the other types of misfits.
Proposition 1. Workarounds can lead to restructuring ostensive aspects and adapting IT artefacts.
Our view on the immediate consequences of workarounds, rooted in the literature on organisational routines and business processes, focuses on the dynamic interplay between performative aspects, ostensive aspects, and IT artefacts, in which each can influence the other two. From our cases, we found that employees can establish socially or individually constructed workarounds in response to a misfit in two ways. First, they can change their work practices (i.e., the performative aspects of a routine) to work around a misfit (e.g., workaround 2 in Case 2, workaround 3 in Case 3) or, in the long term, change the ostensive aspect of a routine. Second, they can unfaithfully appropriate an IT artefact (DeSanctis & Poole, 1994) (e.g., workaround 1 in Case 1, workaround 3 in Case 2), and either re-design the IT artefact or design their own IT artefacts from scratch (e.g., workaround 1 in Case 3). All these strategies are contained in our data set, enabling us to explore how workarounds start to have effects in an organisation beyond the performances of their individual initiators. We conclude that the consequences of workarounds for ostensive aspects and IT artefacts provide important starting conditions for their diffusion through an organisation.
Proposition 2. Workarounds require communication or observation to diffuse in an organisation, subject to coordination patterns.
We found that workarounds spread to the routines performed by co-workers via two diffusion mechanisms. As proposed in Adaptive Structuration Theory (AST), communication (verbal or non-verbal) and observation are seen as mechanisms for instantiating structures, and can trigger behavioural change (DeSanctis & Poole, 1994). This claim is consistent with our case findings. When actors noticed how a co-worker resolved a misfit by performing a process differently or by appropriating an IT artefact unfaithfully to achieve a goal more efficiently, they adopted the workaround in their routines. This behavioural pattern occurred on an individual level-that is, one person copied another's workaround-or even on a group level-that is, multiple persons copied workarounds instantiated by a group of people. To diffuse in an organisation in this way, workarounds had to spread either through communication with co-workers (e.g., workaround 1 in Case 1, workaround 1 in Case 3), or through observation (e.g., workaround 3 in Case 1, workaround 3 in Case 2). Communication was enacted via official channels (e.g., email, memos) or unofficial channels (e.g., informal conversations at work). Observation refers to a co-worker noticing discrepancies between the expected input and output of their performances of a specific routine compared to those performed by others. In many cases, employees realised that the deviation was caused by their co-workers implementing a workaround. Congruent with the literature, as soon as the number of actors performing a workaround reaches a critical mass, others accept the workaround as a source of bottom-up process innovation and integrate it into their own performative routines (Fichman, 1992).
Furthermore, we witnessed that the type of coordination in a process impacts the diffusion of a workaround. In our cases, it mattered if the process was linear (Cases 1 and 2), organised as a network of actors working together, or supported by collaboration tools (Case 3). Workarounds spread more frequently in rather linear processes that follow predefined workflows supported by IT artefacts. Here, the actions of a predecessor in a process triggered a (re-) action by a successor and/or a change in the use of the IT artefact. This effect is also in line with AST, as the interactions between actors and technology are subject to a dialectic of control, in which both (actors and technology) are mutually shaping each other (Giddens, 1984). In contrast, Case 3 exhibits a situation where workarounds spread less frequently because actors exchanged information only casually, without using fixed workflows and regulations as means of coordination. In Case 3, actors were not aware of all the other process stakeholders, nor were they aware of all the activities that contributed to the course scheduling planning process. Only actors who were in direct contact with a person who worked around an IT artefact or process were able to recognise a workaround. Because all stakeholders used their own IT artefact (i.e., the chair administrator used spreadsheets, the administrative office used SharePoint, and room management used a physical planning board), workarounds were enacted but did not diffuse, because of the organisational structure inhibiting communication and observation between departments. Instead, a process participant suspecting that a workaround was performed by a predecessor either rejected it or corrected the input for her own routine, since every group of actors extracted only the data relevant to them, using their idiosyncratic IT artefacts.
Proposition 3. On an organisational level, workarounds can serve as means for re-balancing the triangular relationships between IT artefacts, organisational structures, and performances of routines.
Business processes and IT artefacts are supposed to be aligned to achieve a high task-technology fit (Goodhue & Thompson, 1995). However, in none of our cases was this alignment fully achieved. Our data illustrate how employees perform workarounds to remedy misfits that constrain their day-to-day work. On an organisational level, workarounds served to re-balance the triangular relationships between IT artefacts, organisational structures, and the performances of routines. If a workaround was performed frequently enough, new routines occurred, combining elements of the old routines with innovative elements that originated from workarounds (e.g., workaround 3 in Case 3). In addition, workarounds impacted on the form and function of IT artefacts, leaving traces that coworkers could observe easily and frequently as consequences of the workaround (e.g., workaround 3 in Case 2). As advocated in punctuated equilibrium theory (Eldredge & Gould, 1972), organisations are based on a deep structure comprising multiple elements and activity patterns, which keep a social system stable (Gersick, 1991). However, this deep structure can be disrupted, resulting in a state of imbalance (Gersick, 1991). For example, employees can perceive a misfit while performing their tasks. Beyond that, however, some workarounds can also be adopted as official work practices (e.g., workaround 2 in Case 3), if they are seen as capable of enacting a routine, for example, if the IT artefacts that had been implemented did not provide a function required to perform a task. To remedy this imbalance, the system needs to break up existing structures and shift back to a state of balance through undergoing a reorientation cycle (Gersick, 1991).
Proposition 4. Workarounds can become established as official work practices or remain temporary fixes, subject to existing organisational structures.
Data from our cases show clearly that workarounds are performed on an individual level to fix misfits quickly. In Cases 1 and 2, before they were officially adopted as a new work practice on an organisational level, the workarounds were evaluated with a type of cost-benefit analysis. Complementing recent insights from Zamani and Pouloudi (2021), who found that users evaluate the cost or risks of using an IT artefact when performing a specific task on an individual level, we found that such an evaluation is also conducted on an organisational level (Case 2). It is reasonable to assume, however, that the evaluation of workarounds can also result in an overall negative assessment if it has been identified as inefficient, revealing concerns about the deviance, or even as a violation of law.
Either way, the organisation must have appropriate organisational structures in place to identify and assess a workaround, before deciding whether to adopt or reject it. Workaround 3 in Case 2 (manipulation of the spreadsheet) demonstrated that some workarounds were ignored deliberately over a period of 2 years since they occurred in a sub-ordinated process that, at the time, was not a priority for management. Comparing the results obtained from Cases 1 and 2 (i.e., more rigid inter-organisational scenarios with coordination type flow) with Case 3 (i.e., a more open intra-organisational scenario with coordination type share) suggests that the coordination type linking the activities performed by different employees can strongly impact the ultimate decision of whether or not to adopt a workaround. In Case 3, there was no governance mechanism in place to actively manage this process, thus, workaround 1 was never implemented on an organisational level, even after it had been communicated to coworkers and led to the adaptation of the IT artefact. An important prerequisite for implementing a workaround is the promotion of a culture of trust and innovation among employees. Some employees (like IV6 in Case 1) did not dare to reveal their workaround to their superiors, suspecting that it would be viewed as a negative deviance from standard operating procedures. Workarounds that remained unnoticed or were unfavourably evaluated by managers remained temporary and unofficial fixes that did not result in innovation on an organisational level.
Proposition 5. Workarounds can accumulate, to the point where their existence emphasises the need to adapt processes or IT artefacts in an organisation to resolve the underlying misfits.
Workaround 2 in Case 1 illustrates that top-down and innovation-oriented initiatives are triggered if the number of workarounds that emerge from one single misfit exceeds a certain threshold. We refer to this phenomenon as a workaround stack that can overflow, triggering subsequent actions. The threshold governing this overflow was, however, neither predefined nor standardised, but based on the intuition that eliminating a specific misfit was more effective than waiting for several workarounds to become official work practices. During the period of the case study, all three workarounds in Case 1 resulted in the official adaptation of either the IT artefact or of the process.
Second, workaround 2 in Case 2 illustrates that adaptation initiatives were also started for single workarounds that were critical on an organisational level. Comparing both observations, we conclude that if a workaround, either by itself or in combination with others, is perceived as constituting a substantial threat to an organisation and the underlying misfit cannot be ignored, organisations strive to resolve the misfit by restructuring, redesigning, and innovating IT artefacts, processes, and organisational routines. In Cases 1 and 2, governance structures were employed to manage organisational change projects, whereas in Case 3 such structures were absent and no innovation initiatives were recorded.
Proposition 6. The types of misfits and their individual perception impact the occurrence of workarounds.
As regards the occurrence of workarounds, our data revealed that some misfit categories led directly to workarounds, whereas others manifested indirectly. Clearly, the locus of a workaround is the performative aspect of a routine, referring to the ability of humans to do otherwise (Giddens, 1984), if they deem the workaround necessary to their ability to perform their day-to-day work effectively and efficiently. Category B misfits (a misfit of ostensive aspects and IT artefacts; see Figure 1) did not directly cause workarounds to occur, because they were distant from the human agency inherent in the performative aspects of routines. Instead, they remained latent until they caused considerable constraints to the day-to-day work of employees, prompting them to perform workarounds (e.g., workaround 2 in Case 3). This means that a Category B misfit had to first manifest as a Category A or Category C misfit that constrained the performative aspects carried out by employees. Although this view seems compatible with other categories of misfits and fit proposed by Strong and Volkoff (2010), these authors' perspective places the IT artefact centre-stage, whereas a workaround perspective focuses on performative aspects of routines and, therefore, places human agency centre-stage.
Viewing misfits from the perspective of workarounds, our data provide evidence that misfits do not always trigger workarounds. Although some employees experienced a misfit and (felt that they) needed to work around them, others in the same situation did not perceive it, and, therefore, did not perform a workaround. Beyond the necessary precondition of a misfit being perceived in the first place, another precondition for a workaround to occur lies in a misfit being seen as overly constraining, or hampering, the effectiveness or efficiency of employees performing their day-to-day work. The individual perception of a misfit seems to be mitigated by relevant training that an employee receives. Appropriate training can align the ostensive and performative aspects of routines, which decreases an employee's disposition to implement a workaround in response to a misfit (e.g., workaround 1 in Case 2).
If a misfit was perceived as constraining employees' day-to-day work too much, they searched for viable solutions to overcome it. Across all cases, different types of misfits were perceived on an individual level and manifested in workarounds, which-for the time being-remained undiscovered on an organisational level. In addition, our data reveal that the degree to which employees question their day-to-day work is related to the type of IT artefacts they use. Artefacts that strongly impose an inherent logic on individuals' routines tend to induce a bias that makes it less likely for individuals to perceive a misfit and, therefore, to decide whether they need to implement a workaround.
This finding is in line with related research stating that rigid organisational routines can lower the frequency with which employees change their routines through the implementation of workarounds (Yi & Becker, 2016 activities on an individual level, our empirical findings explore for the first time how workarounds diffuse in an organisation and, thereby, function as generative mechanisms for bottom-up process innovation that impact co-workers' activities, lead to the adaptation of IT artefacts, and restructure organisations as socio-technical systems. Our findings document that workaround diffusion is based on the two mechanisms of communication and observation. Both mechanisms have long been acknowledged as important concepts behind the formation and evolution of social systems (e.g., Luhmann, 1995) but, so far, they had not been associated with the diffusion of workarounds.
Although it is known that new routines spread in an organisation through communication, unfolding over time (Bresman, 2013), the underlying mechanisms of workaround diffusion remained hidden. Likewise, previous research has identified antecedents and types of workarounds (Goh et al., 2011;Zamani & Pouloudi, 2021), but has remained silent on the consequences of workarounds beyond their individual initiators (Tucker et al., 2014;Wolf & Beverungen, 2019). Building on structuration theory and its foundational observations on the duality of social structure and human agency (Giddens, 1984;Schmitz et al., 2016), our results document how workarounds innovate coworkers' activities, organisational structures, and IT artefacts. This view implies that workarounds play an active role for restructuring an organisation, complementing earlier studies that focus on explaining how and why workarounds occur based on the structure of an organisation (Alter, 2014;Davison et al., 2019;Pollock, 2005;Zamani & Pouloudi, 2021). This duality implies that the diffusion of workarounds needs to be analysed regarding their antecedents and consequences, considering that workarounds interact with their environment in a dynamic process. Beyond the extant literature, our results highlight that IT artefacts and coordination patterns have an important role to play in this process, because they can facilitate or constrain communication and the observation of workarounds on an organisational level.
We discovered that innovation processes are open-ended, i.e., constituting endogenous change processes in which workarounds are one particular type of variation that brings about process drift (Pentland et al., 2021). However, we posit that process innovation is neither exclusively endogenous (i.e., caused only by the performances of participants) nor exclusively exogenous (i.e., caused only by designers implementing a new business process). To leverage the full innovation potential of workarounds, ambidextrous change initiatives are necessary; as posited by Ciborra and Lanzara (1994, p. 63), "designing in action" comprises not only incremental changes to organisational routines but also radically new patterns of behaviour. On the one hand, workarounds act as a temporary fix, diffusing from the bottom-up in an organisation and enabling incremental innovations to occur, such as improved system functionalities or more efficient business processes. On the other hand, the accumulation of workarounds can trigger a more comprehensive set of transformations involving the review of organisational structures and business processes. We describe this tipping point with the metaphor of a workaround stack which, when it overflows, triggers a re-design project. Still, if the redesign is top-down driven, it can again lead to the emergence of new misfits  and workarounds, starting a new iteration of workaround diffusion in a socio-technical system.
On a more general level, we posit that the crucial role played by workarounds at the intersection of bottom-up process innovation and top-down process redesign can only be understood by combining the two lenses of business processes and organisational routines. Both lenses have been found to "study organizational processes, [although] they are completely separate" (Pentland et al., 2021, p.969). Recently, IS research has been increasingly discussing the emergent dynamics of business processes  and organisational routines (Feldman, 2000;Feldman et al., 2019). We join this discussion while addressing recent calls to provide an integrated research perspective that bridges the gaps between business processes and organisational routines (Beverungen, 2014;Breuker & Matzner, 2014;Pentland et al., 2020;Pentland et al., 2021). On the one hand, our results contribute to current research that focuses on the top-down (re-)design of business processes, but without having yet conceptualised nor investigated the actions performed by process participants who frequently deviate from pre-defined process blueprints, which can result in process drift. For years, the BPM community has vividly discussed how and why processes tend to drift over time (e.g., Maaradji et al., 2015;Pentland et al., 2020). However, process drift Pentland et al., 2020) is still a black box, obscuring its inner workings.
Although conceptual models like the BPM lifecycle (Dumas et al., 2018, pp.16-24) already consider endogenous change processes, it does not explicitly consider changes that occur through the actions performed by process participants, for example, employees establishing workarounds. Instead, workarounds are treated merely as part of data (traces) to be analysed through process mining before a process is re-conceptualised in an attempt of its purposeful re-design. On the other hand, although organisation science acknowledges routines as patterns of activities and dynamic change, it often ignores the active role played by IT artefacts. For instance, Pentland and Feldman (2008, p. 241) consider artefacts as mechanisms that "ensure the reproduction of particular patterns of action." However, the literature on organisational routines acknowledges that "actions can generate outcomes that make new and different actions possible or necessary" and that "some of these changes have the potential to be continuous" (Feldman, 2000, p. 613), and thus will lead to process drift. Still, the literature on organisational routines often view process change as purely endogenous (Pentland et al., , 2021, not linking it to the purposeful re-design of processes. Our workaround stack provides a conceptual device that bridges both perspectives, acknowledging that workarounds constitute bottom-up process innovation and also pile up to reach a tipping point at which top-down process redesign is conducted to resolve misfits permanently. By complementing both views, we propose that employing the concept of workarounds and studying their diffusion in organisations offers an effective means to investigate the inner workings of process drift, complementing related research that uses methods of BPM to detect workarounds in organisational routines (Beerepoot & Van de Weerd, 2018;Grisold et al., 2020;Weinzierl et al., 2020). In particular, the complex phenomenon of process drift  becomes better accessible by illuminating how workarounds in performative routines act as generative mechanisms that impact socio-technical systems in three distinct spheres, i.e., co-workers' activities, IT artefacts, and organisational structures. With this perspective, we take a meso-perspective on workarounds that is reflected in the Gordian knot of workaround diffusion. We posit that workarounds-especially in digitised processes -can lead to process drift as long as they remain hidden in the performative aspects of routines that process participants carry out, whereas managers take no corrective action. This is because workaroundswhile diffusing in an organisation through observation and communication-act as a generative mechanism for bottom-up process innovation, restructuring organisations, and adapting IT artefacts.
Finally, analysing the diffusion of workarounds adds to the discussion of change in organisations iterating "between the level of the individual agent and the collectivities to which the individual belongs" (Feldman, 2000, p. 625). Workaround diffusion can be viewed as a bottom-up process of organisational learning, knowledge acquisition, and innovation. As indicated in seminal research on knowledge management by Nonaka and Takeuchi (1995, p. 97), innovation and new knowledge "always begin with the individual." Against this backdrop, our findings complement theoretical perspectives on knowledge creation-as a spiral including socialisation, externalisation, combination, and internalisation-constituting organisational learning (Nonaka & Takeuchi, 1995). We posit that workarounds represent new tacit knowledge that is codified into organisational routines performed by employees. Through the mechanisms of observation and communication, co-workers become aware of a workaround performed by others, enabling the diffusion of knowledge in an organisation (Boudreau et al., 2016). Our findings are closely linked to the mechanisms of dynamic creation and conversion of knowledge over time (Nonaka & Von Krogh, 2009) and improvised learning (Boudreau & Robey, 2005), as employees experiment with different options to perform their day-today work more effectively and efficiently. Such "unplanned activities" were found to "spread knowledge among the user community" (Boudreau & Robey, 2005, p. 16) over time.

| Workarounds are temporary fixes of misfits
Our study also provides new insights into participants' experiences of underlying misfits, resulting in the implementation of workarounds. As proposed by Yin (2017), we entered the field with a theoretical framework, drawing on theory related to organisational routines and business processes, to search for three categories of misfits. We provisionally labelled the three misfits, respectively, Category A misfits, Category B misfits, and Category C misfits.
Although Strong and Volkoff (2010) identify misfit domains that partially align with our Categories B (e.g., functionality misfit, or data misfit) and C (e.g., usability misfit), they do not identify Category A misfits. Our interpretation is that in Strong and Volkoff (2010) Category A misfits manifest indirectly as Category B or C misfits, since their study focuses on the relations of enterprise systems with the organisation and its workers. Consequently, they do not consider that misfits can also emerge without the immediate involvement of technology. We agree with this view on misfits for contexts involving enterprise systems because they are designed to directly impose best practices of a routine on the day-to-day work performed by employees through the hard inscription of work practices (Davenport, 1998), leaving workers little opportunity to deviate from standard operating procedures. Enterprise systems-most prominently enterprise resource planning (ERP) systems-are intentionally designed to embed organisational elements and their relationships (Volkoff et al., 2007). Because in such contexts the organisational structures are inscribed into the IT artefact, a misfit between ostensive and performative aspects of a routine would manifest indirectly as a deficiency or an imposition of the enterprise system itself (Strong & Volkoff, 2010). Since we analyse scenarios involving different types of IT artefacts, with some giving employees more freedom than others on how the IT artefact can be used in their day-to-day work, we view misfits from a slightly different angle, enabling us to triangulate the misfit categories identified by Strong and Volkoff (2010). Since the main motivation of this paper is not to theorise new misfit types, but to identify how workarounds diffuse in an organisation, we call on others to investigate Category A misfits further, focusing on scenarios in which employees have agency to establish workarounds to change their performances, re-design IT artefacts, and design new IT artefacts, besides resolving misfits rooted in the deficiencies and impositions related to IT artefacts.
Independently of the misfit categories used, our results indicate that it is the employees' assessment of a misfit and its context that lead them to perform a workaround or related actions (see Table A1 in the Appendix) to compensate for the adverse effects that a misfit exerts on their day-to-day work. With regard to the magnitude of misfits, extant research often described their impact on a scale ranging from minor issues to critical deficiencies (Maurer et al., 2012). However, our findings indicate that not all misfits that are perceived by an individual will lead to the performance of a workaround (van Beijsterveld & van Groenendaal, 2016). Instead, we discovered that the ramification of a perceived misfit is biased by an employee's subjective impression, individual characteristics (e.g., attitude towards change), and capabilities (e.g., proper training). The dependence on an employee's traits fit with the current knowledge base, distinguishing perceived misfits (caused by an individual's characteristics and capabilities) from actual misfits (universally valid) (Goodhue & Gattiker, 2005). Additionally, we posit that critical misfits can cause particularly high organisational costs-causing many workarounds or business-threatening workarounds as remediesthat consequently attract the attention of management and induce countermeasures. Thus, substantial misfits often spawn significant organisational initiatives to resolve the underlying misfits. Since conducting these initiatives takes time, workarounds are needed as temporary fixes until their underlying misfits have been resolved.

| Managerial implications
Although the business environment appears increasingly volatile  and has recently been impacted by exogenous shocks (Roeglinger et al., 2021), managers have become aware that they need to question established routines and implement means allowing for more flexible process updates. There is, however, an inherent, even paradoxical, tension in the requirements of managing such processes. On the one hand, managers face "an T A B L E 6 Managerial implications.

Theoretical propositions Managerial implications
1: Workarounds can lead to restructuring ostensive aspects and adapting IT artefacts Understand that workarounds do not only compensate for misfits, but can also impact on IT artefacts and organisational structures. Identify the occurrences and consequences of workarounds to decide if they are desirable or undesirable for bottom-up process change.
2: Workarounds require communication or observation to diffuse in an organisation, subject to coordination patterns Establish appropriate degrees of communication and observation to facilitate desirable workarounds and prevent undesirable workarounds from diffusing. Manage the role that IT artefacts and coordination patterns play for communication and observation. Be aware that workarounds can diffuse even without managerial attention.
3: On an organisational level, workarounds can serve as means for re-balancing the triangular relationships between IT artefacts, organisational structures, and performances of routines Acknowledge that the triangle of organisational structures, individual performances, and IT artefacts is complex and volatile. Perceive workarounds as a step towards bottom-up process innovation to re-align the triangle.
4: Workarounds can become established as official work practices or remain temporary fixes, subject to existing organisational structures Accept desirable workarounds as temporary fixes to sustain day-today operations. Use desirable workarounds as a backlog for periodic process re-design. Acknowledge that ostensive aspects, performative aspects, and IT artefacts can never be aligned fully, causing a permanent need for managing workarounds as mechanisms for bottom-up process innovation.
5: Workarounds can accumulate, to the point where their existence emphasises the need to adapt processes or IT artefacts in an organisation to resolve the underlying misfits Observe existing workarounds and trace them back to their underlying misfits. Identify those workarounds that need to be addressed with top-down re-design projects but be aware that such initiatives will cause new misfits.

6: The types of misfits and their individual perception impact the occurrence of workarounds
Understand that workarounds point at (perceived) misfits of organisational structures or IT artefacts with performances. Search for relations among different types of misfits. Acknowledge workarounds' essential role to sustain day-to-day work in organisations ahead of top-down process redesign.
increasing level of complexity," while, on the other, "processes also must be implemented more quickly and more frequently" (Beverungen et al., 2021, p.154). Managing the diffusion of workarounds to implement bottom-up process innovation might help to "investigate how performances of business processes may contradict and refine IT artifacts as well as organizational structure" (Beverungen et al., 2021, p.154).
Our results convey actionable advice on how workaround diffusion can be managed in a way that benefits organisations (cf. Table 6). This is not an easy task. Ostensive aspects, performative aspects, and IT artefacts are intertwined in a triangular relationship that is both complex and volatile. As the triangle can never be fully aligned, there is a permanent need for managing workarounds and their diffusion.
Our results re-affirm and extend observations that workarounds can resolve situations in which ostensive aspects or IT artefacts are perceived as constraining participants in their smooth and effective performance of dayto-day work (Strong & Volkoff, 2010). Recognising the essential role of workarounds in sustaining day-to-day work and innovating business processes, we join related calls that dismiss the old misconception in the BPM literature that any processes deviance is undesirable per se (Bagayogo et al., 2013;Mendling et al., 2020), whilst acknowledging the need for compliance with pre-specified processes, and for individual electronic monitoring in specific scenarios (Staats et al., 2016). Workarounds can help to bridge the latency period between detecting and correcting a misfit, keeping organisational operations running. To facilitate these effects, however, managers need to create certain conditions to enable change. Assuming that "change is frequent, relentless, and even endemic" (Brown & Eisenhardt, 1997, p.2) in organisations, and that managing change is critical, the literature draws attention to the need of managers to first understand the underlying reason for change (how and why it comes about), before attempting to manage it. As part of change management, managers should also be willing to adapt business processes alongside reviewing or putting in place appropriate organisational structures.
Managers can use our insights to identify the occurrence and consequences of workarounds beyond their individual initiators, and, importantly, to distinguish desirable from undesirable workarounds. Based on this assessment, managers might decide to train participants in the use of software tools to ensure that standard processes are executed properly y (Strong & Volkoff, 2010), or they might approve the use of a workaround as a temporary fix for a misfit. While either strategy can help to re-align performative aspects with ostensive aspects and IT artefacts, a crucial prerequisite involves encouraging employees to voice the problems they encounter in their day-to-day work processes, because "understanding the different sources of problems […] can help professionals determine how and whether intervening can produce better outcomes" (Volkoff & Strong, 2013, p. 832).
If workarounds are perceived as viable solutions for misfits, they may then diffuse to co-workers through the mechanisms of communication and observation. This diffusion happens through official channels or unofficially, evading managerial attention. To enable managers to manage the diffusion, the Gordian knot of workaround diffusion equips them with the necessary knowledge of how the diffusion of workarounds constitutes either process dynamics (Pentland et al., 2021) or process drift . However, managers often view incremental changes to processes as a misalignment of official business processes and day-to-day practices, that lie beyond their control, while also often lacking a clear understanding of the forces that trigger these dynamics (Pentland et al., 2021).
Beyond using technical methods, including conformance checking (Aalst & Rozinat, 2008) and workaround mining (Outmazgin & Soffer, 2016;Weinzierl et al., 2021), and organisational tools, such as Lean Boards (Weinzierl et al., 2021), as aids to detect process drift (Maaradji et al., 2015), our insights enable managers to understand the active role of workarounds in bringing about process drift . Managers need to establish appropriate communication and observation practices to facilitate or constrain the role that workarounds play for organisational learning and bottom-up process innovation. These responsibilities include managing the role that IT artefacts and coordination patterns play in workarounds.
Beyond managing the diffusion of workarounds, managers should also link endogenous bottom-up innovations through workaround diffusion with top-down redesign initiatives, balancing participants' individual quests to improve day-to-day work with the overarching organisational goal of managing end-to-end business processes. Our metaphor of a workaround stack highlights the risk that workarounds will pile up over time if the underlying misfits remain unresolved, as performative routines become increasingly disconnected from IT artefacts and official work practices.
Managers can use the stack as a backlog to inform top-down initiatives for process redesign, establishing (some of the) workarounds as official work practices. Our results enable managers to "entrain" (Ancona & Chong, 1994) the cycles of workaround diffusion and official top-down redesign processes, that is, aligning the rhythm of both, to solve the paradoxical tension arising from more complex and, at the same time, more volatile business processes .

| CONCLUSION
Although workarounds are commonplace in organisations (Li et al., 2017), the mechanisms of their diffusion are still under-investigated and under-theorised (Davison et al., 2019;Malaurent & Karanasios, 2019;Zamani & Pouloudi, 2021). The extant research focuses on exploring workarounds as user behaviour to remedy constraints that limit the effectiveness or efficiency of performing their day-to-day work (Alter, 2014;Davison et al., 2019;Pollock, 2005;Zamani & Pouloudi, 2021), but it is unclear what consequences they can have at an organisational level, beyond their initiators.
By performing a multiple case study in three organisations, we integrate theoretical knowledge on business processes and organisational routines to identify how workarounds diffuse in organisations, acting as generative mechanisms for bottom-up process innovation. With this view, we make workarounds accessible for further investigations of organisational change, process innovation, and process drift, complementing recent research of workarounds as a software-or hardware-focused accommodating practice, following negative disconfirmation in volitional settings (Zamani & Pouloudi, 2021). We find that workarounds act as generative mechanisms that restructure an organisation from the bottom-up, complementing top-down initiatives of process design and intervention. Our results provide unique empirical insights for research and practice on how to leverage the potential of workarounds as a mechanism for unveiling organisational drift. We provide a fresh lens to the study of drift with conceptual accuracy, uncovering its underlying mechanisms as an interplay between misfits, workarounds, and organisational routines.
Although case study research enabled us to obtain rich empirical insights on workarounds as contemporary phenomena that are inextricably linked with their real-world context, this approach does not come without its limitations.
First, we acknowledge that workarounds might be performed subconsciously, with employees not always aware of deviating from standard processes. Previous research has argued that the ostensive aspects of an organisational routine are subject to interpretation on an individual level (Pentland & Feldman, 2008), obscuring the general patterns of how a routine ought to be performed. If official processes are not explicated with process models or other documentation, or embedded into information systems, individual interpretations might flourish, making it hard for process participants and managers to distinguish a workaround from official control flows that have been purposefully embedded into a business process. Second, all the workarounds we identified in our cases were found to have positive effects on the organisations, alleviating significant misfits. Nevertheless, workarounds are ambivalent as they can result in both desirable and undesirable consequences (Alter, 2014). This dual potential means that our results might overemphasise the more favourable side of the coin. However, although workarounds with negative consequences were not reported in our interviews, maybe because of participants' embarrassment or fear of sanctions to be imposed by management, "one cannot categorically argue that unintended actions are good, any more than one can argue that they are bad" (Boudreau & Robey, 2005, p. 16). Because of this reluctance, the dysfunctional effects of workarounds may not have been given due consideration here, but clearly warrant future research.
Both limitations further strengthen our view that integrating the organisational routines and business process literature can help to substantially advance our understanding of workaround diffusion in organisations, and of their impact in socio-technical systems. Future research can build on our explorative results to explain the complex causeeffect relationships relating to workaround diffusion in more detail. In addition, we anticipate that data-driven methods for identifying and categorising workarounds in an unbiased fashion, for example, process mining (Weinzierl et al., 2021), can augment qualitative case studies that explain how workarounds occur and diffuse, enabling us to develop richer insights on the role that workarounds play in organisations.

ACKNOWLEDGMENT
Open Access funding enabled and organized by Projekt DEAL.

FUNDING INFORMATION
As part of the DIGIVATION project (promotion sign 02K14A220), this research was funded by the German Federal Ministry of Education and Research (BMBF).

DATA AVAILABILITY STATEMENT
Data subject to third party restrictions. Data are available from the authors pending the approval of the organizations (third parties) surveyed in the case study.

Concept Definition References
Decoupling/ Loose Coupling Separating formal rules from actual working practices. Azad and King (2012) Deviance Divergence in how a process is performed over the course of time Pentland et al. (2020).

Fraud
Ill-intentioned employees use the system for prohibited aims. Bagayogo et al. (2013) Interpretive Flexibility Employees assign their own meaning and interpretation to the functions of an information system Orlikowski (1992) Non compliance Security best practices and policies that are avoided. Jenkins and Durcikova (2013) Resistance Behaviours intended to prevent the implementation or use of a system or to prevent system designers from achieving their objectives. Lapointe and Rivard (2005) Sabotage Workplace sabotage is behaviour intended to "damage, disrupt, or subvert the organisation's operations for the personal purposes of the saboteur by creating unfavourable publicity, embarrassment, delays in production, damage to property, the destruction of working relationships, or the harming of employees or customers. Crino (1994) Shadow System/ IT/Work Software applications or extensions to existing software that are neither developed nor controlled by an organisation's central IT department.

Subversion
User modifies the task approach to take advantage of known weaknesses in a system Koopman and Hoffman (2003) System Misuse Perform a behaviour that misuse of IS resources. D'Arcy et al. (2009) Tweaking Deviation from a prescribed work processes by using a system in a slightly different way. Boudreau and Robey (2005) Unfaithful Appropriation Features are designed to promote the technology's spirit, but they are functionally independent and may be appropriated in ways that are not faithful to the spirit.

DeSanctis and Poole
Workarounds A workaround is a goal-driven adaptation, improvisation, or other change to one or more aspects of an existing work system in order to overcome, bypass, or minimise the impact of obstacles, exceptions, anomalies, mishaps, established practices, management expectations, or structural constraints that are perceived as preventing that work system or its participants from achieving a desired level of efficiency, effectiveness, or other organisational or personal goals.
Alter (2014) Workplace Aggression Efforts by individuals to harm others with whom they work, or have worked, or the organisations in which they are currently, or were previously, employed. This harm-doing is intentional and includes psychological as well as physical injury. Baron and Neuman (1996) A.

Ignore Workaround
T A B L E B 3 Workaround 3: Use former PDF list.

IT Artefact
Fit Category C Misfit

Prevent Workaround
Organisational Structure "It is all based on the Software C's calendar. It sets the deadlines, when the basic plan is to be drawn up, when the room planning begins, when it is to be published, etc." (IV4)

Ignore Workaround
T A B L E D 3 Workaround 3: Request room change via email.