Analytical Governance: From User Intent to Governed Analytical Execution

Abstract

Analytical governance is often applied after a query already exists. By then, a more important decision may already have been made implicitly: what analytical object the user intended, and whether that object is entitled to be served. This paper develops analytical governance as an architecture from user intent to governed execution. It identifies two gaps. The intent gap separates ordinary human language from an explicit analytical request. The servability gap separates a faithfully framed request from one that is both adequately supported by current data and analytically established under the laws governing that data. Closing those gaps requires an authoritative middle representation between prose and procedure. Free-text documentation carries meaning but is weak as machine authority. SQL and code are executable but may encode a procedure while leaving analytical identity implicit. A machine-readable governed analytical model makes the requested object, its support requirements, and its lawful derivations independently adjudicable. Only after servability is established should three remaining risks be evaluated: cost, security, and result/application risk. Analytical violations are not risks that can be accepted; they are unestablished requests. The paper further distinguishes the four analytical serving outcomes—Serve, Disclose, Clarify, and Refuse—from Escalate, a governance-process outcome at the constitution's edge; introduces a conformance test for governed analytical systems; distinguishes analytical governance from data governance; and positions analytical governance as a multi-world meta-framework that keeps capability distinct from legitimacy across within-world and crossing contexts.