AI Agents in Action Foundations for Evaluation and Governance 2025

Executive summary This report has been tailored mainly for adopters of AI agents, including decision-makers, technical leaders and practitioners seeking to integrate AI agents into organizational workflows and services. While AI agents are gaining traction, there remains limited guidance on how to design, test and oversee them responsibly. This paper aims to help fill that gap by providing a structured foundation for the safe and effective deployment of these systems. The paper makes three key contributions. Firstly, it covers the technical foundations of AI agents, including their architectures, protocols and security considerations. Secondly, it offers a functional classification that differentiates agents by their role, autonomy, authority, predictability and operational context. Thirdly, it suggests a progressive governance approach that directly connects evaluation and safeguards to an agent’s task scope and deployment environment. Together, these elements guide adopters with a conceptual blueprint for moving from experimentation to deployment. The report highlights the importance of aligning adoption with evaluation and governance practices to ensure that AI agents are successfully deployed while trust, safety and accountability are maintained.This paper explores the emergence of AI agents, outlining their technical foundations, classification, evaluation and governance to support safe and effective adoption. 4 AI Agents in Action: Foundations for Evaluation and Governance