Portal de Periódicos da CAPES

Digital transformation in operations management: Fundamental change through agency reversal

2023; Wiley; Volume: 69; Issue: 6 Linguagem: Inglês

10.1002/joom.1271

1873-1317

Spyros Angelopoulos, Elliot Bendoly, Jan C. Fransoo, Kai Hoberg, Carol Xiaojuan Ou, Antti Tenhiälä,

Quality and Supply Management

The emergence of digital technologies across all aspects of operations management (OM) has enabled shifts in decision making, shaping new operational dynamics and business opportunities. The associated scholarly discussions in information systems (IS) and OM span digital manufacturing (e.g., Roscoe et al., 2019), the digitalization of OM and supply chain management (e.g., Holmström et al., 2019), platform outcomes (e.g., Friesike et al., 2019), and economies of collaboration (e.g., Hedenstierna et al., 2019). For such changes to be successful, however, there is a need for organizations to go beyond the mere adoption of digital technologies. Instead, successful changes are transformational, delving into digital transformation (DT) endeavors (Vial, 2019), which in turn can enable operational improvements in organizational performance (Davies et al., 2017), lead to structural changes in operations processes, and may result in new business models being deployed. Appropriately, DT endeavors are increasingly treated in both the IS and OM literature as an ongoing process rather than an isolated project with a clear start and finish (e.g., Struijk et al., 2022). Here, we adopt this line of reasoning and specifically treat DT endeavors as: "the use of digital technologies to evolve operational activities by creating new or transforming existing processes, cultures, and customer experiences to meet changing business and market requirements." Such a perspective is somewhat distinct from widely adopted definitions of DT in IS and OM (e.g., Vial, 2019), as well as from the strict consideration of radical operational innovation (cf. Hammer, 2004). Specifically, our perspective is neither predicated on "disruption" per se, nor limited by such transformations being fundamentally strategic ones for the focal organization. In other words, DT endeavors can (i) extend into the creation of new organizational processes, (ii) transform existing processes either incrementally or more substantially, (iii) shift decision making with regard to those processes, (iv) enable the consideration of new business models, and (v) largely serve as a source of facilitation and synergy in existing ones. In this special issue, we characterize the specific role of DT in OM as follows: through DT endeavors, digital technologies have the potential to affect OM processes and decision-making with regard to finance, design, production, and the delivery of products, services, or combinations of them. The broader OM literature has already set the stage for the consideration of new business models and innovation tournaments that have been extensively influenced by DT endeavors, such as platform services, omnichannel retail, supply chain information exchange, and Internet of Things (IoT)-enabled operations. This line of research can contribute to contemporary and ongoing discussions within the broader field (e.g., Holmström et al., 2019), including the opportunities for organizations to leverage presence in one market into other areas; the emergence of ecosystems that take into consideration all players in the value chain; the appeal of multi-sided platform business models that bring together disparate actors; the value of new data sources when serving new customers; and the importance of artificial intelligence (AI) in the form of advanced algorithmic solutions as a competitive advantage for organizations. Such scholarly discussions can further consider failures caused by the complexity and comprehensiveness of actions that organizations attempt to undertake during DT endeavors (Struijk et al., 2020, 2023). Empirical research as well as theoretical insights into DT endeavors, therefore, can challenge our established understanding of OM theory and practice, and highlight the importance of organizational dynamics as intertwined with higher levels (Struijk et al., 2022). Our aim here, thus, is to provide an epistemic platform to advance our understanding of how DT endeavors, including the adoption of digital technologies, business model innovations, and innovations in collaboration mechanisms and methods of operations improvement, can affect various aspects of OM. In the discussion that follows, we delineate a review and conceptualization of DT in OM, taking stock of the topic within the field and exploring pathways for moving forward beyond the hype. In doing so, we draw attention to a change in the relationship between humans and technology, where the roles of an agent and a principal are being reversed for the first time in the evolution of the broader IS theory and practice. Specifically, we argue that the transformative nature of DT lies in an agency reversal in many organizational processes that are affected by it. Technology evolution has been a central topic for the broader management literature, due to the transformative effect of technological change on organizations, individuals, and society at large (Grodal et al., 2023). Technology is inherent in OM theory and practice, and its role in the value-adding processes of organizations is crucial to the extent that early management theorists used the word "technology" in place of "process" when discussing what we now know as OM (Thompson, 2017). The evolution of OM, thus, has been tightly linked to the evolution of both physical technology as well as advanced IS, from the invention of the spinning jenny in the early 18th century to modern advanced algorithmic solutions. Our special issue focuses on the latter, within the context of DT and the broader consumerization of digital technologies (Gregory et al., 2018; Struijk et al., 2022). Although we use that term (DT) and argue that the contemporary forms of such technologies bear an exceptional potential for fundamental change, it is still useful to view contemporary technologies within the greater picture of the evolution of organizational IS. In doing so, we see three distinct phases in that evolution as shown in Table 1. This view departs from the idea that the contemporary digital technologies are merely linear extensions of technological evolution, in the sense that they deliver similar benefits as all of the previous technologies such as reducing the costs of data collection, storage, as well as processing, and enable faster and better decision making. Instead, we view the historical development in the role of digital technologies in OM as encompassing three major stages: stand-alone tools, integrated tools, and, contemporaneously, increasingly autonomous tools that have the potential to deliver an unprecedented change in the human-technology relationship, where DT in OM resides. We further discuss these three stages through an elaboration on the leading technologies of the time, providing a brief overview on how various digital technologies have contributed to OM practice. From the 1970s, when IBM developed the COPICS software package for MRP, until the turn of the millennium, when vendors like Manugistics and i2 marketed advanced planning and scheduling (APS) systems for integrated supply chain optimization, the field of OM has experienced an explosion in the use of IS. In those early days, while MRP systems facilitated the day-to-day planning of manufacturing activities, CAD tools were developed to enable the design of complex components with an unprecedented level of precision. To close the loop, CIM systems emerged to facilitate the use and supervision of automated production tools resulting from the evolution of physical technologies. Although such IS combination provided support for the design, planning, and control loop of OM, each one was function specific. As additional IS got added into the picture, including sales support and procurement systems, the inherent standalone nature of such tools created interface maintenance challenges and quality problems due to redundant databases, incompatible protocols, and data formats. Such challenges, in turn, created the need for the first fundamental change in the role of IS, as depicted in Table 1. Instead of providing function-specific support, digital tools would have to provide comprehensive process-wide support. Additional benefits to such integration would ostensibly include reductions in data and software incompatibilities as well as redundancies (Jacobs & Weston, 2007). The challenges in such organizational and technology silos were addressed by a new cohort of IS vendors. Aided by the emergence of the client–server information architecture in the 1990s, companies like SAP embraced the challenge of combining the features of the previously function-specific tools into a single, companywide software suite and database. The implementation of these ERP systems turned out to be fraught with challenges, resulting in many well-publicized failures (Davenport, 1998), yet through their inherent support for business-wide integration (Gattiker & Goodhue, 2005) and process standardization (Cotteleer & Bendoly, 2006), they ultimately proved their worth for many organizations (Tenhiälä & Helkiö, 2015). Nevertheless, it also became evident that a single ERP system was not the optimal solution for everyone, and organizations with lesser needs for integration and standardization could perform well with standalone tools (Tenhiälä et al., 2018). To serve the needs of those organizations, a supplemental group of vendors, including Appian and Pegasystems, emerged to resolve the interface and redundancy problems in organizational workflows with a new digital tool called an iBPM system. As a natural extension to the broadening scope of the support of digital tools from individual business functions to entire business processes, a variety of technologies also emerged to support processes that spanned organizational boundaries, including radio-frequency identification for interorganizational product tracking (Bendoly et al., 2007) and APS systems featuring interorganizational supply network planning capabilities (Stadtler, 2005). By around 2015, the industry began to witness yet another critical development in the use of digital technologies. The decades-long trajectory in physical technologies that had led to ever-increasing industrial automation started to find ways to connect directly to digital technologies without a need for a human mediator. Equipped with sensors and algorithmic solutions, advanced robotics reached a new level of autonomy, leading to breakthroughs in a variety of operational settings from warehouse automation to robotic surgeries (Mukherjee & Sinha, 2020) and increasingly in the domain of knowledge-intensive professional services (Spring et al., 2022). Contemporary robotic solutions can relieve human operators from the physical burden of work or enable doing it beyond humanly achievable precision and consistency. In combination with AI, such solutions could assume an increasing proportion of the cognitive burden, as well. To resolve cognitive challenges, AI needs large datasets for training, which are increasingly drawn from constellations of sensors and communication tools known as IoT. While earlier sensor technologies enabled remote monitoring and predictive maintenance of industrial equipment (Persona et al., 2007) as well as real-time sharing of inventory data (Bendoly et al., 2007), current AI-enabled technologies are increasingly capable of proactively controlling and adjusting equipment to optimize maintenance and the timing and quantities of inventory replenishment. Advances in data analytics and in-memory computing (IMC) have critically improved the performance of these digital technologies, kicking off a trend where humans are no longer so much the users of the technology as they are its mere supervisors. In fact, even such a supervisory role could be already questioned, as recent research shows that human interventions and adjustments to the automated decisions of digital tools are more often counterproductive than they are beneficial (e.g., Caro & de Tejada Cuenca, 2023; Ibanez et al., 2018; Kesavan & Kushwaha, 2020). Although the evolution of IS in OM can be viewed through various lenses and perspectives (Grodal et al., 2023), here we emphasize the changing roles in the human-technology relationship (see Table 1) to better understand DT in OM as far more than a simple extrapolation of prior advancements. Concurrent with the emergence of digital technologies, and the rise of DT in OM, has been the appearance of critical questions related to how decision making can be informed or automated, as well as to how the pervasive use of digital technologies and DT impacts individual responsibilities and shifts power among producers, and consumers. Critically, decision support is increasingly provided by both human-driven analysis of such data, and advanced algorithmic solutions. In the extreme, this can represent a significant role reversal in decision-making, positioning non-human actors as decision makers and directing operational moves carried out by humans (Mims, 2021; Schechner, 2017). To best leverage the potential of both actors in advanced decision-making, human-machine interaction needs to be carefully designed (Gante & Angelopoulos, 2022, 2023; Hoberg & Imdahl, 2022). The spectrum from human driven, technology supported to technology driven, human-supported dynamics—with various degrees of concentration along this spectrum (i.e., a distribution of use)—increasingly characterizes and distinguishes contemporary organizations. This applies to both the case of administrative processes as well as to processes such as order-picking in warehouses (e.g., Sun et al., 2022). Less clear are the costs and benefits of specific levels of agency reversal for organizations, for example, when technology usurps the traditional principal role held by humans, or the pressures that these place on the stewardship of the datasets needed to train algorithmic solutions (Angelopoulos et al., 2021). The broader management literature has long debated agency. Insights from the early work of Chase (1978), for instance, inspired a wealth of subsequent discussions regarding the varying importance of customers as co-producers, critical to the success of service operations (e.g., Cho et al., 2022; Damali et al., 2022; Dellaert, 2019; Yalley, 2022). Certainly not all service processes benefit from a high degree of customer contact, and thus not all service outcomes are highly reliant on customer (inter)actions; however, some service processes are. As organizations maintain a range of service processes, the degree of customer reliance becomes a distribution bound by various levels of reliance. Further, many service operations that involve customers have a certain level of discretion regarding the quality of service and customer experience provided (Hopp et al., 2007). Organizations are accustomed to understanding—and strategizing around—the customer co-producer role, and they increasingly realize the need of customers to be viewed, in some instances, as partners rather than arms-length entities—not unlike often-referenced close ties between some supplier and buyer organizations. What organizations have only recently begun to consider, however, is a similar reliance on digital technologies, acting either as advanced agents or as quasi-principals. Such co-production has received limited scholarly attention. Xue et al. (2005) noted a paucity of discussion almost two decades ago, and a contemporary review of the management literature indicates that not much has changed in this regard. Their work discusses the critical, mediating role that particular digital technologies provide to customers positioned in co-producer roles. At this point, digital technologies were already starting to play a role in co-production, albeit still predominantly relegated to a static, or at best responsive, resource status. Currently, many advanced digital technologies can follow rules as well as make their own rules based on their exposure to datasets. In other words, many digital technologies have a capacity to learn, act upon such learning, and give rise to new dynamics. A contemporary example that has taken both academia and industry by storm is ChatGPT, which—according to itself—is a large language model trained to assist in generating human-like text based on provided input. There are analogies in how learning occurs across a range of digital technologies. Critical to appreciating such learning is a consideration of how digital technologies can leverage what they learn. When customers are considered as co-producers, they are seldom given the opportunity to make impactful changes to operating processes. In a product customization context, for instance, this relates to the use of combinatoric configurations to customize shoes (e.g., model, materials, and colors) without options to add free-form features that alter predefined designs (Randall et al., 2005). Such insights might be gleaned through product and service feedback, but specific and actionable solutions are unlikely to emerge from customers no matter how embedded they are as co-producers. Customers as co-producers, thus, are unlikely to appear beyond the level of agent in a relationship with organizations. Digital technologies pose a striking distinction in this regard, which we depict in Table 2, relative to the potential for customers as co-producers. Traditional Product purchase or limited engagement service settings Conditionally Typical High service-contact settings where success relies on customer involvement Highly Atypical Extreme MtO/DtO settings where customers fully direct value-added processes Traditional Information support settings constrained by interactivity limits; largely pre-scripted Pervasive Settings where sufficient detail is available to yield responsive, automated support towards actions Increasingly Typical Settings where advanced learning and autonomy permit AI-derived orders/placement/staffing There are both upsides and downsides implied by these shifting roles. From an OM perspective, DT clearly has the potential to empower both customers and organizations, however, it can also take some of the decision-making power out of their hands. From a customer empowerment perspective, DT can be operationalized in ways that increase transparency and help customers rationalize benefits and costs/risks associated with a wider array of options (e.g., Clemons et al., 2006). Shifting cognitive burdens onto automated decision support systems can also allow customers to focus on critical aspects of goods and services they might otherwise overlook. That is, digital technologies as an increasingly decisive co-producer can prove a valuable companion to customers in the broader decision-making context of product, and process selection. This is the case, for example, when it comes to consumer feedback. In the early stages of expansion of digital technologies, the facilitation of IT-supported consumer feedback represented an unprecedented expansion of insight into real-time market performance and emerging trends. Organizations, however, were still reliant on that feedback being voluntarily provided. The increased power of digital technologies, however, can also be used by organizations in place of insights otherwise gained through coordination with customers. With the increasing volume, velocity, and variety of data, advanced algorithmic solutions can now discern shifting interests of customers before they voice their preferences (Zuo et al., 2022, 2023). Currently, every action that a consumer makes, from shopping cart placement to returns, from questions asked of search engines to social media chatter, from click streams to even biometric data, can flow regularly, and be collected, stored, and analyzed. Insights extracted from such large collections of micro-signals, analyzed through advanced algorithmic solutions, are now far more insightful for retailers than the often biased and sporadic nature of volunteered commentary that online retailers have been accustomed to. The co-producer role has very much shifted from the consumer to the system in this case, giving rise to timely and topical questions: What does this mean for the value of customer engagement? What does it imply for the value of investments oriented towards customer care? If the value of "partnership" shifts away from customers in their limited co-producer capacity and towards that of digital technologies in their exponentially growing capability, then one might anticipate at least some losses in benefits to certain customer sectors (e.g., perhaps those more likely to physically patronize but of less economic value and purchase frequency). We cannot, of course, simply presume that organizations will harness all benefits and opportunities provided by DT, abandoning efforts to foster customer relationships they have long supported, in place of more-lucrative data-intense or data-convenient ones. There are also clear arguments to be made for the value of increased agility that can provide benefits for all the involved parties. A related discussion can be had around human-machine interaction in the domain of forecasting. Many organizations apply advanced algorithmic solutions along with a variety of data sources to create highly granular forecasts, such as those for daily demand per product. A great deal of relevant information including sales history, promotions, transportation costs and condition, weather forecasts, and a host of macro- and micro-economic conditions (borrowing rates, changes in tax policy, etc.) are now readily available in highly structured formats and can be used as inputs for algorithmic solutions. Other information such as local events or income-dependent plant closures may be less structured and are only available to demand planners with "an ear to the ground." As a result, demand planners, who are also ultimately responsible for forecasting performance, often have the opportunity to adjust ostensibly powerful forecasts generated by algorithmic solutions (Perera et al., 2019). Prior research shows that the extent to which humans add value in forecasting varies greatly (Katsagounos et al., 2021; Khosrowabadi et al., 2022). A critical question, thus, arises as to how to best manage and use the human input in such processes. Rather than risk losses due to well-meaning but potentially biased human judgment, such additional insight might be viewed simply as an additional data stream, feeding into the prediction process as generated by algorithmic solutions, weighing that insight in accordance with its perceived value. Given its variable contribution, as more information becomes available for training and testing algorithmic solutions, the weight given to contextual insights is increasingly constrained (Angelopoulos et al., 2021; Kar et al., 2023). The co-producer role of planners in the production function is, consequently, also diminishing. This, however, also gives rise to an important question for the broader field: What implications does this hold for the standing of traditional planners in the apparatus of the organization? One might expect the responsibilities of such agents, or next-generation planners, to transition towards more nuanced organizational considerations that are not positioned for absorption through automation. These might include greater involvement in relationship development internally (e.g., with design teams) or externally (e.g., with strategic partners). While such value-adding shifts for planners and process managers may seem ideal, they are not forgone conclusions of the DT expansion. As we depict in Table 3, a consideration similar to that regarding customers can be made with regard to the shifting role of operations personnel and digital technologies. Here we can think of the potential for both personnel and technology to emerge from traditional, tactical roles into more strategically relevant positions in the value chain of an organization. For example, algorithmic solutions are becoming increasingly accessible to organizations interested in leveraging IoT for predictive, condition-based maintenance. The use of algorithmic solutions can create a complex criteria structure, such that real-time signals from sensors embedded within equipment can be quickly interpreted to reveal cost-effective cases for preventative maintenance, avoiding impending failures, and corrective maintenance incidents. The mining of process data (e.g., Van der Aalst et al., 2004) can similarly rely on algorithmic solutions to identify opportunities for eliminating unnecessary touchpoints and redundancy loops, some of which may have gone unnoticed for years. In these ways, digital technologies are solidly positioned as an agent capable of "getting things done," on par with operations personnel. Increasingly there are even discussions of algorithmic solutions working far more proactively (i.e., as principals), taking the adjustment of scheduling and processing flow into its own hands, unprompted by a human actor (e.g., Homayouni et al., 2023; Sun et al., 2022), further shifting the balance of agency towards the principal functions in decision making, as we illustrate in Table 3. Traditional Largely monitoring and formulaic execution; business continuation focus Conditionally Typical Providing contextual insight into, and operational solutions aligned with, strategic planning Highly Atypical Extremely integrated functional settings, where operations experts drive strategic decisions Traditional Information support settings, constrained by interactivity limits; reporting/flagging Pervasive Settings where data and technology yield predictive policies and prescriptive insights when needed Increasingly Typical Settings where advanced AI learning and autonomy preemptively pose and enact process adjustments A very similar transformation is taking place in the shop-floor. For instance, warehouse workers may be following autonomous robots equipped with algorithmic solutions on their picking route. Such workers may be conducting relatively simple, post-processing operations following robotic ones, such as etching or polishing after fully automated rotation molding and welding. The organizational-behavior implications of these changes in "process leadership," requires attention. The broader OM discipline should be wary of treating human workers as a-emotional agents, like Taylor did more than a century ago, given these shifting co-dependencies. A challenge in this respect is to manage DT through the involvement of current decision makers, especially since in many planning decisions the value of algorithmic solutions is not yet fully clear. Especially for complex, multi-period, supply-chain planning decisions under uncertainty, algorithmic solutions still lack the ability to take on principal roles. In such cases, with the human still acting as the principal, overcoming algorithm aversion is critical, where research suggests that humans need a certain level of autonomy to function well (Dietvorst et al., 2015, 2018). Finally, one important matter to realize is that in operations and supply chains, much of the produced or collected data are proprietary, have strategic value, and are often either generated or—purposefully—manipulated by humans. Many digital technologies rely on such data being reliable, of high quality, and readily available (Struijk et al., 2023). Despite the increasing presence of IoT sensors, current industrial practice shows that such data inputs (in particular so-called master data) require extensive human labor. All of these points come into play as individuals rationalize the trust they place in digital technologies as a part of decision processes (Little, 1970). Just like humans, digital technologies that have been designed to automate—or at least support—decision making are fallible, as well as subject to manipulation and occasionally the source of security concerns (Ou et al., 2022). As isolated sources of decisions, this can prove catastrophic. More virtuously, the combination of human decision makers and digital technologies through a human-in-the-loop approach could provide critical checks and balances in highly impactful activities and achieve collaboratively what neither humans nor digital technologies can achieve on their own. Building on the aforementioned arguments, it is also likely that more overarching changes in business models are likely to emerge through DT as digital technologies and algorithmic solutions open up new strategic opportunities. These developments may be accompanied by natural tensions between competitive priorities such as cost, flexibility, and speed (Olsen & Tomlin, 2020), though intelligence emerging through DT may also make possible heretofore unrealized synergies among these priorities. Associated are implications for internal business models and inter-organizational structures (supply chains and vertical partnerships) as we visually depict in Table 4. Traditional Product purchases only with limited services Increasingly Typical New service offering based on digital interaction with existing supplier-customer networks Increasingly Typical New products or services to existing customer-supplier networks leveraging digital interaction Traditional Product purchases only using new channels and partners Pervasive Service offering enhanced by value-added services offered by new partners integrated for current customers Increasingly Typical Settings with new products or services with new partners or disintermediation Access to data of high volume, velocity, and variety enables organizations to build closer relationships with their customers as they can now better monitor and optimize the use of their assets (Porter & Heppelmann, 2014). Data provided by IoT sensors such as production equipment, cargo containers, or aircraft turbines can enable suppliers to better unde