Digital twins have been used in many settings ranging from spacecraft and wind turbine simulations to chemical processes and urban planning. This editorial focuses on the digital twin of an organization (Kerremans and Kopcho 2019). However, we first provide a more abstract definition of a digital twin using Fig. 1 (van der Aalst 2021a; Fuller et al. 2020).
A digital model is a reflection of reality that is created manually and functions in an offline manner, i.e., the model does not change when reality changes. An example is the classical use of simulation tools like Arena, AnyLogic, Vensim, or Simul8 modeling a production line or supply chain. Such a digital model can be used to reason about reality and answer what-if questions. A digital shadow goes one step further. The model is now automatically derived and changes when reality changes. The digital shadow can also be used to reason about reality and answer what-if questions. Although the digital shadow is based on data tapped from reality, there is no automated real-time feedback loop. The insights produced by the digital shadow do not automatically trigger changes. This is still done manually after interpreting the results. The Internet of Production (IoP) developed at RWTH Aachen University provides collections of digital shadows supported by an infrastructure that is using AI/ML techniques with a focus on process mining (van der Aalst et al. 2021; Liebenberg and Jarke 2020). A digital twin goes one step further. Results of the digital twin directly impact reality. For example, when the simulation model predicts a delay, the production process is reconfigured automatically.
The idea of a digital twin is appealing. In the virtual world, all possible decisions can be evaluated without causing harm, waste, and costs. Typically, stochastic models are used to cope with uncertainty. This explains why simulation tools play a crucial role in the development of digital twins. In selected application domains, it is already possible to create reliable digital twins that can automatically respond to behaviors observed in reality. Early examples of digital twins were relatively simple, focusing on a single physical thing (e.g., a wing of an airplane). However, over time more complicated settings were considered. As discussed in recent literature, digital twins now play a key role in cyber-physical systems, Industry 4.0, Internet-of-Things, smart cities, aviation, energy, and healthcare (Fuller et al. 2020; Kritzinger et al. 2018; Caporuscio 2020). These successes led to the idea of a Digital Twin of an Organization (DTO).
Although the term "Digital Twin of an Organization" (DTO) existed before, it only became a topic of discussion when Gartner started to promote the concept a few years ago (Kerremans and Kopcho 2019). Gartner uses the following definition: "A digital twin of an organization (DTO) is a dynamic software model of any organization that relies on operational and/or other data to understand how an organization operationalizes its business model, connects with its current state, responds to changes, deploys resources and delivers exceptional customer value." Creating a DTO can be seen as one of the grand challenges in information systems.
Why is it so challenging to create a DTO? There are two main reasons:
The boundaries of an organization and, therefore, also a DTO are not so clear, i.e., an organization has customers, suppliers, employees that collectively influence the processes.
Human and organizational behavior may be irrational and change over time (influenced by regulations, social interactions, and personal preferences).
For most organizations, it is not feasible to create a DTO that captures reality well. However, the desire to model, visualize and understand the complex context in which an organization operates is compelling. One can view process mining as a concrete technology to facilitate such a DTO (van der Aalst 2016). Using process discovery, one can discover the so-called "control-flow model" (represented using Petri nets, process trees, or BPMN models) and by aligning event data with the control-flow model, it is possible to add other perspectives (time, costs, resources, decisions, etc.). The resulting more elaborate model can be simulated (van der Aalst 2016). Several process-mining tools provide such a simulation facility (e.g., ProM, Celonis, and Apromore). Also, business process modeling and simulation tools (e.g., Signavio, Aris, and Simul8) have added process-mining capabilities to automatically learn simulation models. Using process mining, it is relatively easy to create a digital shadow (see Fig. 1b). However, due to the challenges mentioned before, it is extremely difficult to create a model that behaves like a real organization. Also, multiple processes interact and compete for resources concurrently. The importance of concurrency and complex interactions between objects (customers, orders, products, machines, people, etc.) is elaborated on in Van der Aalst (2021a). Hence, it is not sufficient to consider one process in isolation. Moreover, to create a digital twin, as shown in Fig. 1c, the DTO also needs to automatically take action. Action-oriented process mining provides initial steps for this. For example, the Celonis Execution Management System (EMS) can trigger corrective workflows using the Integromat integration platform.
Although process mining provides initial capabilities to create a DTO, it is fair to say that, currently, DTOs are more a vision than a reality. Moreover, to make DTOs resilient, we need humans in the loop (Abdel-Karim et al. 2020) to cope with disruptions. This is illustrated in Fig. 1d. As reality changes due to disruptions, the digital twin should still be useful for human decision-makers. In other words, we want the combination to be resilient.