AI Ecosystems: Assistants, Agents and Platforms

Permanent URI for this collectionhttps://hdl.handle.net/10125/112500

Browse

Now showing 1 - 6 of 6

Orchestrating Generative AI-Based Multi-Agent Systems for Complex Knowledge Work Automation: A Design Science Research Approach
(2026-01-06) Fetzer, Dominik; Gimpel, Henner; Schoch, Manfred
The pursuit of automation has been a key objective throughout industrial history. Advancements in information technology, such as robotic process automation, accelerated this progress for knowledge work. However, complex knowledge work was off-limits to automation. The emergence of generative artificial intelligence (GenAI) coupled with multi-agent systems (MAS) pushes the boundaries of what is technically feasible and economically viable. While several technical frameworks for developing GenAI-based MAS are available, the systems’ orchestration remains largely based on ad-hoc trial and error. The paper addresses this gap by developing design knowledge for GenAI-based MAS. Based on design science research, we present a morphological box of orchestration options and derive four propositions regarding GenAI-based MAS orchestration. Our research contributes to academic and practical understanding by offering design knowledge for GenAI-based MAS development, facilitating the automation of complex knowledge work.
Agentic AI Platform Networks
(2026-01-06) Schmidt, Rainer; Alt, Rainer; Zimmermann, Alfred
Agentic AI is a recent development in artificial intelligence (AI) technologies that is capable of automating task execution and decision-making. In the context of platforms, they provide novel approaches in interconnecting platforms. This enhances the traditional hub-and-spoke architecture of multi-sided platforms and leads to the formation of networks of platforms. This research analyzes four major AI platforms and identifies three distinct integration patterns: isolated architecture (Alexa), mediated integration (ChatGPT), and protocol-based networking (Claude/Gemini). The findings reveal that integration-centric architectures demonstrate higher cross-platform connectivity than traditional, interconnection approaches . They extend platform theory by introducing network orchestration as a new competitive strategy and provide empirical evidence of an emerging paradigm shift from platform competition to interconnection.
Integrating the AI Artifact in the IS Artifact: A Conceptual Framework and Research Agenda
(2026-01-06) Albrecht, Valerie; Viale Pereira, Gabriela
The AI artifact has become relevant in theory and practice. Empirical studies and conceptual frameworks explore the role of artificial intelligence, its impact, and application. However, the rapid development of new technologies leaves little time for theory to be developed at the rate of practical advances in the field of AI. This paper builds on previous work that contextualizes the IS artifact in general systems theory and updates the conceptual framework of the IS artifact for the context of AI. To achieve this, a narrative literature review of 30 papers has been conducted. The resulting first conceptualization of a potential framework indicates that the AI artifact adds complexity to the model by bridging existing categorizations. Agency is attributed to the AI artifact, requiring a new conceptualization and an additional environmental layer that requires systemic interconnectedness. The analysis also shows areas for further research
Evaluating and Improving Prompt Quality in LLM-Based Assistants: A Synthesis of Criteria and Indicators
(2026-01-06) Reinhard, Philipp; Sajzev, Vladimir; Li, Mahei; Leimeister, Jan Marco
Generative AI (GenAI) assistants, particularly large language models (LLMs), are gaining increasing relevance across domains. The quality of outputs generated by these systems is highly contingent on the input prompts, giving rise to new professional roles such as prompt engineers. In this study, we systematically examine evaluation criteria and optimization methods that can improve prompt quality. Drawing on a systematic literature review, we identify key criteria, including clarity, accuracy, and precision, and initial measurement techniques. In addition, we synthesize common optimization methods such as iterative refinement and shot-based prompting. Our work contributes to the growing efforts to standardize the evaluation and improvement of prompts in interactions with LLM-based assistants, thereby fostering a more rigorous and coherent understanding of the prompt quality construct.
Mapping the Moral Foundations of Machines: A Vignette-Based Inquiry into Moral Reasoning Across Six Large Language Model Platforms
(2026-01-06) Oliver, Alison; Crawley, Alan; De, Soumyajit; Jeay-Bizot, Lucas
Large language models take on an ever-increasing role in our societies. In particular, they can guide and inform human decisions. Understanding the moral profiles of LLMs, whether they are stable within the same models over time and across models, is key to ensuring LLM-informed decisions are not unduly shaped by model-specific moral emphases. Drawing from Moral Foundations Theory (MFT), we evaluated the moral profiles of six different LLMs developed in different sociotechnical contexts to measure stability of moral profiles across models. We measured test-retest reliability of LLMs’ explanations of their moral judgments using content analysis to test the stability of output patterns within LLMs. We found mostly stable moral profiles within and across models with few exceptions. We frame these cross-model differences and exceptions as actionable guidance for model selection and routing in morally relevant use cases.
Introduction to the Minitrack on AI Ecosystems: Assistants, Agents and Platforms
(2026-01-06) Schmidt, Rainer; Zimmermann, Alfred; Alt, Rainer

Browse

Recent Submissions