A new hire sits down on a trading floor with credentials, a desk, and a screen full of live data. Nobody has told them what they can approve on their own.

No limit. No guidance on when to ask for help, or when to stop and check with someone senior. So one of two things happens. They freeze and ask permission for everything, which makes them useless. Or they guess and approve something they shouldn’t have, which makes them dangerous.

Give it a week and the floor adjusts. Someone takes them aside, explains the boundaries, and things settle.

Now replace the new hire with an agent

Same missing rules. But the agent can’t read the room, can’t sense an awkward silence, and never feels unsure. It acts at speed and at scale, every second, with no hesitation and no lunch break.

Unclear rules cost a new hire a slow first week. Unclear rules can cost an AI agent a quarter of bad decisions before anyone notices.

So who owns it

The person whose number moves when it works.

IT installed the tool. IT doesn’t miss a target when the tool underperforms. Give the decisions to whoever does.

One sales team proved it. Their AI tool sat with IT while it shaped sales output. Results stayed flat and nobody adjusted anything. Then ownership moved to the sales lead, whose target depended on the tool performing, and she made three changes in her first week that IT had never considered.

IT never felt the pain of a missed number. She felt it every day.

Ownership follows outcomes, not technology.