Agentic AI Platform Presentation for Keycloak KeyConf 2025

Martin Besozzi presentation for Keycloak KeyConf 2025

Martin Besozzi presents “Agentic AI Platform for Enterprise” at Keycloak KeyConf 2025 (recorded session).

Martin Besozzi explores how Agentic AI systems can be safely introduced into enterprise environments using open IAM standards and zero-trust principles.

The session examines the growing challenge of AI agents acting autonomously across enterprise systems and why traditional application-centric security models are insufficient for this new paradigm.
Martin explains how Identity and Access Management (IAM) becomes a foundational layer for governing agent behavior, decision-making, and scope of action.
The presentation connects OAuth 2.0, OpenID Connect, and the emerging OpenID AuthZEN specifications to real-world agentic use cases, showing how standards-based authentication, authorization, and policy enforcement can be applied to:

• Secure AI Agent in the context of MCP (On-Behalf-User)
• Enforce fine-grained, relationship-based authorization
• Decouple policy decisions from application logic
• Apply zero-trust governance to AI Agents

Rather than focusing on proprietary AI frameworks, the talk emphasizes interoperability, policy as code, and standards-driven governance, positioning IAM as the control plane for enterprise-grade Agentic AI platforms.

This presentation is complemented by a detailed article that expands on the architecture, design principles, and security implications of agent-driven systems in regulated and complex enterprise environments.

If you’d like to learn more, let’s stay connected through TwoGenIdentity and arrange a call to continue the discussion.