Jon W. Hansen, FCIPS · Procurement Insights | Hansen Models™
Julie Simou, Chief Data & Analytics Officer, asked one of the better questions I have seen this year about agentic AI: we talk a lot about controlling what an agent may do, but who controls what it is allowed to trust?
Her answer is a two-plane model. A control plane governs the action. A data plane governs the context. Her post has drawn more than a hundred comments, and it deserves them. An agent can hold every right permission and still act on context that is incomplete, inconsistent or out of date.
Original graphic and model: Julie Simou, on LinkedIn.
The plane underneath
Both of Julie’s planes govern the agent. What neither shows is where the data comes from.
Data is the residue of how people actually work. In 1998, the delivery data at the Department of National Defence showed orders arriving late in the day. It did not show why: technicians were holding orders to protect their service-call numbers. The data was accurate. Every order really did arrive late, and any data governance check would have passed it. But an agent trusting that data would have learned the wrong lesson, that the suppliers were slow, when the real cause was how the technicians were measured.
That is the operating plane, and it sits underneath the other two.
Guardrails and fences
It also changes where guardrails belong. Guardrails usually go in after an agent has been sent out without real control over it, dispatched within a frame people know, and that frame is often limited. They are useful. But they are not a firm fence.
A firm fence can only be built by understanding how the business actually works: who does what, why, and where the workarounds live. That is why so few get built, and why shadow operations have always existed.
So before asking what an agent may do, and what it may trust, there is a prior question.
How does the work really happen?
My thanks to Julie Simou for a post that moved the conversation forward. The third plane is offered in the same spirit.
Truth Is Believing. Accuracy Is Knowing. Outcome Is Proof.™
-30-
Related
Agentic AI Needs Two Planes. It Rests on a Third.
Posted on October 2, 2026
0
Jon W. Hansen, FCIPS · Procurement Insights | Hansen Models™
Julie Simou, Chief Data & Analytics Officer, asked one of the better questions I have seen this year about agentic AI: we talk a lot about controlling what an agent may do, but who controls what it is allowed to trust?
Her answer is a two-plane model. A control plane governs the action. A data plane governs the context. Her post has drawn more than a hundred comments, and it deserves them. An agent can hold every right permission and still act on context that is incomplete, inconsistent or out of date.
Original graphic and model: Julie Simou, on LinkedIn.
The plane underneath
Both of Julie’s planes govern the agent. What neither shows is where the data comes from.
Data is the residue of how people actually work. In 1998, the delivery data at the Department of National Defence showed orders arriving late in the day. It did not show why: technicians were holding orders to protect their service-call numbers. The data was accurate. Every order really did arrive late, and any data governance check would have passed it. But an agent trusting that data would have learned the wrong lesson, that the suppliers were slow, when the real cause was how the technicians were measured.
That is the operating plane, and it sits underneath the other two.
Guardrails and fences
It also changes where guardrails belong. Guardrails usually go in after an agent has been sent out without real control over it, dispatched within a frame people know, and that frame is often limited. They are useful. But they are not a firm fence.
A firm fence can only be built by understanding how the business actually works: who does what, why, and where the workarounds live. That is why so few get built, and why shadow operations have always existed.
So before asking what an agent may do, and what it may trust, there is a prior question.
How does the work really happen?
My thanks to Julie Simou for a post that moved the conversation forward. The third plane is offered in the same spirit.
Truth Is Believing. Accuracy Is Knowing. Outcome Is Proof.™
-30-
Share this:
Like this:
Related