Technological Embeddedness and Organizational Change

Organization Science

informs

Vol. 18, No. 5, September–October 2007, pp. 832–848 issn 1047-7039 eissn 1526-5455 07 1805 0832

®

doi 10.1287/orsc.1070.0288 © 2007 INFORMS

Technological Embeddedness and Organizational Change Olga Volkoff

Faculty of Business Administration, Simon Fraser University, 8888 University Drive, Burnaby, British Columbia, Canada V5A 1S6, [email protected]

Diane M. Strong, Michael B. Elmes

Department of Management, Worcester Polytechnic Institute, 100 Institute Road, Worcester, Massachusetts 01609 {[email protected], [email protected]}

W

hile various theories have been proposed to explain how technology leads to organizational change, in general they have focused either on the technology and ignored the influence of human agency, or on social interaction and ignored the technology. In this paper, we propose a new theory of technology-mediated organizational change that bridges these two extremes. Using grounded theory methodology, we conducted a three-year study of an enterprise system implementation. From the data collected, we identified embeddedness as central to the process of change. When embedded in technology, organizational elements such as routines and roles acquire a material aspect, in addition to the ostensive and performative aspects identified by Feldman and Pentland (2003). Our new theory employs the lens of critical realism because in our view, common constructivist perspectives such as structuration theory or actor network theory have limited our understanding of technology as a mediator of organizational change. Using a critical realist perspective, our theory explains the process of change as a three-stage cycle in which the ostensive, performative, and material aspects of organizational elements interact differently in each stage. Key words: enterprise systems; organizational change; embeddedness; structure; agency; critical realism; grounded theory

Introduction

In particular, as various organizational elements, e.g., routines and roles, become embedded in the technology, they acquire a material aspect that plays a key role in organizational change. To develop our theory, we conducted an intensive, longitudinal case study using grounded theory methods to observe the unfolding process of organizational change from a fresh empirical perspective not bound to existing theories. As we collected and analyzed our data, we identified theoretical categories that led us to examine related concepts in the literature. In keeping with a grounded theory methodology, we did not enter the field with a set of predefined theoretical concepts or hypotheses or with a particular body of literature in mind, but instead sought a “practical middle ground” (Suddaby 2006, p. 635) that drew iteratively from the empirical data we collected and from our knowledge of substantive theories that pertained to the theoretical concepts emerging from our data. This avoided Suddaby’s (2006) concern that grounded theory should not be an excuse to ignore the literature. During our analysis, we also identified critical realism as an appropriate lens for examining the actions and interactions of stakeholders in their appropriation and use of technology, while simultaneously accounting for the role of technology. Critical realism also incorporates a temporal aspect that fits well with a change process, which by definition occurs over time. For clarity of presentation, we discuss this literature prior to presenting our results, but in reality, we examined it as data collection and analysis indicated its relevance.

The introduction of information technology (IT) into an organization is generally accompanied by changes to organizational form and function. While early research took a deterministic perspective to explain such changes, contradictory results underscored the weakness of such an approach and suggested that greater value would be derived from studying the process of change rather than the (often idiosyncratic) outcomes (Robey and Boudreau 1999). Using a variety of constructivist approaches, various models of the change process have been proposed. These include studies based on structuration theory (Orlikowski 1992, Orlikowski and Robey 1991), institutional theory (Avgerou 2000, Gosain 2004), and actor network theory (ANT) (Latour 1996, Walsham 1997). Not only do these different perspectives present conflicting views about how technologymediated organizational change occurs, but each in its own way is problematic. Those based on structuration theory or actor network theory tend to focus on the actions of agents, ignoring the technology, while those using institutional theory tend to ignore agency. Furthermore, technology itself is often treated as a unitary object, ignoring each technology’s distinctive characteristics, which should be acknowledged in a theory of technology-mediated organizational change. In this paper, in response to the research question “how does technology mediate organizational change,” we propose a new theory that addresses the specific role of technology, while also incorporating the effects of agency. 832

Volkoff, Strong, and Elmes: Technological Embeddedness and Organizational Change Organization Science 18(5), pp. 832–848, © 2007 INFORMS

Background Literature Organizational Change: The Case of Organizational Routines Analysis of our data suggested that we consider how technology affects various organizational elements as our level of analysis and that we start by understanding the constituent parts of each of those elements. Within the literature, the study of organizational routines provided a model for how to proceed. Organizational routines have been defined as “repetitive, recognizable patterns of interdependent actions, carried out by multiple actors” (Feldman and Pentland 2003, p. 95). Since Nelson and Winter (1982) first recommended organizational routines as a unit of analysis for examining organizational change and stability, a vast literature on routines has accumulated (see Becker 2004 for a comprehensive review). While much of this work has focused on the stability and even inertia of routines, Feldman and Pentland examined the changeability of routines and the endogenous processes that are responsible (Feldman 2000, 2003; Feldman and Pentland 2003; Pentland and Feldman 2005). They also opened the black box of routines to identify the constituent parts. Using Latour’s (1986) language, they distinguished between the ostensive or structural aspect of routines, and the performative, or agentic aspect. The ostensive aspect is the “abstract, generalized idea of the routine, or the routine in principle” (Feldman and Pentland 2003, p. 101), while the performative aspect “consists of specific actions, by specific people, in specific places and times the routine in practice” (ibid). Feldman and Pentland used a structuration perspective (Giddens 1984) to discuss how these two aspects of routines are related; the ostensive aspect guides, helps account for, and enables the performative, and the performative aspect helps to create, maintain, and/or modify the ostensive. Pentland and Feldman (2005) subsequently extended this analysis by looking beyond these two aspects of a routine to the various physical artifacts associated with the routine. They highlighted the need to study the relationships among each aspect and various artifacts ranging from written rules and machines or computers to office layout. While they called for research on such relationships, they did not present specific findings or hypotheses. Our grounded theory study started with a general objective of examining how technology affects organizations, not a specific focus on any particular organizational element such as routines. During data collection and analysis, we recognized that what we were seeing related to and, among other contributions, extended the work of Feldman and Pentland (2003) in a number of ways. First, in examining changes in organizational routines, we observed that in addition to ostensive and performative aspects, routines also have a material aspect

833

that is embedded in the technology. Whereas Pentland and Feldman (2005) included IT in the broad set of physical artifacts associated with routines, we observed that the IT artifact is different from those other artifacts because it is an integral part of those routines, not just part of the context within which routines are executed. As will be shown, the material aspect of routines plays a critical and direct role in the change process. Second, routines are not the only organizational elements that IT changes. In our study, we found that IT also affected roles and data. Furthermore, these other organizational elements can be seen to have an ostensive, a performative, and a material aspect that interact in specific ways. Third, in addition to the interactions of the different aspects of each organizational element, change arises from the relationships among the material aspects of different organizational elements. For example, the IT changes the relationship between routines and roles by embedding this relationship into the system. Fourth, there are other organizational elements such as mindset and organizational culture that, while they do not become physically embedded in the technology and so do not acquire a material aspect, are changed with the introduction of IT. These changes are second-order effects that arise because of the connection between these unembedded organizational elements and those that do become embedded. Finally, while Feldman and Pentland’s (2003) use of a structurationist lens has focused attention on the ostensive and performative aspects of organizational elements, it neglected the material aspect. By using a critical realist lens instead, we are better able to address the inherent materiality of technology. Organizational Change: A Mediating Technology Because different technologies have different properties, the choice of which technology to study will color what we learn about technology’s effects on organizations. For our research, we chose to look at enterprise systems, defined as, comprehensive commercial software packages designed to support and integrate organizational processes across functional boundaries (Davenport 1998). They are built on a common database and are used to execute the complete range of organizational transactions (Klaus et al. 2000, Markus and Tanis 2000). As packages, they are designed to offer generic functionality for use by many companies, not custom-built solutions for a specific organization. To accommodate organizational variation, these systems are configurable, but not infinitely malleable. Unlike simpler systems that may only affect an individual or a small work group, an enteprise system (ES) has organization-wide effects, and as such is an excellent choice for studying the relationship between technology and organizational form and function.

834

Volkoff, Strong, and Elmes: Technological Embeddedness and Organizational Change

Organizational Change: Different Theoretical Perspectives Two papers illustrate the way different theoretical perspectives color the analysis of such systems. At one end of the spectrum, Boudreau and Robey (2005) take a human agency perspective and argue that an ES can be resisted or reinvented. From this perspective, as users redefine the technology through their actions, it loses salience and its role in affecting organizational change is weakened. By contrast, Gosain (2004) employs institutional theory to argue that enterprise systems are objects of institutionalization during configuration and carry the institutional logic during use. From this view, technology actively constrains human agency. Researchers have employed a variety of approaches to bridge these two extremes, with varying success. Structuration theory, the constructivist perspective most commonly used to examine the technology-organization relationship, views technology as “interpretively flexible” (Orlikowski 1992, p. 405), enacted and defined at the moment of use. In this perspective, the material aspects of technology often play a minor role in the change process, with attention focused on the actions and interactions of individuals. Thus, in Barley’s (1986, 1990) work examining CT scanners, the effects of technology on the organizations were examined exclusively through the interactions between people and the implications for their roles in the organization. The technology provided an “occasion for structuring” (Barley 1986, p. 78), but was otherwise a bit player in the unfolding drama. Under structuration theory as delineated by Orlikowski (2000), technology stops being an artifact, and becomes a malleable “technology-in-practice” (p. 409), whereby users constitute (or reconstitute) emergent technology structures through their actions. In structuration theory, the relative neglect of technology’s role in the change process arises from Giddens’ (1984) insistence that structure only exists in the moment of instantiation as traces in the mind. Without an actor, there is no structure. While this perspective may be appropriate for conceptualizing social structures that have no concrete form, it ignores the inherent materiality of technology (Jones 1999). Once we acknowledge that technology, while interpretively flexible, is not infinitely so, we must be specific about its material aspects (Monteiro and Hanseth 1996). Various researchers, most notably those who subscribe to an ANT perspective, have challenged the short shrift given to the technology artifact by structurationalists when analyzing the technology-organization relationship (e.g., Hanseth et al. 2004, Monteiro and Hanseth 1996, Walsham 1997). ANT downplays the distinction between structure and agency, putting technology on the same footing as individual actors (collectively referred to as actants), and viewing them all

Organization Science 18(5), pp. 832–848, © 2007 INFORMS

as participants in a network of heterogeneous components (Mutch 2002). Under this perspective, an analysis of organizational change entails examining the process of negotiation whereby different actants work to enroll each other as allies, and through which their interests are translated and become aligned (Law 1992, Walsham 1997). Despite the inclusion of technology as one of the actants, the focus on negotiation, a human activity, privileges human actors. The technology is viewed as a receptacle where an actor’s perspective can be inscribed and frozen. This is particularly useful for examining the design and implementation phase of technology, but we suggest that once technology has been implemented and is in use, it still plays a role in organizational change. Thus, ANT offers a high-level explanation based on negotiation, but fails to capture the active role of technology in the negotiation. While ANT acknowledges the material aspects of the technology much more than structuration theory, its conflation of agents and structures reduces its capacity to examine how technology mediates change in organizations (Mutch 2002). It is also primarily descriptive rather than explanatory (Howcroft et al. 2004, Walsham 1997). Moreover, by examining only social action at a microlevel, the context in which that social action occurs— the relatively enduring institutionalized relationships that constitute structure and constrain agency—is ignored (Reed 1997). Recently there have been calls to apply critical realism (CR) to the study of IT-mediated organizational change (Dobson 2001; Mingers 2002, 2004; Mutch 2002), although to our knowledge, no empirical studies have yet done so. Critical realism has the potential to address the shortcomings identified above. Organizational Change: A Critical Realist Perspective Critical realists carefully separate structure and agency, claiming that other ontological perspectives in social science either ignore structure (behavioral approaches), ignore agency (institutional theory), or conflate the two (structuration theory and ANT) (Archer 1995). Only by keeping them conceptually distinct, can we logically discuss their effects on each other. While distinct, structure and agency cannot exist without each other. Specifically, social structure logically predates the practices it engenders. Furthermore, through those practices different structures may emerge, but these changes to structures necessarily postdate those actions (Bhaskar 1979, Collier 1994). Thus, Archer’s (1995) critical realist theory of social change involves cycles of three phases, which we use in our theory of technology-mediated organizational change, namely, (1) structural conditioning, the preexisting structural properties that are the consequence of past actions, followed by (2) social interaction, during which agents engage with, and are constrained and enabled


by, the preexisting structural conditions, followed by (3) structural elaboration/reproduction, the modification of previous structural properties and the introduction of new ones or the reenforcing of existing structures. While social interaction is continuous, at any moment the emergent structure depends on past activities, not on the actions of current agents. Agents in turn are shaped and reshaped as they engage with the structures they confront, but did not create (Archer 1998). Use of any theoretical perspective requires the use of a methodology that matches the underlying ontology and epistemology. Ontologically, in CR, there are three nested domains, namely, the real (structures or mechanisms that exist independently of us), which contains the actual (events that those mechanisms could potentially generate, whether or not they occur), which in turn contains the empirical (the subset of those events that are observed or experienced) (Bhaskar 1998, Mingers 2002). Because the mechanisms operate in open systems, their effects may or may not occur, and those that occur may or may not be observed (Bhaskar 1978, Outhwaite 1998). The task of the researcher, then, is to use perceptions of empirical events to identify the mechanisms that give rise to those events (Collier 1994). Our grounded theory approach of constantly comparing data units to abstract higher-level concepts (Glaser 1978, Glaser and Strauss 1967, Suddaby 2006) was well aligned with CR’s epistemology, namely, the use of observed events to infer the mechanisms through which the underlying real but unobservable structures generated those events (Collier 1994). Critical realist theories, like our new theory, are explanatory process theories. They explain an event not by predicting what will happen nor describing what did happen but, rather, by identifying the mechanisms that generate what we observe in the empirical domain. Following Archer (1995), these are the mechanisms by which preexisting structural conditions shape agency, and in turn the mechanisms by which agency changes or reinforces these structural conditions. Critical realism, like Gidden’s (1984) structuration theory, is a grand theory that does not mention technology. The role of technology in organizational change, however, is easily conceptualized as part of the structural conditions agents encounter, making CR an ideal perspective for developing midrange theories of technologymediated organizational change. Thus, we view an ES as a source of structural conditioning that is relatively independent and enduring, existing materially in the real domain, rather than primarily as a malleable structure, existing only in the empirical domain at the moment of instantiation. Using a critical realist perspective, we can discuss the interplay between structures and human agents, and examine the generative mechanisms or mediators through which agents affect structure and, of greater importance for this study, how structures shape agency.

835

Methodology

The implementation of an ES is a lengthy process, taking from several months to several years. To uncover and examine the changes occurring from such an implementation, we followed three phases of a multiyear, multiphased implementation at ACRO, a multinational producer of precision industrial products. Capturing the complete picture of the change process necessitated following each phase studied from before its go-live, so we could establish a baseline, through the implementation and initial period of use, to a period of stable use. Thus, our research involved a longitudinal, intensive case study, and employed grounded theory methods (Glaser and Strauss 1967, Strauss and Corbin 1998). With grounded theory methods, data collection, data analysis, and theory building are overlapping and iterative processes that enable theory to emerge organically, as researchers associate increasingly more abstract conceptual categories with the data (Suddaby 2006). Through these methods—especially as we analyzed our data and let these data tell us what was changing in the organization as a result of implementing the ES and why—we identified CR as an appropriate lens through which to identify the theoretical concepts that would explain how organizational members interpreted the organizational changes they observed (Suddaby 2006), and to begin to build theory from these concepts. Strauss and Corbin (1998, p. 268) stipulate four criteria for evaluating a qualitative, grounded theory study. One makes judgments about (1) the data, i.e., the validity, reliability, and credibility of the inputs to the research process, (2) the theory itself, i.e., the credibility of the output of the theory-development process, (3) the adequacy of the research process through which the theory is generated, focusing on analysis methods, and (4) the empirical grounding of the research, i.e., the grounding for the resulting concepts and theory. These mirror the assessment criteria of Miles and Huberman (1994), who approached qualitative research from a critical realist perspective. The following sections on our research site and data collection methods cover Criterion 1. The sections on our analysis and theory building methods cover Criteria 3 and 4. The resulting theory and coverage of Criterion 2 are presented later in the paper. Overview of the Research Case ACRO is headquartered in the northeast United States, with approximately 20,000 employees working in plants around the world. In 1998 the company decided to implement SAP’s R/3 software suite, and run it as a single system across the entire company. To make implementation manageable, the company divided the project into six phases. When we started our observations, the first two phases were already finished. Both were

836


relatively small, localized implementations used to solve some immediate problems and to let the company gain experience with the software. We began data collection in August 2000, before Phase 3, the first major phase, went live. This phase focused on component production and assembly operations, and entailed implementing a broad set of SAP modules including Finance, Production Planning, Sales Order Management, and Materials Management. Once this phase moved past go-live, the company started Phase 4, which brought the ES to the maintenance and repair operations. Phase 5, the last we observed, extended the ES to the research and development (R&D) group. Data Collection To ensure the validity, reliability, and credibility of the data collected, we used multiple methods and informants. Specifically, we collected data through observations, interviews, and informal conversations. Over the course of 150 visits (approximately one trip per week over three years), we conducted 72 formal interviews with 60 different people (see Table 1). In addition to these formal interviews, we had casual conversations as we accompanied people to meetings, as we observed them during testing and training sessions and in the war rooms during go-live, and as we shadowed them as they solved implementation problems. We also had the opportunity to listen in on group conversations, both casual and formal, and collect documents such as training material and meeting handouts. Over our three-year engagement at the site, we got to know our key informants well, and they appeared to be candid in voicing their thoughts about the new system and its effect on the organization. Observations were captured in field notes taken on site and written up after each visit. With few exceptions, formal interviews were taped and transcribed. During initial interviews, lasting from 30 to 60 minutes, we asked people to describe their job—their responsibilities, activities, and interactions with others—and to speculate on what would change after the ES was implemented. During follow-up interviews conducted shortly after go-live, we asked the same people, and others they suggested, to talk about what had changed and to indicate what they particularly liked and disliked about the new software. Final interviews were conducted about a year after go-live. As interviews progressed, we supplemented our initial questions with additional ones based on our emerging understanding of technology-related change. Our key informants were, for the most part, power users—individuals who were not members of the ES team, but who had been selected from operational units to work with the team during testing, to conduct training, and then to return to their units as resident experts and the first line of support. To ensure broad theoretical


Table 1

Distribution of Interviews Across Phases, Locations, and Roles Number of times each person was interviewed

Phase 3 Location 1

Phase 4 Location 1

Location 2

Location 3

Phase 5 Location 4∗

Totals

6 key informants 1 manager 3 power users 2 users 6 nonkey informants 1 manager 5 users 2 key informants 1 power user 1 user 4 nonkey informants 4 users 6 key informants 6 power users 8 nonkey informants 1 manager 3 power users 4 users 10 nonkey informants 3 managers 2 power users 5 users 7 key informants 5 power users 2 users 11 nonkey informants 9 power users 2 users 21 key, 39 nonkey∗∗ 6 managers 29 power users 25 users

1 person × 2 times 3 people × 2 times 2×2 1×1 5×1

1×2 1×2 4×1 5 × 3, 1 × 1 1×2 3×2 2 × 2, 2 × 1 3×1 2×1 5×1

1 × 3, 2 × 2, 2 × 1 1 × 1, 1 × 2 3 × 2, 6 × 1 2×1 8∗∗∗ 53∗∗∗ 31∗∗∗

∗ While most of the operations related to Phase 5 occur at Location 4, several of these individuals also spend time at Location 1, and some of the interviews occurred at Location 5, where the ES team was located. ∗∗ Key informants were those people we not only interviewed, but also shadowed during our weekly site visits as they attended meetings, performed testing, delivered training, or solved problems. Nonkey informants were those who we interviewed, but did not shadow. ∗∗∗ Because some of the individuals were interviewed together, there were only 72 separate interviews conducted to obtain these 92 person interviews.

sampling, we also interviewed some managers and general users. Interviewees worked in different functional areas at four different plants. Operations ranged from assembly, to repair and maintenance, to R&D. Our interview subjects included customer service representatives, material buyer planners, inspectors, engineers, expeditors, shipping clerks, inventory clerks, plant managers, accounts payable managers, and plant controllers.


Data Analysis and Coding Consistent with a grounded theory approach, our data collection and analysis occurred iteratively. As field notes and interviews were transcribed, they were coded. Analysis followed Strauss and Corbin’s (1998) recommendations for open, axial, and selective coding of all interview data. Specifically, each passage (from one to several sentences in length) in every interview was assigned one or more codes reflecting what the coder perceived the speaker to be talking about. The code labels were suggested by the words used in the interviews. The NVivo software package was used to track the coded interviews and field data. Theory development occurred iteratively with coding. Thus, as coding progressed, we organized codes into trees, compared similarly coded passages to generate more abstract theoretical concepts, and wrote memos to propose abstract concepts and potential relationships. The different research traditions of the three researchers enabled us to challenge and critique each other’s ideas and to ensure that our research process stayed grounded in the data but also drew from relevant literature. During the open coding stage, our codes were primarily substantive using the vocabulary of the interviewees, e.g., “misfits,” “responsibilities,” and “impacts of SAP.” The first few interviews were coded by all three researchers. We then met to discuss the codes to ensure that we all used the initial labels to mean the same thing. Subsequently, each interview was coded by the researcher who conducted it because recalling the body language used to generate the words on a page helps to keep the interpretation of the words aligned with the speakers’ intentions. As we each added to the initial list of codes, we shared the labels and their intended meanings with each other to ensure consistent coding. Our final list included 53 codes covering the broad set of concepts in our interviews. As concepts of interest emerged from our open coding, we initiated axial coding. Strauss and Corbin’s (1998) paradigm recommends axial coding once a phenomenon (category) is identified and further explanation is desired. Similar to Glaser’s (1978) recommendation of coding around a core category, the researcher uses axial coding once open coding has uncovered phenomena of interest, in our case misfits and other organizational-level effects of the ES. We followed Strauss and Corbin’s (1998) axial coding recommendations, but treated this technique as a method through which to discover the relationships in the data, rather than as a restrictive set of rules. We did this to avoid Glaser’s concern that axial coding is overly restrictive, with questions emanating from the researcher rather than the data (Locke 2001). Suddaby (2006) expressed similar concerns, noting that grounded theory techniques require active interpretation by the researcher, not mechanical application of techniques. Thus, while we

837

used the axial coding recommendations of Strauss and Corbin (1998), i.e., coding for conditions (causal, intervening, and contextual), actions/interactions, and consequences, we also used Glaser’s six C’s coding category, i.e., coding for conditions, causes, covariances, contingencies, context, and consequences, as well as Glaser’s dimension family, i.e., coding for different types within a category, in our case different types of organizational effects, when we observed particular dimensions of ES effects (Glaser 1978). Adjusting the coding categories to fit the data rather than trying to force the data into predetermined categories is supported by both Glaser and Strauss (Urquhart 1997). As an example of our coding, a passage originally coded as an “impact of SAP” and “misfit” during open coding was further coded during axial coding as a “data misfit” (a type of misfit), and more specifically as a “data quality problem” due to “missing data.” This passage was also coded as a “contingent intervening condition” because the employee was having difficulty performing a task for which some required data were missing. After axial coding of all of our interview and field data, our code list had expanded to 376 codes. This large number of codes reflects two characteristics of our coding process. First, because we did not know what theory might emerge from our data, we coded broadly to cover several possible theoretical approaches to understanding technology-mediated organizational change, e.g., we coded for, but did not use, employees’ assessments of how well the implementation was going. Second, we coded for more depth than needed for the resulting theory, e.g., 37 codes are labeled consequences with some qualifier such as time, workload, visibility, or master scheduling. These are in addition to other codes labeled misfits or impacts of SAP, which are also consequences. Although such detailed coding was useful in the constant comparison process, not all of it contributed to the resulting theory. Theory Building Selective coding is a process of integrating and refining categories with the goal of building and refining theory (Strauss and Corbin 1998). Thus, selective coding moves the analysis from a potentially large number of codes produced from axial coding to a few theoretical categories that form the foundation of a new or revised theory. Selective coding involves integration and abstraction through comparison of coded passages, as well as comparison to the literature. To facilitate this process, we used NVivo to produce various reports of all the text coded with particular subsets of the axial codes. We then read and reread these NVivo reports, comparing coded passages to each other both within and across categories for similarities and differences related to conditions, actions, and consequences.


838


Table 2

Coding Statistics for Key Codes Code name

Number of passages with this code

Number of sources with passages having this code

Functionality (61 codes)

Process gaps Process misfits General impacts on processes Transaction reversal Interfaces Reports/reporting

24 25 30 27 18 22

14 14 17 18 13 19

Roles (16 codes)

Role gaps General impacts on roles Workload

24 30 56

16 12 34

Data (32 codes)

Data Data Data Data

32 28 9 43

19 16 7 21

Control (23 codes)

Control Visibility

56 39

26 20

Mindset (13 codes)

Discipline Organizational culture Mindset

61 8 25

34 7 17

Code group

quality availability detail (other)

Throughout selective coding, our focus moved iteratively among consequences, causes and conditions, and reading and integrating relevant literature. For ease of presentation, we report the process for consequences followed by that for causes and conditions as though they were performed in sequence. We initially focused on categories of consequences, i.e., the effects of the ES in the organization. As a result, we collapsed many codes from axial coding into five overall categories of organizational changes, namely, changes to functionality, roles, data, methods of control, and mindset, each with subcategories. See Table 2 for the number of detailed codes that apply to each of these five categories, and the number of passages and data sources that had codes representing each subcategory. As we examined causes, we realized that the first three categories corresponded to organizational elements that were embedded into the software, namely, routines, roles, and data, and thus, these consequences arose directly from this embedding, i.e., they were first-order effects. The other two, control and mindset, were consequences or second-order effects of the embedding of the three organizational elements and the embedding of the relationships between them, e.g., embedding the relationship between routines and roles. Furthermore, as we viewed these consequences through the lens of CR, we realized that our theory should focus on causes, i.e., mechanisms of change, rather than the observed organizational consequences. Understanding the abstract categories of the observed consequences, however, was necessary for developing the theoretical categories related to causes, as the epistemology underlying CR indicates. Thus, we examined the passages that were coded as conditions and causes. We compared similarly and differently coded passages, the same process we used

for understanding effects and consequences. We also compared causes to develop a theoretical understanding of the causes of the observed changes. As a result of these comparisons, an overall cause emerged—the organizational element was embedded in the ES software, e.g., an organizational routine was coded into the software resulting in changes to how that routine could be performed. Thus, from this analysis, the abstract concept of embeddedness emerged. That is, the causes and conditions for the organizational changes were due to embedding particular organizational elements, e.g., work routines and role definitions, and relationships among these organizational elements into the software. While the overall cause or mechanism of change was the embedding of organizational elements in the software, our data showed that the result of that embedding differed for different organizational elements. Viewing this observation through the lens of CR, we arrived at the conceptual understanding that each embedded organizational element takes on a material aspect. These ideas are more thoroughly presented in the next two sections. The first presents our results, focusing on grounding the concepts of embeddedness and material aspect in our data, but including a brief overview of organizational consequences so that readers can see how the observed consequences resulted from embedding organizational elements. The second presents our resulting theory, as informed by CR.

Field Site Observations of Embedded Organizational Elements

The presentation of our results in this section is organized around the three core organizational elements—organizational routines, roles, and data—that were changed.


We argue that the organizational changes we observed are attributable to the embedding of these organizational elements into the ES and the resulting tension between the elements as embedded and as previously experienced. Embedding an organizational element in technology introduces a material aspect of those elements that is different from but related to the ostensive and performative aspects. While this paper is about the organizational change process, not outcomes, we present some of the consequences of introducing an ES because it is from these observed elements that an understanding of the process is derived. Discussing consequences also allows us to distinguish between direct, first-order effects on those elements that were embedded and other effects, such as how control was exercised and how organizational mindset changed, that were second-order consequences of these first-order effects. Embedded Routines As we collected data about changes associated with the ES implementation, we observed many instances of changes to routines that had become embedded in the technology. Organizational routines are embedded in the ES in the form of system-executed transactions— sets of explicitly defined steps that require specific data inputs to automatically generate specific outcomes. For example, with the appropriate inputs an ES will issue a purchase order to suppliers for component parts. Transactions make changes to data stored in the ES, e.g., by recording the new purchase order and the quantities being purchased, and also recording when and by whom it was executed. As a consequence of the embedding of organizational routines, the routines performed by our respondents differed substantially from the routines they performed before the ES implementation. Some changes took the form of substitution. For example, certain routines in the shipping area, such as the preparation of paperwork to accompany parts going out for repair, were no longer performed by shipping employees. Instead, the documents were automatically generated by transactions performed by inspectors deciding to send a part for repair. Thus, embedding the routine automated it. Other changes did not replace, but instead altered the routine. For example, material buyer planners still performed a routine to order inventory, but the way they did this was new. The embedded transactions dictated a certain set of steps that were different from the steps in the old routine. Such changes in routines were generally expected by employees as part of using a new IT, and while the new routines had to be learned, they often significantly improved work practices. A less welcome change in work routines was a loss of flexibility arising from the reduced variety in the ways each routine could be performed. Each step of an embedded routine must be specified in advance, and

839

the number of alternatives is limited. This is necessary not only to avoid excess complexity, but also to support standardization, which is an expected outcome of implementing an ES. For example, ACRO’s management chose to implement its ES as a single system across the company to increase its ability to manage the organization. Thus, the same embedded routines were used throughout the organization, e.g., product assembly and new product development used the same inventory handling transactions. For product assembly, the revisions to their routines were within the range of expected changes, but in new product development, employees reported that the routines available to them were overly restrictive and interfered with their tasks. As a development engineer and a planner supporting development noted: In development there is a lot of change. The system is rigid, it’s not flexible enough to take a lot of the change and in the time frame that’s required. It doesn’t offer the flexibility for an engineering type business that doesn’t do the same job each and every day.

In addition to embedding individual routines, the ES embedded work sequencing, that is, the order in which routines could and should be performed and the connections between routines. In this way, the ES enabled process integration, which is achieved when different organizational units (and individual actors within those units) can work together in a seamless fashion with minimal intervention or delay (Gattiker and Goodhue 2005, Lawrence and Lorsch 1967, Volkoff et al. 2005). When integrated, not only do the individual routines become tightly coupled, but the chains of linked routines also become longer. The necessary inputs are available in the right formats, and the outputs are automatically sent to where they are needed, i.e., the ES automatically synchronizes the transactions producing outputs with those using them. Many of the embedded work sequences were not new, i.e., they were already the espoused form of the work sequence, but before the ES, employees had more flexibility, which they often used to work on another activity when the next activity in their work sequence was unexpectedly delayed or to vary the routine to accommodate differences because of location or novel circumstances. The new ES removed this flexibility in work practices because embedded transactions were performed the same way every time they were executed. Many managers and some other employees wanted and welcomed a more disciplined adherence to process sequencing, but others found that the ES-enforced sequencing of routines resulted in a loss of flexibility and interfered with their work. As one manager noted: “SAP has driven complete discipline into our processes.” Embedding of routines and routine sequences in the technology happens prior to use and creates a “material”

840


aspect to them. Although ES vendors design and implement transactions to represent “best practices” and organizations tailor them by configuring the ES to represent how routines “should be” performed in that organization (Davenport 1998), this material aspect is qualitatively different from the ostensive aspect. Whereas the ostensive aspect of a routine is idealized and abstract— it reflects the general principle underlying the routine (even if that general principle is interpreted differently by different people) – the material aspect is concrete and specific. Because it consists of transactions hard-coded into the system, it is the same for everyone, and individual interpretations do not affect how transactions are performed. At the same time, the material aspect is not the same as the performative aspect because the transaction is not performed but is executed by the technology. Whereas the performative aspect depends on the choices of the actor, the material aspect is predefined, and there are no choices to be made. That said, the material aspect attempts to capture the ostensive in its coding, and, like the performative aspect, executes particular instances of the routine. By tying both the ostensive and performative aspects to the material aspect, ES technology has the effect of bringing the ostensive and the performative closer to each other, which instills discipline into work practices. Embedded Roles Other organizational elements are also embedded into technology and exhibit ostensive, performative, and material aspects. For example, during an ES implementation, roles are defined and embedded; i.e., in the ES, every user ID has an assigned role. At ACRO, over 20,000 user IDs were mapped into one to two thousand roles. Furthermore, each role had a fixed set of ES transactions it was authorized to perform. These user ID-tied transaction authorizations were the role’s material aspect, in contrast to each role’s ostensive aspect—a general understanding of the responsibilities assigned to each role—and its performative aspect—the responsibilities assumed by individuals through their actions. The embedding of roles and the assignment of transactions to them had several direct consequences. First, individuals could not operate outside their roles—a significant change from the prior practice of individuals with experience helping out wherever needed. Even the weekend shift plant supervisor could not create purchase orders, but had to call in a materials buyer planner. ACRO has many employees with 20 plus years of experience in a variety of related roles, but this experience became unavailable in everyday work practices, which also slowed operations when an authorized individual was not available. Second, task assignment was no longer a dynamic management decision that enabled flexible responses to dynamic workloads, but a static assignment embedded in the ES. On the other hand,


having static and well-defined authority was consistent with management and accounting principles that emphasized clear and limited employee responsibilities. As one development planner noted: Sometimes we have to get parts from the production side of the house and we don’t have that access anymore. The materials people make the decision whether or not. Now that’s always been in place, it’s just that we were able to do it. We used to be able to go around the system. SAP does not allow you to do that anymore. So our hands are tied in that aspect.

As a consequence of the embedded assignment of transactions to roles, employees experienced role changes because the content of a role changed, which contributed further to the loss of organizational expertise. Some people were assigned to perform tasks they had not done before, and preferred not to do, while others lost tasks they had performed in the past. For example, design engineers were expected to complete engineering change documents, previously the responsibility of process engineers, but they lost responsibility for ordering parts for new products. As one process engineer noted: In the old paper world we could work their [change order] independent of any outside influence. Now with the new system, that is gone. In order to work the [change order], the electronic feed from the engineering system has to be made for the change or else the process engineer cannot work that change.

Not surprisingly, with such changes, workloads changed—we were frequently told that the material buyer planner role had become much more onerous, but the shipping role involved less work. The combined consequences of embedding both roles and work sequences had a paradoxical effect. Embedding work sequences promoted process integration, which worked to reduce borders and create a seamless pan-organizational environment, which in turn created a need for role integration. That is, for integrated processes to work effectively, employees needed to work as though they were part of a seamless process. The embedded roles, on the other hand, strictly and narrowly defined role authority, heightening role differentiation and boundaries and making it difficult for employees to work as though they were in a seamless and borderless environment. As one employee noted: There should be some safeguards, but if you put so many, then it doesn’t help. And what that does by making it smaller and smaller, your scope of what you do makes very few people that understand the whole system. The only way an ERP system, in my opinion, that you can make this work, is everybody has to be smarter.

For example, data from part inspection flowed seamlessly to the receiving area, producing an integrated process that automatically matched repair request data


generated by inspectors to paperwork accompanying repaired parts arriving in receiving from repair vendors. Such integration was expected to speed up receiving by automating verification, and to improve control by insuring verification. Unfortunately, if either the inspectors or the vendors entered erroneous transaction data, the match would fail and the part could not be received, even if desperately needed. Yet because of role differentiation, even an obvious or trivial error could not be corrected by receivers, but only by inspectors with authority to run the transactions that could change that data. The initial result was a rock pile of parts that could not be received or used in the plant. In one plant, these issues were resolved over several months as receiving took responsibility for working with inspection to resolve problems and to find methods for avoiding them. In another plant, they worked around the problem by ordering more parts, resulting in significant inventory cost increases. Thus, although the consequences of embedding work sequences was more integration of data and processes, the role differentiation arising from embedded roles resulted in fewer or even negative productivity gains from integration. As one employee noted, If I don’t do this correctly, or if it wasn’t set up correctly, how do I correct it so I can do the transaction?, and then if I don’t do it correctly, who is it going to affect? We’re definitely going to have to work with each other and we can’t be independent like we are today.

Another way ACRO resolved this tension between narrow, well-defined roles and the need for integrated work across roles was through superusers who had the authority to perform all ES roles. While superusers were intended as temporary roles to handle problems until users learned the system well, ACRO was still using superusers well after go-live because of the difficulty of coordinating across roles to solve problems that could be easily resolved by one person with wider authority. While a few well-trained superusers were unlikely to create problems, providing all users with broad authority (i.e., integrating roles) would lead to a loss of control and create the potential for chaos. Embedded Data The third organizational element changed through being embedded in the ES was data which were stored in a computer and made available to computer users. When embedded, data take on a material aspect related to the level of detail and format and how they are manipulated or transformed, presented to users, and changed. The ostensive aspect of a piece of data is the meaning users ascribe to it, and the expectations of why it has been collected and what it will be used for. The performative aspects of data include the specific uses made of the data by different users. While the legacy systems had embedded data, the ES made substantial changes to the

841

embedding of data. First, the ES generally embedded more data, which were often tracked at a greater level of detail. For example, the ES tracked who did what and when, and this was available to anyone with access. Second, the ES changed when and how frequently data were collected or updated, with much more data available in real-time. For example, in the legacy systems, raw data were transformed into accounting data once a month when operational data were sent to accounting for roll-up and reporting. In the ES, controllers could and did run daily trial balance reports. Furthermore, on-demand trial balances could be run at the level of individual orders moving through the plant. As one plant controller noted: And if it’s over your threshold, you can go to the floor very quickly and say, “You were [x] times over on your hours. I need to know what happened.”

Third, the ES changed the location of embedded data by storing them in a common database, rather than in local legacy systems. The direct consequence of this change to the material aspect of the data was to provide broad visibility to operational data. The increase in visibility changed role content by empowering users. For example, because customer service representatives had access to financial and production data in the ES, they knew whether a customer had been invoiced as well as the progress of a customer’s order through the plant; inventory planners had access to inventory data in other plants. Before the ES, these data were locally embedded in finance and manufacturing, and customer service representatives had to rely on whatever information manufacturing might provide to them about the status of an order, and inventory planners had to rely on inventory information from another plant. As one customer service representative noted about the ES: In the impact on customer feedback, questions that are coming in from the customers, today we have a lot more capability to get information with a lot less effort than we did before. What you think would be fairly simple things, but back when we used [old system], it might not have been so simple.

At the same time that this material aspect of the data empowered workers, it also changed control. Not only managers, but peers had the information needed to exercise control over other workers in an integrated process. Thus, workers in receiving exercised control over inspectors because receivers were able to find and report inspectors’ mistakes. The changes in empowerment and control that we observed were attributable to embedding detailed, real-time data about work activities in a common database. Another consequence of embedding data in a common database was the need to standardize the material form of the data, in terms of its definition, format, and level of detail. For example, at ACRO finance and production

842


were asked to select a common definition for the location of a part, resulting in defining it in terms of which group had responsibility for it, rather than the definition favored by production, the physical location of the part. The consequence was that finance always knew who was responsible for every part, but production spent more time searching for parts. In addition to embedding data, the ES also embedded the relationship between transactions and data, prohibiting transactions from being executed if the required data were not in the common database. For example, a repair order could not be placed until the specific part needing repair and the repair price were specified. While requiring the part number was reasonable, the price was typically not known until the repair process had assessed the extent of work needed. Parts could not be ordered by the material buyer planners if a blueprint containing those parts was not finalized by engineering, even though the need for those parts was obvious. As one planner in development noted: If the part number doesn’t exist in the database, you can’t order it. If the blueprint is not available, you can not receive the part into stores. That was always the rule, but we were able to get around that. We would supply preliminary blueprints, or hand carry a marked up blueprint down to the inspector, so he could inspect the part and receive it. in SAP it’s very strict, everything runs in series.

Further Consequences of Embeddedness As we have discussed, not only does an ES embed individual organizational elements, i.e., individual routines in the form of transactions, individual roles, and individual data elements, it also embeds the relationships between organizational elements, i.e., the relationships among routines in the form of work sequences, the relationships between transactions and roles, and the relationships between data and transactions. Each of these embeddings takes on a material aspect that affects employees as they perform their work activities. With so many organizational elements and their relationships embedded in the ES, employees are no longer able to pick and choose routines and roles to perform and data to use. Much of their work is prescribed by the material aspects of organizational elements, and thus, employees face a different environment. The consequences of embedding task sequences, the relationship between transactions and data, narrowly defined roles, and visibility to data were high costs or sanctions for mistakes. Erroneous transaction data, e.g., from inspectors, immediately flowed though the process and everyone knew the source of the poor data. Furthermore, the narrowly defined roles meant that reversing the problem required all workers in the process to back out any of their processing that used that data. Thus, the ES served to embed sanctions for poor work


performance and for any performance of routines that was inconsistent with ES-embedded work practices. As a consequence of these implicitly embedded sanctions, a culture of disciplined work emerged as workers came to recognize that the ES only worked smoothly when they performed their work routines with strict discipline. As a development material planner noted: Everybody now realizes what tasks have to be done before you move on, so now we’ve learned. And we’ve learned that if you try to do the end-around, it just adds more time to your process, so do it right the first time.

Before the ES, individuals established a sense of expertise and power by being able and willing to perform a variety of tasks and using these skills to overcome unexpected obstacles through creative solutions. With the ES, honor accrued to those individuals who best understood proper use of the ES. The ability to work around the system, a respected competency in the legacy system environment, was considered disruptive. Heroic measures to ship product on time were no longer valued; a mindset gradually developed that if the work practices were followed with discipline, product was more likely to ship on time than under the old practices. As one materials planner noted: It’s just the mindset is different and you have to make sure that you do what you’re supposed to do. Everything is out in the open.

Summary The effects we observed from the ES implementation were caused by embedding organizational elements and their relationships in the ES. These effects were not, however, fully determined by that embedding. For example, in one plant, the response to the rock pile of parts in receiving was to work toward understanding the cause of the problems and resolving them, whereas in another plant, the solution was to order more parts. Some employees found the visibility to more data empowering and made use of that data to perform their jobs better or to understand and enhance the performance of integrated processes, while others stayed within the narrowly defined roles prescribed by the ES. Overall, by embedding routines and their sequencing into the ES, as well as roles, relationships between roles and transactions, data, relationships between data and transactions, and relationships among all these embedded elements, employees found it necessary to perform their work according to the organizational element’s material aspect which was embedded in the ES. This moved the performative aspect of each organizational element closer to its ostensive aspect.

A Theory of Technology-Mediated Organizational Change

As discussed in the introduction, our objective is to build a midrange process theory of organizational change.


843


Unlike variance models, process models do not identify causal factors and predict outcomes, but instead explain how the process unfolds, identifying the mechanisms that move it along. The concept of embeddedness is at the core of our explanation of how technology-mediated organizational change unfolds. While the term embeddedness has been used in various ways,1 we use it to describe the way in which technology introduces a material aspect to organizational elements such as routines, roles, and data. While the material aspect (e.g., an electronic transaction or a user ID-linked authorization) is physically manifested in the technology, it is more than simply an artifact or part of the technological context. It is an integral part of the associated organizational element and is the key to the change process. When organizational elements such as routines, roles, and data become embedded in technology, their material aspect interacts with and affects their ostensive and performative aspects. Understanding that organizational elements such as routines, roles, and data have material as well as ostensive and performative aspects is necessary but not Figure 1

sufficient for building a theory of technology-mediated organizational change. In our view, it is also necessary to move from typical constructivist perspectives such as structuration theory to an approach based on critical realism so as to incorporate the temporal aspect of the change process and to address the materiality of the technology. Unlike structuration theory, which examines structure and agency and their relationship at particular moments of instantiation, critical realism insists that we consider a time line that distinguishes among prior structural conditions, social interaction, and poststructural conditions. This allows us to propose a theory of technology-mediated organizational change, shown in Figure 1. For purposes of clarity during the exposition of the theory, we use routines as an example of an organizational element. The comments apply equally, however, to other organizational elements such as roles and data that become embedded in the technology. Structural Conditioning The routines enacted through social interaction during ES use do not construct the system—the system exists

A Critical Realist Theory of Technology-Mediated Organizational Change Structural conditioning

Social interaction

Ostensive aspects (O): Organizational structures as understood in principle by designers Routines as understood (ON)

Ostensive aspects (O): Org structures as understood by users

Structural elaboration/reproduction

Performative aspects (P): Org structures as performed by users

od od od od od

on′ on′ on′ on′ on′

od od od od od

PX

or or or or or (2)

Material aspects (M): Technology under design and construction

(5)

(4)

Embedded transactions (MT)

mt mt mt

Embedded data structures (MS)

Embedded transactions (MT′) mt′ mt′ mt′

mt mt mt ms ms ms

ms ms ms

Embedded authorizations (MA)

MX

ma ma ma

Embedded authorizations (MA′)

Summary:

O

P Summary:

Time O: Ostensive aspect P: Performative aspect M: Material aspect

N: Routines D: Data R: Roles

MX′

ma′ ma′ ma′

ma ma ma

M

Embedded data structures (MS′) ms′ ms′ ms′

MX

M

Leads to Interaction Integration

(6)

Material aspects (M): Technology under revision

Material aspects (M): Technology as built

Embedded data structures (MS)

or′or ′ or′ or′ or′

pr pr pr pr pr

(1)

Key

OX′

Roles as understood (OR′)

or or or or or

O

od′ od′ od′ od′ od′

pn pn pn pn pn

OX

Embedded authorizations (MA)

Data as understood (OD′)

(3)

on on on on on

OX

Roles as understood (OR)

Summary:

Routines as understood (ON′)

pd pd pd pd pd

Data as understood (OD)

on on on on on

Embedded transactions (MT)

Ostensive aspects (O): Organizational structures as understood in principal by designers and users

T: Transactions X: Other organizational elements S: Data structures A: Authorizations

O′

M′

844


prior to its use, having been designed and built by a vendor, and configured by an implementation team. These design and configuration activities create the prior structural conditions within which use eventually occurs. The design and construction/configuration of technology consists of articulating and explicating in detail how each routine should operate (i.e., the ostensive aspects of the routine, shown as ON in Figure 1) and then turning this understanding into embedded transactions realized as lines of code (i.e., the material aspect of the routine, shown as MT). During that process, designers collect varying interpretations of a given routine, decide on the best one or ones, and encode this choice into the software. For example, consider a routine for issuing a purchase order. While several variations of this routine may be built into the ES (i.e., different routines may be established for purchasing a part versus purchasing a repair service), the number of alternatives that can be included is finite. While the ostensive aspect directly guides the development of the material aspect, technical considerations may limit or extend what is possible, i.e., the material aspect may influence the ostensive aspect of the routine and vice versa (shown in Figure 1 as Arrow 1 between the ostensive aspects, O and the material aspects, M). Furthermore, technology such as ESs has predefined versions of the routines (MT) built into it. Vendors and consultants regularly urge purchasers of such systems to conform to these built-in best practices and thus, the material aspect may again influence the ostensive. While the ostensive and material aspects of a routine interact during design and construction/configuration, its performative aspect is latent; the performance of the technology-mediated routine cannot occur until the technology is implemented. The objective of design and configuration activities is to bring the ostensive and material aspects into alignment to the extent possible. The outcome of the interaction between the ostensive and material aspects of the technology in our theory marks the prior structural conditions within which performance of the routine can eventually take place. As noted by Feldman and Pentland (2003), the ostensive aspects are interpreted differently by different people. During structural conditioning, the ostensive aspects of organizational elements as interpreted by those responsible for design and configuration are put into play. Social Interaction Once the technology is implemented, users start to perform the organizational routines. This social interaction phase, during which changes to organizational elements become evident, follows and is influenced by the structural conditioning phase (Arrow 2). In the social interaction phase, the performative aspect of routines (shown as PN in Figure 1) becomes active as users perform those routines. The performative and ostensive aspects


interact and influence each other as discussed by Feldman and Pentland (2003), but in our theory that interaction (Arrow 3) is constrained/enabled and moderated by the material aspects (Arrow 4). Note that during social interaction it is the users’ interpretation rather than the designers’ interpretation of the ostensive aspect that is salient. In general, while ostensive aspect can guide actors, and enable them to account for or refer to their performances, and the performative aspect can help to create, maintain, and modify the ostensive aspect, the material aspect of the software circumscribe these interactions. For example, as individuals perform the purchase order routine, their understanding of the purchase order process in principle will guide them, but the system-embedded transactions that allow a purchase order to be submitted will constrain them. During their work, they may encounter situations that do not fit well with the routines and work sequences as designed. For example, the R&D group may want to place orders for parts in products not yet fully designed. At ACRO in the ES as designed, purchase orders could only be issued for items that had fully specified bills of material. Thus, the material aspect constrained the performative. The workers did identify some creative workarounds such as setting up bills of material for dummy items, but even these workarounds had to conform to the systemimposed constraints. Because the material aspect both frames how people interpret the ostensive aspect and influences how they perform it, the performative and the ostensive move closer together—that is, as people are required to behave in a more disciplined manner, their performance becomes more closely aligned with the routine in principle. During this social-interaction stage, the ostensive and performative aspects do not affect the material aspect—it is static, as systems such as these, once adopted, cannot be easily altered. Structural Elaboration/Reproduction Social interaction (i.e., the execution of the routines) will lead to either reproduction of those routines, or elaboration of them (Arrow 5). To the extent that the constraints imposed by the material aspects are in conflict with the ostensive aspect, people may redefine their interpretation of the right way to do things to conform with the embedded versions, i.e., change the ostensive aspect. Alternatively, they may want to change the technology. The nature of enterprise systems is such that changes to the technology cannot be made through casual, ad hoc adjustments by individuals as they work. Instead, a formal project must be initiated to change the material aspect. Thus, after the social interaction inherent in the performance of routines, there is the possibility of structural elaboration or reproduction, the next stage of our theory, where the ostensive and material aspects are either confirmed or redefined through further interaction (Arrow 6).


The Change Process Overall, the change process is not one of continuous, gradual evolution, as suggested by constructivist approaches that focus on individual actions. Instead, it is a set of alternating cycles where the ostensive and the material aspects become mostly aligned during design and construction, enabling significant changes to organizational elements over short periods of time, followed by periods where the performative and ostensive aspects interact, but are constrained by the material aspect, leading to new interactions between the material and the ostensive. A parallel discussion could be presented with respect to roles (the three aspects of roles are shown as OR, MA, and PR in Figure 1), data (shown as OD, MS, and PD), or the embedded relationships between them (shown as the small arrows within and between the boxes that represent different organizational elements). Similarly, there could be other organizational elements (e.g., communication channels) that become embedded in the technology and acquire a material aspect. In Figure 1, we indicate these other elements as OX, MX, and PX. Finally, while the relationships between the material aspects of the different organizational elements are still in flux during design and construction, after implementation these relationships are fixed in the technology as built. Moreover, for a technology such as an ES, it is these fixed embedded relationships that define integration (the straight lines in Figure 1). While the performative aspects of such relationships are still under some degree of negotiation, they are strongly influenced by the fixed embedded relationships (the long curved arrows). Comparison to Other Theoretical Approaches This use of critical realism, the introduction of the concept of embeddedness, and analysis at the level of individual organizational elements gives our midlevel process theory of how technology-mediated change unfolds a number of advantages over traditional models of organizational change. In particular, it provides a temporal separation that distinguishes between two types of processes, namely, a process of negotiation between the ostensive and material aspects, where the performative aspects are latent, and a separate process of negotiation between the ostensive and the performative aspects, moderated by the material aspects. Under structuration theory, change only occurs through an interaction of performance and structure. Because the material aspect is not distinguished from the ostensive aspect in the concept of structure, the technology loses salience and the effects of different types of technology cannot be examined. ANT does allow the technology to become an independent actant in the change process, but because it is treated in the same way as every other actant, the two different processes that occur cannot be distinguished. Institutional theory tends to ignore the performative aspects altogether.

845

Examining organizational change at the level of organizational elements such as routines, roles, and data, combined with the temporality of critical realism has further advantages. First, only at the level of organizational elements can we discuss the distinctions among the material, the performative, and the ostensive aspects. Moreover, different elements are likely to have different temporal cycles in our theory. For example, organizational routines are likely to change on the same time cycle as system upgrades. Roles, which are embedded through authorizations, might be changeable over a shorter time cycle. By contrast, data structures, once designed, are unlikely to change for a long period. Focusing on organizational elements also allows the examination of differential effects such as our discovery that the introduction of an ES serves to integrate routines and data, but differentiate roles.

Conclusion

This study developed a new theory of technology-mediated organizational change. Our theory emerged as a result of employing grounded theory techniques to analyze data collected during an intensive, longitudinal case study of an organization implementing an ES. Our data told us that technology-mediated organizational change should be examined at the level of organizational elements, namely, as changes in, for example, organizational routines, roles, and data, and as changes in the relationships between these elements. The changes observed were the starting point for understanding how the technology served as the source or cause of these changes. The core theoretical concept emerging from our data was embeddedness. Embedding a routine, role, or data in the ES changed that element and how it could be enacted by employees. As we coded our field data, constantly comparing coded passages to derive these theoretical concepts, we also investigated existing literature and found two research areas relevant to our emerging theory: critical realism and organizational routines. Critical realism, with its theoretical emphasis on objects existing in the real domain, supported the view of technology emerging from our data—specifically that technology has an inherent materiality that influences work practices, rather than technology as an object purely constructed through subjective interpretation. Critical realism also contributed to how we included temporality in our theory. Research on organizational routines, especially Feldman and Pentland (2003), supported the view emerging from our data that organizational change occurred through changes in organizational elements, such as organizational routines and roles. This research also provided the concepts of ostensive and performative aspects of routines, which we applied more broadly to other organizational elements in our theory.

846


Our theory contributes to research at the intersection of organizations and technology in several ways. First, our theory of technology-mediated organizational change explicitly includes technology. Technology’s role is to embed organizational elements, which then have a material aspect that affects how employees are able to enact and interact with those organizational elements. This view of technology differs from the role of technology in ANT, in which it is just another actor in a negotiation process. It also differs from the role of technology in structuration theory, in which technology is “interpretively flexible” (Orlikowski 1992, p. 405), enacted and defined as part of structure in the instance of use by agents. Second, our theory of technology-mediated organizational change covers a number of organizational elements and their relationships, including routines, roles, data, relationships between routines in the form of sequences of routines, relationships between routines and roles, and relationships between routines and data. This extends the organizational research on organizational routines as sources of organizational change to other organizational elements. In addition, our theory applies the concepts of ostensive and performative aspects of organizational elements—first articulated by Latour (1986) for power, and then applied by Feldman and Pentland (2003) to organizational routines—to other organizational elements, including roles and data. This provides the basis for introducing the concept of embeddedness, and allows us to analyze how the material aspect of organizational elements influences and is influenced by the ostensive and performative aspects of these elements, as well as moderating the interaction between them as employees interact with the technology to perform their work. Third, our theory of technology-mediated organizational change is an explanatory process theory, answering calls for such work (Robey and Boudreau 1999). Our theory not only explains how embedding organizational elements into a technology leads to changes in both the ostensive and performative aspects of those elements, and thus to changes in organizational form and function, but it also has an explicit temporal dimension. The temporal dimension consists of the three critical realist stages identified by Archer (1995): structural conditioning, social interaction, and structural elaboration/reproduction. Within these stages, the process of change, i.e., how the material, ostensive, and performative aspects of organizational elements interact, was identified from our data. These stages and the interactions within each provide an understanding of how and why implementing a technology leads to organizational change. This approach differs from theories taking a structuration theory perspective, which explain organizational change in terms of the interactions between structure and agency, but are less likely to explain how these interactions play out over time.


Finally, our theory provides an example of a theory developed from a critical realist perspective, which to our knowledge is the first such example in IT research. This addresses the claims of several researchers that critical realism would provide a better lens for technologymediated organizational change theory development than structuration theory or ANT (Dobson 2001; Mingers 2002, 2004; Mutch 2002). Our theory of technology-mediated organizational change also contributes to practice. With organizational systems like ESs installed in most medium and large organizations, understanding technology-mediated organizational change is becoming a required skill for successful managers. Our theory suggests that they need to consider more than just the planned changes to data and functionality. Such systems are likely to affect a broader range of organizational elements, such as roles, relations between data and routines or routines and roles, forms of control and mindset. In addition, our theory helps managers understand how organizational elements differ in their changeability and their time cycle for changing. Because our theory of technology-mediated organizational change is expressed in general theoretical terms, it can be applied to technologies other than the ES we studied. To consider other technologies, one must determine which organizational elements are embedded in the technology and examine their material aspects. Those material aspects will vary in their effects on the ostensive and performative aspects of those organizational elements. For example, the material aspect may be highly flexible and malleable or its use may be optional, resulting in fewer constraints on users of the technology than what we observed at our field site. Furthermore, the effect of the material aspect on the interaction between the ostensive and performative aspects during technology use may lead to second-order effects other than the ones we observed on control and mindset. However, we do not yet have empirical evidence that our theory is applicable to technologies other than the ES we studied and, thus, cannot attest to the generalizability of our theory. The ES we studied, however, is the market leader and thus our theory is likely to apply to the other installations of SAP or similar ESs. Similarly, our intensive, longitudinal case study method required a focus on a single organization, limiting the generalizability of our results beyond that organization. We compensated as much as possible for this by using theoretical sampling to cover different locations and functional areas of the organization, ranging from a highly structured production environment to a much less structured R&D environment. In this paper we develop, rather than test, theory. The next step is to test our theory’s ability to explain technology-mediated organizational change in other organizations and with other technologies, both from a research and practice perspective. As we and other


researchers do this, our theory, like other theory, is likely to evolve in various ways that we cannot predict. Another avenue for further research is to investigate links between our theory and other organizational theories. In particular, our theory of IT-mediated organizational change might be enriched by considering the extent to which embedding organizational elements in technology requires a shift in the balance between mindful and less mindful behavior (Elmes et al. 2005, Langer 1989, Levinthal and Rerup 2006, Weick and Sutcliffe 2006). Finally, we urge researchers who are interested in the technologyorganization nexus to consider using the lens of critical realism, which proved to be an invaluable perspective from which to develop new theory consistent with a grounded theory methodology. Acknowledgments

This research was supported in part by the National Science Foundation under Grant 0114954. Any opinions, findings, and conclusions or recommendations expressed in this paper are those of the authors and do not necessarily reflect the views of the National Science Foundation. The authors thank the organization we studied for generously providing us with open access to observe their implementation process, attend meetings, and interview employees, at a time when they were extremely busy. The authors also thank the many employees who willingly talked about their experiences with the newly implemented ES.

Endnote 1

847

Boudreau, M.-C., D. Robey. 2005. Enacting integrated information technology: A human agency perspective. Organ. Sci. 16(1) 3–18. Collier, A. 1994. Critical Realism: An Introduction to Roy Bhaskar’s Philosophy. Verso, London, UK. Davenport, T. H. 1998. Putting the enterprise into the enterprise systems. Harvard Bus. Rev. 76(4) 121–131. Dobson, P. J. 2001. The philosophy of critical realism—An opportunity for information systems research. Inform. Systems Frontiers 3(2) 199–210. Elmes, M. B., D. M. Strong, O. Volkoff. 2005. Panoptic empowerment and reflective conformity in enterprise systems-enabled organizations. Inform. Organ. 15(1) 1–37. Feldman, M. S. 2000. Organizational routines as a source of continuous change. Organ. Sci. 11(6) 611–629. Feldman, M. S. 2003. A performative perspective on stability and change in organizational routines. Indust. Corporate Change 12(4) 727–752. Feldman, M. S., B. T. Pentland. 2003. Reconceptualizing organizational routines as a source of flexibility and change. Admin. Sci. Quart. 48(1) 94–118. Gattiker, T. F., D. L. Goodhue. 2005. What happens after ERP implementation: Understanding the impact of interdependence and differentiation on plant-level outcomes. MIS Quart. 29(3) 559–585. Giddens, A. 1984. The Constitution of Society. University of California Press, Berkeley and Los Angeles, CA. Glaser, B. G. 1978. Theoretical Sensitivity. Sociology Press, Mill Valley, CA. Glaser, B. G., A. Strauss. 1967. The Discovery of Grounded Theory: Strategies for Qualitative Research. Aldine, New York.

The term “embeddedness” is often used to describe personal relations and positioning within social networks and their effects on economic behavior (Granovetter 1985, 1992). That is not the sense used in this paper.

Gosain, S. 2004. Enterprise information systems as objects and carriers of institutional forces: The new iron cage? J. Assoc. Inform. Systems 5(4) 151–182.

References

Granovetter, M. 1992. Problems of explanation in economic sociology. N. Nohria, R. G. Eccles, eds. Networks and Organizations: Structure, Form, and Action. Harvard Business School Press, Boston, MA, 25–56.

Archer, M. S. 1995. Realist Social Theory: The Morphogenetic Approach. Cambridge University Press, Cambridge, UK. Archer, M. S. 1998. Introduction: Realism in the social sciences. M. S. Archer, R. Bhaskar, A. Collier, T. Lawson, A. Norrie, eds. Critical Realism: Essential Readings. Routledge, London, UK, 189–205. Avgerou, C. 2000. IT and organizational change: An institutionalist perspective. Inform. Tech. People 13(4) 234–262. Barley, S. R. 1986. Technology as an occasion for structuring: Evidence from observations of CT scanners and the social order of radiology departments. Admin. Sci. Quart. 31(1) 78–108. Barley, S. R. 1990. The alignment of technology and structure through roles and networks. Admin. Sci. Quart. 35(1) 61–103. Becker, M. C. 2004. Organizational routines: A review of the literature. Indust. Corporate Change 13(4) 643–677. Bhaskar, R. 1978. A Realist Theory of Science. The Harvester Press, Brighton, Sussex, UK. Bhaskar, R. 1979. The Possibility of Naturalism. The Harvester Press, Brighton, Sussex, UK. Bhaskar, R. 1998. General introduction. M. S. Archer, R. Bhaskar, A. Collier, T. Lawson, A. Norrie, eds. Critical Realism: Essential Readings. Routledge, London, UK, ix–xxiv.

Granovetter, M. 1985. Economic action and social structure: The problem of embeddedness. Amer. J. Sociol. 91(3) 481–510.

Hanseth, O., M. Aanestad, M. Berg. 2004. Actor-network theory and information systems. What’s so special? Inform. Tech. People 17(2) 116–123. Howcroft, D., N. Mitev, M. Wilson. 2004. What we may learn from the social shaping of technology approach. J. Mingers, L. Willcocks, eds. Social Theory and Philosophy for Information Systems. John Wiley & Sons Ltd., Chichester, UK, 329–371. Jones, M. 1999. Structuration theory. W. Currie, B. Galliers, eds. Rethinking Management Information Systems: An Interdisciplinary Perspective. Oxford University Press, New York, 103–135. Klaus, K., M. Rosemann, G. Gable. 2000. What is ERP? Inform. Systems Frontiers 2(2) 141–162. Langer, E. J. 1989. Mindfulness. Da Capo Press, Reading, MA. Latour, B. 1986. The powers of association. J. Law, ed. Power, Action and Belief. Routledge & Kegan Paul, London, UK, 264–280. Latour, B. 1996. Social theory and the study of computerized work sites. W. J. Orlikowski, G. Walsham, M. R. Jones, J. I. DeGross, eds. Information Technology and Changes in Organizational Work. Chapman & Hall, London, UK, 295–307.

848


Law, J. 1992. Notes on the theory of the actor-network: Ordering, strategy, and heterogeneity. Systems Practice 5(4) 379–393. Lawrence, P. R., J. W. Lorsch. 1967. Organization and Environment: Managing Differentiation and Integration. Harvard University Press, Boston, MA.


Orlikowski, W. J. 2000. Using technology and constituting structures: A practice lens for studying technology in organizations. Organ. Sci. 11(4) 404–428. Orlikowski, W. J., D. Robey. 1991. Information technology and the structuring of organizations. Inform. Systems Res. 2(2) 143–169.

Levinthal, D., C. Rerup. 2006. Crossing an apparent chasm: Bridging mindful and less-mindful perspectives on organizational learning. Organ. Sci. 17(4) 502–513.

Outhwaite, W. 1998. Realism and social science. M. S. Archer, R. Bhaskar, A. Collier, T. Lawson, A. Norrie, eds. Critical Realism: Essential Readings. Routledge, London, UK, 282–296.

Locke, K. 2001. Grounded Theory in Management Research. Sage, Thousand Oaks, CA.

Pentland, B. T., M. S. Feldman. 2005. Organizational routines as a unit of analysis. Indust. Corporate Change 14(5) 793–815.

Markus, M. L., C. Tanis. 2000. The enterprise system experience— From adoption to success. R. W. Zmud, ed. Framing the Domains of IT Management: Projecting the Future Through the Past. Pinnaflex Educational Resources Inc., Cincinnati, OH, 173–207.

Reed, M. I. 1997. In praise of duality and dualism: Rethinking agency and structure in organizational analysis. Organ. Stud. 18(1) 21–42.

Miles, M. B., M. A. Huberman. 1994. Qualitative Data Analysis, 2nd ed. Sage, Thousand Oaks, CA. Mingers, J. 2002. Real-izing information systems: Critical realism as an underpinning philosophy for information systems. TwentyThird Internat. Conf. Information Systems, Barcelona, Spain, 295–303. Mingers, J. 2004. Re-establishing the real: Critical realism and information systems. J. Mingers, L. Willcocks, eds. Social Theory and Philosophy for Inform. Systems. John Wiley & Sons Ltd., Chichester, UK, 372–406. Monteiro, E., O. Hanseth. 1996. Social shaping of information infrastructure: On being specific about the technology. W. J. Orlikowski, G. Walsham, M. Jones, R. J. I. DeGross, eds. Information Technology and Changes in Organizational Work. Chapman & Hall, London, UK, 325–343. Mutch, A. 2002. Actors and networks or agents and structures: Towards a realist view of information systems. Organization 9(3) 477–496.

Robey, D., M. Boudreau. 1999. Accounting for the contradictory organizational consequences of information technology: Theoretical directions and methodological implications. Inform. Systems Res. 10(2) 167–185. Strauss, A., J. Corbin. 1998. Basics of Qualitative Research: Techniques and Procedures for Developing Grounded Theory, 2nd ed. Sage Publications, Thousand Oaks, CA. Suddaby, R. 2006. From the editors: What grounded theory is not. Acad. Management J. 49(4) 633–642. Urquhart, C. 1997. Exploring analyst-client communication: Using grounded theory techniques to investigate interaction in informal requirements gathering. A. S. Lee, J. Liebenau, J. I. DeGross, eds. Information Systems and Qualitative Research. Chapman & Hall, London, UK, 149–181. Volkoff, O., D. M. Strong, M. B. Elmes. 2005. Understanding enterprise systems-enabled integration. Eur. J. Inform. Systems 14(2) 110–120.

Nelson, R. R., S. G. Winter. 1982. An Evolutionary Theory of Economic Change. Harvard University Press, Cambridge, MA.

Walsham, G. 1997. Actor-network theory and IS research: Current status and future prospects. A. S. Lee, J. Liebenau, J. I. DeGross, eds. Information Systems and Qualitative Research. Chapman & Hall, London, UK, 466–480.

Orlikowski, W. J. 1992. The duality of technology: Rethinking the concept of technology in organizations. Organ. Sci. 3(3) 398–427.

Weick, K. E., K. M. Sutcliffe. 2006. Mindfulness and the quality of organizational attention. Organ. Sci. 17(4) 514–524.