Workshop outline
Artificial intelligence is shifting from systems that generate content and recommendations to systems that act. Agentic AI can plan, use tools, and carry out multi-step tasks with limited human involvement, and this changes the nature of the risks organisations need to manage. This practical masterclass, designed for senior leaders and decision-makers, offers a clear and technically grounded view of what changes when AI moves from advising to acting with autonomy. Drawing on Gradient Institute's frontier research into the behaviour and governance of multi-agent systems, we examine how agentic AI systems work, where they fail in genuinely new ways, and what safer and more responsible deployment involves. No prior technical knowledge is required. The emphasis is on the practical mechanics of risk and control, giving you a structured basis for setting direction.

We will explore

• How agentic AI works, and why systems that act, rather than only generate or recommend, are meaningfully different from earlier AI and conventional software.
  The novel and system-level risks that emerge when autonomous agents use tools, access data, and interact with one another, including failure modes that do not appear in single-model systems.

• Practical, evidence-based approaches to identifying, assessing, and reducing these risks across the agent lifecycle, and how governance connects the technical, executive, and board levels of an organisation.

Practical outcomes

• Gain a clear and realistic understanding of what agentic AI can and cannot reliably do, and where its risks genuinely differ from those of earlier AI.

• Move from broad unease about autonomous systems toward a concrete and structured grasp of their failure modes and the approaches available to manage them.

• Leave with practical next steps for governing agentic AI, connecting technical safeguards with the governance decisions executives need to make, and an awareness of how these link to board oversight and technical implementation.

Who should attend
This workshop is designed primarily for senior leaders and executives responsible for AI strategy, risk, governance, and safe innovation as their organisations explore and adopt agentic AI. It will also be valuable for system owners, board members, and technical leaders who need a shared, evidence-based understanding of agentic AI risk. No prior technical knowledge of these technologies is required for participants.