Agentic AI is entering the enterprise faster than it can be governed. Unlike the previous generation of AI that produced predictions for a human to act on, agents now act directly — making decisions, calling tools, and moving data across systems with limited human intervention. This is a structural change in how software behaves, and it breaks the governance assumptions that most organisations still rely on.
The visible result is sprawl: agents spun up in individual business units, each consuming budget, each touching sensitive systems, each a potential liability under regulation, such as the EU AI Act, with no central visibility, no shared accountability, and no portfolio view of value or risk. The enterprises that win will not be the ones that deploy fastest but the ones that build the system to scale agentic AI safely.
This talk makes the case that three capabilities must be built together, not in sequence: an AI Vision that sets direction and risk appetite; an AI Portfolio Management System that decides where to invest and keeps every agent visible, owned, and scored; and a Governance Structure that defines accountability and embeds proportionate, tiered controls into the delivery platform itself. The session closes with a pragmatic four-stage maturity path and a single first move that works for nearly every organisation in the room: making every agent visible.