Agentic AI

Artificial intelligence (AI) is shifting from simple automation (Janiesch et al. 2021 ) and content generation (Feuerriegel et al. 2024 ) towards solving complex goals across enterprise landscapes (Baird and Maruping 2021 ). Enabled by advances in generative AI (GenAI) and large language models (LLMs) (Feuerriegel et al. 2024 ; Schneider et al. 2024 ), tool learning (Qin et al. 2025 ; Qu et al. 2025 ), as well as large action models (LAMs) (Zhang et al. 2025 ), agentic AI systems are emerging in practice and research (Holldack et al. 2026 ; Allmendinger et al. 2026 ). With GenAI capabilities gradually entering the enterprise context via cloud-based enterprise platforms (with on-premise alternatives emerging) (Haki et al. 2025 ), and with agent modes from OpenAI and Google, multi-agent frameworks such as Manus, and technologies like SAP’s Joule or Microsoft’s Agent Framework, agentic AI capabilities are becoming an integral part of enterprise platforms and can enable novel ways to integrate existing systems and technologies. Organizations can benefit from this development, as many previously unaddressed tasks span heterogeneous systems, creating a need for interoperability and requiring contextual interpretation that cannot be predefined at design time. Yet, realizing this potential requires first addressing challenges around governance, coordination, and human oversight.