Every day, across a wide range of industries, companies must adapt employee access profiles to reflect changes in authorization requirements. A key challenge lies in translating the needs of an employee to a technical authorization concept. Due to the rise of generative AI, it makes sense to consider Large Language Models (LLMs) for processing permission requests. This is especially relevant in large-scale access control environments, where the high number of permissions and roles makes oversight difficult for human administrators. Thus, in this contribution, we conduct quantitative research and test whether LLMs are suitable for permission request processing. We find that when embedding an access control document, an LLM performs quite well when translating a permission request into the required permission but struggles with suggesting a suitable role. We show that instead of using an LLM for both permission and role suggestions, it is more efficient to apply a hybrid approach, and first apply an LLM-based permission translation and then algorithmically search for a role that contains the translated permission.
Despite the widespread use of domain models, the modeling process remains underexplored, particularly regarding the interactions among agents, products, and activities. Building on prior work that identified 16 recurrent moments of dissatisfaction (“pain points”) experienced by modelers during these interactions, this study offers a deeper analysis to clarify the significance of these pain points and support improvements in modeling practice. Through an online survey of 49 modelers, the study provides empirical evidence on the frequency of these moments, the reasons behind the frustrations they cause, and the strategies modelers use to address them. The descriptive analysis offers valuable insights into these aspects, revealing interesting patterns among modelers. These findings have implications for practice and academia, offering a foundation to enhance the modeling experience and improve the value of domain modeling efforts.
When analysing business processes through process mining, cases are typically considered in isolation. A case is seen as one process execution, whose control-flow is only influenced by case-specific features. However, this isolated view on cases does not consider inter-case effects, which likely occur when cases compete for the availability of a fixed number of shared resources. In this paper, we study the stochastic implications of (1) transition life cycles and (2) resources. That is, we extend stochastic Petri nets with the notion of a transition life cycle to capture the non-atomicity of resources executing transitions. This then allows us to describe the competition between cases for resources stochastically and analyse how scarcity of resources influences the behaviour of real-life processes. We provide an according automated discovery method, and evaluate our technique on multiple real-life event logs by showing its applicability and feasibility, and demonstrating that new types of insights can be obtained for real-life processes.
Business processes in which decisions are made by human resources suffer from noise. Noise is unwanted variability, which leads to inconsistent and unrepeatable decisions and impacts trust negatively. In this paper, we present the Decision Noise Instrument (DNI) to quantitatively estimate the noise that can be attributed to unknown factors affecting the actors who carry out the business process. For our estimation, we solely use readily available data from business information systems that support the execution of the process. In our method, we limit the influence of factors that cannot be attributed to actors. The DNI makes it possible to compare the noise levels between and within business processes. Based on that comparison, further investigation into the problem and noise reduction efforts can be prioritized. We evaluate the DNI on a claim handling process of UWV, a public service provider in the Netherlands. Our results show that we can estimate differences between the noise levels of several decisions within the selected process of UWV. By drilling down on a decision, the subgroups that have the most impact on the noise level are identified.
Data Science (DS) technologies provide crucial competitive advantages in contemporary business environments through optimized processes. However, DS-based service systems require collaborative efforts between IT and business divisions, necessitating structured frameworks to bridge organizational gaps and ensure alignment between technical requirements and business objectives. This research examines the integration of DS systems within Enterprise Architecture (EA) models to address complex implementation challenges. EA modeling provides an indispensable, structured framework for ensuring the optimal performance of DS systems, particularly in the context of digital transformation initiatives, and enables the seamless integration of data science capabilities into organizational frameworks. This work presents a thorough literature review that critically examines existing research on integrating DS projects into EA models, offering a comprehensive view of DS integration to enhance architectural frameworks. Future work will demonstrate these holistic models in practical enterprise applications and conduct case studies to refine the model and evaluate the effectiveness of conceptual modeling approaches.
