Case Studies

Evidence for mapping before AI-influenced work moves.

Three cases show how AI can affect naval maintenance, defense analysis, and human review when execution paths, assumptions, and approval boundaries are not visible enough.

Request Data Flow Mapping Review Evidence

Proof of Method

Evidence that execution paths have to be visible.

Each case shows where AI assistance meets human approval and authority, and where mapped controls can catch drift before it influences a real operating decision.

Operational systems Authority changes as work phases, materials, testing, and release conditions change.
Analytical systems A rejected concept can re-enter the baseline unless terminology is governed.
Interaction systems The interface itself can alter continuity even when the model is not factually wrong.

Operational Governance

Runtime Authority Enforcement in Naval Nuclear Maintenance

A naval availability scenario showing how AI assistance must remain bounded before it influences planning, material control, execution, testing, or final release.

Runtime authority
Dynamic re-authorization
Final release gates

Read case study

Analytical Governance

Unauthorized Concept Drift During AI Analysis

An analytical case in which an unauthorized framework term was introduced, rejected, and later reintroduced without traceability to the approved baseline.

Unapproved terminology
Traceability failure
Corrective action

Read case study

Human-AI Interaction Governance

Context Disruption Following User Interface Intervention

An observation report showing how a system-generated interface prompt disrupted conversational continuity and caused loss of shared attention.

Shared attention
Interface governance
No model failure required

Read case study