Securing Agentic AI Without Slowing Down Innovation: Fireside chat

Shadow AI is a growing challenge because the barrier to building AI agents has never been lower, with no-code and low-code options commonplace. This means adoption can quickly outpace an organisation's visibility of what's happening. Before you can govern agent access, you need to know what agents exist, where they're operating and what they're connecting to.

Not on its own. By the time an organisation needs to stop an agent, damage may already have happened, making continuous visibility and early intervention critical.

Human oversight can't mean approving every action when we're talking about agentic scale. Organisations need clear boundaries for routine activity, with human intervention focused on unusual behaviour or high-risk tasks. 

Potentially. The demands of agentic AI can expose weaknesses in existing human identity governance, which could ultimately drive organisations to address gaps that have previously gone unnoticed or unresolved.

An agentic gateway sits between AI agents and the systems they need to access, providing a central point to manage authentication and enforce access policies. This means developers don't need to build identity controls into individual agents, while identity teams can maintain control over what agents are allowed to access.

A good place to start is with discovery. Understand which agents and platforms are already in use, then prioritise a small number based on value, use or risk. From there, identity, security and agent-building teams can work together to establish the controls and build a blueprint for wider adoption.