When AI Starts Acting - UKAI Agentic AI Roundtable

On 25 March 2026, UKAI convened a small group of founders, technologists, investors and operators at HSBC Innovation Banking for a closed roundtable exploring a simple but increasingly important question: what changes when AI systems move from tools to agents?

The discussion focused on how agentic behaviour is already beginning to appear inside organisations often not through a single deliberate decision, but gradually through integrations, orchestration layers and workflow automation. Many organisations are now finding that systems are not just producing outputs, but initiating actions, triggering processes and operating across tools with increasing autonomy.

The conversation explored where organisations are already seeing this in practice, how teams are thinking about delegation between humans and machines, and what meaningful oversight looks like when systems are able to act within live environments. A significant part of the discussion also focused on accountability particularly in situations where systems act across teams, platforms or organisational boundaries.

One of the clearest themes from the session was that agentic behaviour is already here, but governance, language and operating models are still catching up. Many organisations are currently operating in a grey area where systems are taking action, but responsibility, visibility and intervention models are still being worked out in real time.

The session marked the first formal roundtable of the UKAI Agentic AI Working Group, and the insights from the discussion will help shape the group’s next phase of work, focusing on practical approaches to oversight, accountability and operational control as AI systems begin to take a more active role inside organisations.