The proliferation of data-intensive applications, such as artificial intelligence and smart manufacturing, has significantly increased the need for effective data management strategies. However, existing data management strategy frameworks are often too generic and lack practical implementation guidance. This leads organizations to develop ad hoc or custom-built solutions. We address this gap by developing a lightweight data readiness framework (DRF) specifically designed for data-intensive applications. Utilizing a tailored design science approach and a systematic literature review of 42 relevant studies, we identify five dimensions of DR: metadata management, data provenance and lineage, data quality, data interoperability, and data governance. The resulting DRF enables a structured classification and assessment of DR aspects in data-intensive projects. To evaluate the framework, we applied it to three digital twin demonstrators with varying levels of practical complexity. The evaluation provided early indications of the framework ‘s applicability, utility, and robustness.

In view of the growing significance of artificial intelligence (AI) in healthcare, compliance with the European Union’s AI-Act (AI-Act) is crucial for ensuring patient safety, data security and legal conformity. This research-in-progress paper examines the impact of the AI-Act on the design of healthcare processes that are supported conversational agents. It highlights the challenges arising from the new regulatory requirements of the AI-Act and discusses preliminary approaches for integrating these provisions into healthcare processes. The focus is on analyzing the regulatory effects on the automation of healthcare processes. Our first findings provide a basis for further research and the eventual implementation of the AI-Act’s requirements. The paper underlines the need for clear risk classifications and precise definitions of AI systems, as well as their potential effects on efficiency and patient safety in healthcare.

Die Fragmentierung von Informationsflüssen und Zuständigkeiten im Gesundheitswesen erzeugt erhebliche Ineffizienzen und einen hohen Koordinationsaufwand für Patient:innen. KI-gestützte Agenten, die über standardisierte Schnittstellen wie das Model Context Protocol (MCP) an heterogene Ressourcensysteme angebunden werden, bieten einen vielversprechenden Ansatz für eine vermittelnde Koordinationsschicht. Offen bleibt jedoch, welche agentische (auf autonomen, zielgerichteten KI-Agenten basierende) Architektur – zentralisierte Single-Agent- oder dezentralisierte Multi-Agent-Architektur – sich für die komplexen Anforderungen der Versorgungssteuerung besser eignet. Dieser Beitrag verfolgt einen Design-Science-Research-Ansatz: Zwei konkurrierende Prototypen werden auf einer gemeinsamen MCP-Infrastruktur entwickelt und anhand von zehn systematisch konstruierten Testfällen evaluiert. Die Ergebnisse zeigen, dass Single-Agent-Systeme bei atomaren Anfragen mit geringer Werkzeugdichte überlegen sind, während Multi-Agent-Systeme bei komplexen, domänenübergreifenden und sicherheitskritischen Aufgaben in der Regel besser abschneiden. Ein zentraler architekturunabhängiger Befund ist das Persistenzproblem sicherheitskritischer Kontextinformationen: LLM-eigenes Wissen zu Kontraindikationen und Risikofaktoren verblasst im Verlauf längerer Konversationen – unabhängig von der gewählten Architektur. Aus der vergleichenden Evaluation werden vier Designprinzipien abgeleitet: domänenbasierte Spezialisierung, deterministischer Safety-State, explizite Übergabe-Schemata und eine architekturangepasste Komplexitätsschwelle. Der Beitrag liefert präskriptives Gestaltungswissen für die Entwicklung robuster, skalierbarer und auditierbarer Agentenarchitekturen im Gesundheitswesen.

The traditional process of conducting literature reviews requires a significant amount of manual work and is often constrained by the limitations of small sample sizes. Although novel computational approaches to language analysis offer great opportunities, they are rarely applied to literature reviews. We aim to demonstrate the potential of ontology-based computational literature reviews. Therefore, we conducted a Design Science Research (DSR) project to apply and extend a method for this type of review. Since DSR shows a considerable diversity in theoretical foundations and methodological approaches and can serve as an insightful example, we first developed a machine learning classifier to identify DSR articles. Second, we applied and extended a method for ontological annotation and sentence classification. Finally, we conducted a computational literature review of 6235 DSR articles, focusing on the distribution of theories, methods, and topics. We also developed an interactive dashboard prototype with selected results from our study.

Digital twins (DTs) can transform after-sales services by supplying real-time data, predictions and remote control, yet firms lack clear guidance on turning these features into business value. Anchored in explorative Business Process Management (BPM), this study develops a design-oriented taxonomy that links DT capabilities to revenue-focused process redesign. Using Nickerson et al.’s iterative method, we combined concepts from after-sales management, DT engineering and explorative BPM and refined them through three development cycles. The resulting taxonomy comprises seven dimensions, namely: Service Type, Customer Effort, DT Information Product, DT Media Format, DT Information Value and Customer-to-Process Interaction and Organisation-to-Process interaction. We demonstrated utility by classifying 30 automobile diagnostic and repair services for automobiles launched between 2016 and 2024. The study contributes (1) a taxonomy for DT-enabled after-sales and (2) a practical lens for determining high-value redesign opportunities.

This study examines how digital platform introduction drives business model change in small- and medium-sized enterprises (SMEs). Drawing on a case study of four textile firms, we analyze three phases of platform introduction: pre-platform, engagement, and post-platform launch. The findings reveal a dynamic platform orientation, understood as a cumulative and capability-contingent pattern of SMEs’ engagement with the platform over time. All firms initially adopt an efficiency-focused orientation as a resource-conserving entry logic. However, only those that embed the platform deeply enough to enhance their digital, agile, and network capabilities develop episodic innovation-oriented engagement, which may become dominant and enable more transformative business model change. Others remain primarily efficiency-oriented, achieving only incremental gains. The study demonstrates that pre-existing organizational capabilities shape how the platform orientation unfolds and that improvements in these capabilities determine whether an innovation orientation becomes viable. This study theorizes dynamic platform orientation as the mechanism explaining why similar SMEs participating in the same joint digital platform introduction experience divergent business model outcomes. This mechanism advances research on SME digital transformation and clarifies how tightly coupled SMEs can leverage digital platforms for business model change.

Conversational AI systems, such as chatbots and virtual assistants, are becoming increasingly important to digital business processes. However, the established Business Process Model and Notation (BPMN) standard faces challenges when representing dynamic, context-sensitive interactions. This paper addresses this methodological and practical research gap by developing a standard-compliant BPMN extension (BPMN4CAI). Using Design Science Research methodology, this paper develops an approach that systematically extends existing BPMN elements and incorporates specialized components. The applicability and relevance of the BPMN4CAI framework are demonstrated and evaluated through a case study. The results show that the BPMN4CAI extension facilitates adaptive decision-making processes, robust context management, and transparent interactions for Conversational AI within business processes.

Die Verfügbarkeit und Qualität von Metadaten sind zentrale Erfolgsfaktoren für die effektive Nutzung von Open Data. Unvollständige oder inkonsistente Metadaten erschweren jedoch die Auffindbarkeit, Interoperabilität und Wiederverwendung offener Datenbestände erheblich. Dieser Beitrag untersucht, inwieweit Open-Source Large Language Models (LLMs) zur automatisierten Generierung hochwertiger Metadaten beitragen können. Aufbauend auf dem Data Catalog Vocabulary (DCAT) wird ein prototypischer Ansatz entwickelt, der Open-Source-LLMs in den Upload-Prozess von Open-Data-Portalen integriert. Die Evaluation anhand realer Open-Data-Datensätze zeigt, dass die generierten Metadaten in vielen Fällen mit Expertenbewertungen übereinstimmen und insbesondere die Konsistenz und Kategorisierung verbessern. Herausforderungen bestehen jedoch in der präzisen zeitlichen Zuordnung von Daten sowie in der Skalierbarkeit auf unterschiedlichen Datenformaten. Der Beitrag beleuchtet methodische sowie technische Optimierungsmöglichkeiten und gibt Hinweise auf potenzielle Verbesserungen hinsichtlich Skalierbarkeit und zeitlicher Einordnung von Metadaten.

The process of conducting scientific literature reviews is becoming increasingly complex and time-consuming due to the rapid expansion of available research. Popular academic search engines offer limited filtering capabilities and suffer from low precision. Machine learning-enhanced approaches tend to target rather specific areas, and novel approaches based on generative artificial intelligence suffer from hallucinations. Drawing on information foraging theory, this article presents a design science research project aimed at generating design knowledge for developing domain-specific search systems for research articles. Our contributions include: (1) integrating domain ontologies with large language models to design ontology-based search systems, (2) generating descriptive design knowledge by exploring the problem space, (3) generating prescriptive design knowledge for developing domain-specific search systems, and (4) presenting an ontology-based search engine prototype. Our results indicate that the proposed solution supports researchers in conducting literature reviews by increasing information gain while reducing interaction costs.

In the Information Systems (IS) discipline, central contributions of research projects are often represented in graphical research models, clearly illustrating constructs and their relationships. Although thousands of such representations exist, methods for extracting this source of knowledge are still in an early stage. We present a method for (1) extracting graphical research models from articles, (2) generating synthetic training data for (3) performing object detection with a neural network, and (4) a graph reconstruction algorithm to (5) storing results into a designated research model format. We trained YOLOv7 on 20,000 generated diagrams and evaluated its performance on 100 manually reconstructed diagrams from the Senior Scholars' Basket. The results for extracting graphical research models show a F1-score of 0.82 for nodes, 0.72 for links, and an accuracy of 0.72 for labels, indicating the applicability for supporting the population of knowledge repositories contributing to knowledge synthesis.

Many company networks, especially those comprising small- and medium-sized enterprises (SMEs), face the challenge of digitally transforming their value co-creation (VCC). However, this topic, despite its high relevance, remains vastly under-researched. We thus conducted a case study in an SME network to investigate how these companies adapt their internal processes to enable an overarching VCC process in the network and use IT to support it. We first derived a framework of four propositions showing that the modularization of value creation, equality of actors, efficient information and knowledge flow, and inter-organizational information technology support facilitate VCC in these networks. The propositions framework then became our lens to analyze the SME network. Our empirical study enabled us to gain deeper insights into the relationships of the proposed facilitating factors for VCC in company networks and the role of inter-organizational information systems in this context.

Originating from product life cycle management, more and more Digital Twins (DTs) are implemented in research contexts and industrial use cases. DTs allow for continuous information provision, a mandatory requirement, e.g., for the circular economy. Implementing a DT faces integration challenges and requires precise planning for selecting specific components of a DT and its functionalities. A process model supports making informed decisions, structuring the application and its demand for certain DT elements, and streamlining implementation processes. Albeit specific implementations come with individual requirements, a generic process model allows for the transfer of general instructions and rules to particular cases. Our ongoing research focuses on developing a process model for implementing DTs. The process model builds upon the VDI norm 2206 for implementing cyber-physical systems and enhances the VDI norm by providing more granular sub-modules. This paper reports on the first development steps and preliminary artifacts, e.g., a visual inquiry tool in a DT implementation canvas. Our research builds upon a thorough analysis of theoretical and practical requirements and existing frameworks and architectures for DTs.