This one day masterclass provides an intensive deep dive into the architectural and engineering concepts around Agentic AI. The course provides a refresher on AI security and the AI Security Services Reference Architecture, and then describes the evolution of LLMs to the current state of understanding of AI agents.
The course studies the risks of Agentic AI, both threat and opportunity driven, and derives the controls and enablers necessary to maintain agentic activity within risk appetite. It covers the governance for agentic AI in the form of an enhanced ISO42001 AI management system, with a delegation approach which takes autonomous agents into account. Agentic identity is discussed together with architecting agentic access and privileges. The course covers the Hindsight approach to agentic memory engineering and presents a series of agentic AI design patterns. It concludes by reviewing the evolving protocols for addressing agents across the internet.
