An Intelligent Museum Agent Framework (IMAF): A Design Science Research Approach to Agentic AI in Museums

Although museums are increasingly integrating large language models (LLMs) into their operations, current applications remain largely confined to narrow tasks such as chatbots and document management. While the emergence of agentic AI offers more advanced, autonomous capabilities, existing frameworks are primarily designed for general-purpose environments. Consequently, they provide limited guidance for museums, which function as complex socio-technical systems governed by strict institutional mandates, ethical responsibilities, and operational constraints. Due to these complexities, the practical deployment of agentic AI in real-world museum settings remains highly challenging. At present, a system-level architecture for agentic AI tailored specifically to the unique requirements of museums is notably absent. To address this gap, this study employs the Design Science Research (DSR) paradigm to propose the Intelligent Museum Agent Framework (IMAF). The proposed framework extends an existing general agent architecture by incorporating a bifurcated module for contexts to systematically enforce institutional and operational constraints. Additionally, it features an integrated memory structure grounded in cultural heritage data to ensure factual reliability. Ultimately, the IMAF provides a conceptual systems integration guideline for museums exploring domain-constrained autonomy and offers a basis for future implementation and empirical validation.