Use of Agent-Based AI Applications in Research Institutions

Research institutions face increasing pressure to ensure methodological rigour, transparency, and governance compliance under conditions of growing complexity and limited human resources. While generative artificial intelligence is often discussed as a tool for efficiency gains, its use in academic contexts raises fundamental concerns regarding responsibility, decision authority, and the risk of implicit automation of scholarly judgement. This paper addresses these challenges by proposing a governance-aware, agent-based approach to AI-supported research processes. The study develops a structured framework for integrating agent-based AI support into scientific workflows without delegating epistemic or managerial responsibility. Instead of treating AI as a monolithic system, the proposed approach decomposes support functions into specialised agents that perform clearly bounded tasks such as structural checks, methodological consistency analysis, citation validation, and formal compliance review. Human actors retain full control over interpretation, prioritisation, and decision-making at all stages of the research process. Methodologically, the paper follows a design-oriented research approach. It introduces a reference process for AI-supported academic work, a role- and agent-based interaction model, a multi-agent system architecture, and a governance and operating model that ensures transparency, accountability, and institutional control. The applicability of the approach is demonstrated through an illustrative use case covering the development of an exposé, manuscript preparation, Internal Review, and submission readiness. The contribution of the paper is threefold. First, it provides a practically implementable model for AI-supported research workflows that preserves scholarly autonomy. Second, it translates abstract governance principles into concrete organisational and technical design choices. Third, it offers research institutions a transferable blueprint for deploying generative AI as a supportive infrastructure rather than a decision-making authority. The paper concludes by discussing limitations and outlining avenues for future research on institutional AI governance in academia Keywords: AI governance, managerial decision support, organisational design, human-in-the-loop systems, digital transformation in research organisations, responsible AI management, process governance, strategic use of artificial intelligence