At EVOLVE [25], Graeme Cox argued that AI was moving from co-pilot to autopilot, and that agentic systems would reshape how work gets done within two years. Twelve months on, he returns to mark his own homework. This session is a field report from the agentic era so far. Drawing on real deployments inside UK businesses, alongside the latest public developments in agentic AI, Graeme will explore what is actually working, what is failing in interesting ways, and what is already being rewritten around teams, oversight, accountability and decision-making. The talk will look beyond the hype to examine the practical reality of agents in the workplace. Where are they creating real value? Where are they exposing new risks? And has the idea of “human in the loop” already started to age faster than expected? Graeme will also explore the emerging regulatory picture, from the EU AI Act and US policy shifts to the UK’s increasingly serious work on AI safety and pre-deployment evaluation. For British businesses, the question may no longer be whether to deploy agents, but whether the UK is quietly moving into a more defensible position than either of its neighbours.