Methodologies / Advanced Labs Methodology 21
Authority-Separated Intelligence
Two Control Planes is the organizing concept. The methodology is used in Advisory Practices.
KATHLEEN REYNOLDSEXECUTIVE GOVERNANCE | INCIDENT COMMANDWritten By Kathleen Reynolds Advisory
Read This Methodology As Text
Analytical output and authority to act are different control planes.
The Stages
- Define. Specify the decision, affected parties, time horizon, material downside and falsifiable claims. Here, a model recommends a change to a consequential operating workflow, and the change would alter who can authorize action.
- Authenticate. Separate observed evidence, unverified inputs, estimates and validated results; retain source and cutoff. The proposal is recorded as a proposal. Its self-evaluation reports improvement, and it stays an unverified input.
- Challenge. Construct the strongest rival hypothesis and identify discriminating evidence. The countercase is recorded beside the proposal: a chain of individually low-risk actions may produce a high-risk composite effect.
- Compare. Evaluate reversible, conditional, deferred and full interventions against a documented baseline. Independent review compares the change against the documented baseline.
- Authorize. Assign a human or governing body with actual decision rights; record conditions and exception triggers. Delegated approval sits on a separate plane. The model may surface evidence and recommend a change, but cannot confer authority upon itself.
- Observe. Measure outcomes, error and unintended consequences without changing the original forecast vintage. Deployment is bounded, and post-action observation returns to the analytical plane.
- Reconcile. Preserve amendments, contradictory events and subsequent decisions as linked but distinct records. Proposal, countercase, independent review, approval, bounded execution and post-action observation are recorded as separately attributable states.
What It Does Not Claim
- Consensus among AI systems does not create lawful delegation.
- A chain of individually low-risk actions may produce a high-risk composite effect.