THE REGULATORY STACK: REGULATING THE ENDS AND THE MEANS

Most organizations operationalize their obligations by converting them into controls. This has been the dominant approach in information security for years, and it is now being carried into AI governance.
The difficulty is that obligations come in different regulatory designs. Some require outcomes, some require performance, some require practices to be followed, and some require rules to be adhered to. These are the ends and the means, from macro to micro, and each one needs its own kind of regulation.
The diagram shows how this works for AI in the Operational Compliance Model (OCM), developed by Lean Compliance over the last ten years. Each function in the organization operates as a regulator, a pattern drawn from cybernetics, the science of regulation. A regulator has a purpose, a model of what it regulates, an observ
er, the intelligence to decide when and where to intervene, and a channel of control. Governance leads, programs steer, systems control, and processes execute.
Each regulator feeds forward to set the purpose of the one below it. It takes feedback to measure whether it is achieving and maintaining what it exists for: effectiveness, performance, conformance and adherence.
When obligations are mapped directly to controls, every obligation is treated as a rule. A policy, a named owner, an output filter and a shutdown switch are each tested for whether they exist and operate. That is adherence, the bottom loop of the stack. Effectiveness, performance and conformance are left to chance, and organizations end up with a constellation of controls in the hope that they will produce the outcomes they promised.
We often talk about governing the ends and the means. The OCM is how you operationalize it.





