Business process management in the age of AI – three essential drifts

Rosemann, Michael; Brocke, Jan vom; Van Looy, Amy; Santoro, Flavia Article — Published Version Business process management in the age of AI – three essential drifts Information Systems and e-Business Management Provided in Cooperation with: Springer Nature Suggested Citation: Rosemann, Michael; Brocke, Jan vom; Van Looy, Amy; Santoro, Flavia (2024) : Business process management in the age of AI – three essential drifts, Information Systems and eBusiness Management, ISSN 1617-9854, Springer, Berlin, Heidelberg, Vol. 22, Iss. 3, pp. 415-429, https://doi.org/10.1007/s10257-024-00689-9 This Version is available at: https://hdl.handle.net/10419/315073 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ EDITORIAL Information Systems and e-Business Management (2024) 22:415–429 https://doi.org/10.1007/s10257-024-00689-9 1 Introduction - from business process management to process science Business Process Management (BPM) is a holistic management discipline (Rosemann and vom Brocke 2015) that encompasses methods, techniques, and tools to support the management of business processes throughout their lifecycle, from discovery to execution, monitoring and mining (Dumas et al. 2018). The discipline of BPM covers an established field of research with common building blocks and comprises a wide variety of both technical aspects (e.g., process modelling, execution, mining, etc.) and managerial aspects (e.g., capabilities, governance, mindset, etc.) (Van Looy 2020). Although BPM was characterized by a clear boost in the 1990s and early 2000s due to the waves of business process reengineering (Hammer and Champy 1993), process innovation (Davenport 1993) and workflow management (van der Aalst and van Hee 2003), the underlying idea and importance of managing operational tasks and value chains have been recognized for centuries and this Published online: 3 September 2024 © The Author(s) 2024 Business process management in the age of AI – three essential drifts MichaelRosemann1,2· Jan vomBrocke2,3· AmyVanLooy4· FlaviaSantoro5 Jan vom Brocke [email protected] Michael Rosemann [email protected] Amy Van Looy amy[email protected] Flavia Santoro [email protected] 1 Queensland University of Technology, Brisbane, Australia 2 European Research Center for Information Systems, Münster, Germany 3 University of Münster, Münster, Germany 4 Ghent University, Ghent, Belgium 5 Institute of Technology and Leadership, Sao Paulo, Brazil 1 3 M. Rosemann et al. already since the early rise of factories and the notion of scientific management (Taylor 1919). Other disciplines explicitly recognize the operational level as an additional management layer complementing tactical and strategical viewpoints and which is needed to run a business (Ross et al. 2006). In recent years, the BPM discipline is been challenged by turbulent or dynamic business environments that are characterized by fast emerging technologies (e.g., artificial intelligence, Internet of Things, etc.) (Grisold et al. 2023) and external shocks (e.g., political instability, economic crises, pandemic) that can disrupt an organization’s default way of working (Roeglinger et al. 2022). This had amplifed the need for business processes that are not just cost efficient, but also agile, innovative and resilient by design (Van Looy 2021). In response, the BPM discipline has been broadening its foci to a wider set of strategic directions as well as to new technology, advancing BPM to be more adaptive to change (such as robotic process automation, predictive process monitoring, data-driven process optimization), while still sticking to its core building blocks and process lifecycle models (Van Looy 2020). Research has contributed to reconceptualize BPM accordingly. The BPM Billboard, for example, is a visual inquiry tool (Avdiji et al. 2020) that links BPM to an organization’s strategy, its context and its capabiltites, and it suggests specific projects to futher develop these and to measure related results (vom Brocke et al. 2021). In order to advance theorizing about the BPM discipline, and to decouple it from fast paced technological changes, Process Science has been proposed (vom Brocke et al. 2021; Brocke et al. 2021b). Process Science is considered a post-disciplinary approach, since it puts processes (as the phenomonen of interest) in the center of attention and invites contributions from as many disciplines that can make a contribution to identifying, understanding and intervening into processes. Process Science provides a platform for process research that is agnostic of disciplines. Hence, processes are generally defined as a coherent series of changes unfolding over multiple levels, contemporarily involving human actions with the use of digital technologies. Specifically, Process Science makes use of digital trace data in order to capture and analyze various process dynamics in real-time and based on naturally occurring data. Three types of Process Science have been distinguished so far: (1) descriptive Process Science, (2) explanatory Process Science and (3) prescriptive Process Science. A fourth one, generative Process Science, may be emerging with the rise of advanced AI solutions (Beheshti et al. 2023). This special issue demonstrates many avenues of process research to further develop. Specifically, we show some selected drifts that will become prevalent for the next generation of process research and BPM as a field. Next, we turn to the limits of the first generation Business Process Management. 2 The limits of first generation business process management Frederic Taylor’s book ‘The Principles of Scientific Management’ (Taylor 1919) is regarded by the Fellows of the Academy of Management as the most influential management book of the twentieth century. This is because Taylor’s relentless search for the one best way to conduct a specific task triggered a century-long investigation 1 3 416 Business process management in the age of AI – three essential drifts into how to identify and address underperforming business processes. For instance, Industrial Engineering grew out of it as a discipline, Harvard designed its first-year curriculum based on Taylor’s book, and various related disciplines have been developed addressing specific process issues including operations management, industrial automation, logistics and others. More recent approaches to process management are covering also white-collar business processes. Lean Management (Womack et al. 2007), for example, defines seven types of waste and proposes ways for how to overcome these. Six Sigma concentrates on variations as the root cause of process problems and deploys a set of mostly statistical methods to uncover and address these. This is also the case with process standardisation and its search for economies of process scale (Goel et al. 2023). Toyota’s continuous improvement (aka Kaizen) defines the elimination of process ‘pain points’ as an ongoing organizational routine (Maksberry 2011). Alternatively, Walmart’s just-in-time logistics takes the perfect synchronisation of independent processes to new levels of maturity (He 2023). And a number of process-aware information systems have been developed to address and reduce human-induced process errors (Dumas, van der Aalst, ter Hofstede 2005). Common process lifecycle models materialise this ‘problem-elimination mindset’ by first identifying process problems and their root causes, and then addressing these. As a result of all of these endeavours, there is consensus in this first generation of BPM regarding the stable north star of Business Process Management, namely the streamlined, ‘friction-free’ process that is also described by terms such as straightthrough processing, zero touch or simply, the optimised process. Cost and time reductions are the dominating aims of common process improvement. This traditional core of BPM has served organisations very well. The costs of process management have decreased, productivity increases have been noted across all industries and the ability to identify and address process issues is now a widespread capability. This increase in BPM maturity has led to new levels of transactional efficiency and made first generation BPM credible, if not indispensable. However, the ongoing race towards the ideal of a friction-free business process comes with a number of limitations. First of all, it is a race towards 0. Ultimately, one can envisage the fully digital, mobile business process which is ‘hyper-convenient’, and still only a hygiene factor. This is the case when customers are accustomed to flawless process execution and as a result do not respond with significant appreciation. Such a state can already be observed with ERP-supported business processes that nowadays are the common core of large and medium-sized organizations, and are providing a baseline operational capability. As a consequence, the future argument might be ‘business processes do not matter’ to re-phrase Carr (2003). Second, the ongoing elimination of what looks like non-value adding tasks comes with a cultural price (Schmiedel et al. 2020). Process optimisation and automation have taken a significant amount of human activity out of the world’s business processes and has cut millions of jobs globally for the gain of organizational performance. As a result, BPM can have the stigma of stressing process performance more than people’s well-being (Van Looy and Shafagatova 2016), making it a challenging choice in organizations that strive for purpose and are much aware of offering an attractive work culture to attract limited talent. 1 3 417 M. Rosemann et al. For these two reasons, a renewed view on BPM and its primary objectives is required. No longer is the existing reductionist approach to business processes centred on the elimination of issues sufficient for the next generation of BPM. Rather, new environmental conditions require new BPM capabilities. This motivational moment in the history of BPM triggered this Special Issue and its call for papers that asked for new BPM conditions and capabilities, and ultimately entirely new perspectives on the future of BPM. 3 Next generation business process management Embracing a process view of the world and leaving the beaten tracks of first generation BPM as described above, we see important drifts in contemporary and emerging BPM. In this editorial, we will briefly outline three exemplary drifts of BPM that are grounded in technology-enabled new conditions leading to a demand for newly or reinterpreted BPM capabilities. Concept drift has been introduced as a BPM term commonly describing the slow but continuous change of process behaviour (Bose et al. 2011). Here, we use drift not on a process level, but a meta level describing a significant shift in BPM overall. In fact, not just one drift, but three. The following three drifts (Fig. 1) are only describing selected phenomena and by no means strive for completeness in capturing all characteristics of the next generation of BPM. Yet, we intend to introduce these drifts as stimuli and invite fellow researchers to identify and conceptualize drifts through their research in order to further develop Business Process Management so it embraces fundamentally new technological opportunities. Fig. 1 Three BPM drifts 1 3 418 Business process management in the age of AI – three essential drifts 4 From transaction to conversation Business processes are typically defined from a transactional view as processes can be seen as orchestrated transactions. Each transaction consists of an initiating state (e.g., order arrived), a task (e.g., verification), and a resulting state (e.g., order has been verified). This transactional paradigm manifests itself in the default notion of active and passive nodes in process models, in logfiles that feed process mining algorithms, and in large enterprise systems (e.g., SAP’s notion of transaction codes). There has been clearly a transaction-dominant logic within the first generation of Business Process Management (e.g., Georgakopolous et al. 1995). As a result, processes are broken down into discreet steps, and this is how processes present themselves not only to their users, but also to business analysts eager to improve these processes. Generative AI, and with it an increased ability for natural language-based interactions, is now facilitating an additional conversational view on business processes complementing the existing and established transactional view. Unlike transactions, conversations are not constrained by formal start and end states, but are much more fluid and natural. They are less predictable in their outcome, can overlap and be redundant, and are driven by the human user, and not the transactional system. We define conversational BPM as the ability to interact in natural language with process information and related exsternal data. This definition recognises that the next generation of BPM will go beyond a narrow focus on event logs and interrelate these closely with context information to derive at situational explanations (Grisold et al. 2024). Increased conversational process engagement has become popular as part of chatbots (Kurtz et al. 2022; Grisold et al. 2023) and the way customers can interact with a process (‘What is the status of my claim?’, ‘When can I expect delivery of my order?’). However, it equally applies to all other internal process-related stakeholders. No longer do these have to adopt a process mindset – thinking in discreet transactions. Rather they can interact, as process owners, analysts or users, with a process in a much more intuitive conversation as opposed to engaging with the additional cognitive load that comes with comprehending transactions. This could include investigative questions facilitating process forensics such as ‘What are the differences between processing this customer order and the previous one?’ or ‘Why did this process take so much longer than the average?’ Beyond clarifying or investigative purposes, conversations can be used to train or configure a BPM solution. For example, a process analyst might state any of the following: ●This process is a positive deviant (making it a benchmark for future process improvements) (Setiawan and Sadiq 2013). ●This customer prefers a human case manager who speaks her native language (impacting resource allocation). ●Assess the regional impact of the upcoming cyclone on demand and resourcing of our supply chain process. Finally, a process might also report on progress in conversational forms. 1 3 419 M. Rosemann et al. ●Demand for the process is declining and therefore I [the robot] reduced the price of the product by 4%. ●The most experienced case worker is currently on sick leave, so I need to know if I should delay the process or allocate a less experienced alternative? BPM and its methods (e.g., process modelling) typically require users to think like a (transactional) business process. The rapidly increasing capability of generative AI is now starting to facilitate a conversational interface with business processes, and such makes BPM systems more behave like a human. However, research and professional BPM practice are still in their infancy in comprehending the requirements and possibilities of conversational BPM. The next generation of BPM, therefore, will need to expand a transactional view on processes with a significant conceptualisation and formalization of a conversational view on processes. Speech-act theory (Austin, Urmson, Sbisa 1975) or communication theory (Miller 2005) might be fertile theories in this context. Thus, BPM needs to move on to a new institutional logic (Tumbas et al. 2015). The different persona taking part in such process conversations need to be identified and categorised recognizing that some of these persona will be non-human agency. Finally, models of transactional and conversational views on processes need to be integrated, e.g., what transactional data should inform what conversations? Hence, we conclude: Next generation BPM is characterized by a drift form transactional to conversational logic, where both humans and machines engage in a conversation including process and contextual information. 5 From automation to autonomization The digitalisation of business processes is largely a story of increasing and maturing process automation. The first low-hanging fruit for automation within a process was task automation. Simple automation of calculations led in the 50s/60s to scalable bill-of-materials processing, which facilitated material requirements planning. This became the first comprehensive module within more complex manufacturing (MRP II) solutions, and increasingly all main tasks (from long-term forecasting to scheduling and order release planning) were supported. Next, administrative tasks within areas such as finance (e.g., invoice verification), asset management (e.g., depreciation) or human resources (e.g., payroll) were automated, too. As an Integrated solution, this was called an ERP system (e.g., SAP), namely a system that at its core automated the tasks. It was up to its human users to trigger these tasks (e.g., by prompting a transaction code in SAP) in a specific sequence and to feed the system with the data and business rules it required (so called customization). Second, the 1990s saw the emergence of workflow management, i.e. the automation of control flows (van der Aalst and van Hee 2003). Once a task has finished and reached its end state, the workflow management system takes over and triggers the next task according to the underlying process (workflow) model. Relevant resources are identified using role resolution where the most adequate staff is selected based on 1 3 420 Business process management in the age of AI – three essential drifts criteria such as qualification, previous experiences (e.g., with the case or customer), and availability. This meant that human users no longer had to initiate the subsequent task in a business process. Third, and more recently, robotic process automation (RPA) has also allowed to automate the interactions of a user with a task (Syed et al. 2020). By either manually training or screen scraping, the routines performed and rules followed by human users are understood and where possible replicated by a software robot. In these three types of automation, technology is taking over tasks from humans so these can focus on value-adding tasks requiring empathy and other human-only capabilities (‘from mundane to humane’). Automated processes are conducted faster, instantly, and cost-effectively, at least in terms of their variable costs. A specific feature of process automation is that it has been designed, tested and optimised by humans. Machines tasked with automation are programmed so that the outcomes are precisely as predicted and according to defined rules (e.g., a workflow management system will route work according to the defined resource allocation strategy). The fourth and next stage of the deployment of technology to business processes, however, is distinct from these first three types. Enabled by, but not depending on artificial intelligence, it is now possible to delegate decisions, or, more broadly, the governance of a process to machines. This means, machines, and not just humans, can make a decision and also change decision making processes. For example, an autonomous BPM system might identify additional events that matter as part of a process execution (very much like how AI in medical imaging has identified previously unknown symptoms as part of a diagnosis). Where humans are constrained in their selection of datasets that matter in the context of a business rule, a robot could scan and assess a broader set of contextual data for their impact on a business process. This could, for example, include evaluating various company-external variables (e.g., weather, holidays, festivals) in terms of their correlation with the demand for the process. It could also lead to revised resource allocation rules due to identified new correlations between resources and process performance, or including additional external events (e.g., sentiment analysis) in the configuration of a campaign processes, for example. We define process autonomization as the empowerment of a business process to make decisions in light of defined goals and constraints. These decisions could relate to events, control flow, data, duration, and resources involved. Figure 2 summarises these four stages leading from automation to autonomization. Autonomous decisions can be situational and case-constrained (e.g., a specific customer order is assigned a higher priority because the customer also just submitted a very large request-for-quotation), or structural and with this apply to all subsequent cases (e.g., an additional decision point is inserted into the process to enable a more adequate resource allocation). Autonomous processes are very distinct to automated processes. Whereas the latter are efficient forms of arriving at a desired and predictable outcome, autonomous processes are not predictable. Both in terms of the emerging process model and in the process outcome, they are sensitive and adaptable. 1 3 421 M. Rosemann et al. Thus, the next generation of BPM needs to invest academic and professional bandwidth in better understanding and conceptualizing autonomization as a new process capability. Related research, for example, could be dedicated to. ●Developing a taxonomy of process autonomization. ●Define extensions of process modelling languages to describe process autnomization in an at last semi-formal notation. ●Create procedure models for process analysts to decide on the adequate level and form of process autonomization. ●Explore the notion of responsible autonomization. Hence, we conclude: Next generation BPM is characterized by a drift from an automation to an autonomization logic, where machines will take over an increasing part of process governance. 6 From simplification to sophistication A typical to-be process looks much simpler than the corresponding as-is process as it tends to have less re-work, less variants, less approval steps, less paper-related activities (e.g., printing, copying, filing) and overall less non-value adding tasks. Simplified processes are easier to manage and govern, and have reduced costs of process execution. The skills to achieve process simplification are readily available as process analysts are trained in the methods and techniques that help them to identify end eliminate cost-driving process complexities. This includes in particular activity-based costing and its embedded identification of process cost drivers. In fact, many organizations have made the ‘one process’ notion (e.g., one claims engine at an Australian insurance company) the main mantra of their organization-wide BPM initiatives. Therefore, process simplification is often seen as a proxy for process Fig. 2 From automation to autonomization 1 3 422 Business process management in the age of AI – three essential drifts vom Brocke J, Mendling J, Rosemann M (2021) Planning and scoping business process management with the BPM billboard. In: vom Brocke J, Mendling J, Rosemann M (eds) Business process management cases, vol 2. Springer, Berlin, Heidelberg vom Brocke J, van der Aalst WMP, Grisold T, Kremser W, Mendling J, Pentland B, Recker J, Roeglinger M, Rosemann M, Weber B (2021) Process science: the interdisciplinary study of continuous change. SSRN Electron J, 1–9 vom Brocke J, Weber M, Grisold T (2021) A matrix for context-aware business process management: Empirical evidence from Hilti. In: de Leoni, M., Song, M., Röglinger, M. (eds). Industry Forum of the 19th International Conference on Business Process Management, 6–10 September 2021, Rome. https://ceur-ws.org/Vol-3112/paper5.pdf (Accessed April 2024) vom Brocke J, Zelt S, Schmiedel T (2016) On the role of context in business process management. Int J Inf Manag 36(3):486–495 Womack JP, Jones DT, Roos D (2007) The machine that changed the world: the story of lean production. Free, New York Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. 1 3 429