The formal definition of Contract Intelligence™. The standard every GovCon AI claim must be evaluated against.

Papers 1 and 2 establish the theoretical foundation — the GovCon-as-contract-execution-system argument and Governing Object Theory. Paper 3 is the definitional anchor. It names Contract Intelligence™, defines it with the precision required to withstand technical scrutiny, establishes the five architectural properties, and presents the seven implementation conditions. Every other paper in Canon II references this definition. The governing paper is the definition paper, not the foundational argument paper.

Explainability and responsibility are one of the five properties (Deterministic Audit Traceability) — necessary but not sufficient. Contract Intelligence™ additionally requires that the AI operates inside a contract-governed data model (Property 1), that operational rules are inherited from the contract at entity creation (Property 2), that contract state changes propagate immediately via typed events (Property 3), and that every recommendation is evaluated against live policy constraints before surfacing (Property 4). An explainable AI system that lacks Properties 1–4 satisfies one condition in isolation while failing the standard.

xpdOffice is the reference implementation — the platform built from the ground up to satisfy all five properties and seven conditions simultaneously. The standard is stated precisely enough to evaluate any platform against. A platform built on a contract-governed data model with event-driven propagation, operational inheritance, policy-grounded AI inference, and deterministic audit traceability would satisfy the standard. The distinction is that xpdOffice is the only GovCon platform architecturally designed from the start with the contract as the governing object — rather than retrofitting contract governance onto a ledger-centric or project-centric architecture.

Policy-grounded means every AI recommendation is evaluated against the applicable FAR/DFARS cost allowability constraints, CLIN ceiling headroom, LCAT qualification requirements, and period of performance boundaries of the governing contract — before the recommendation is surfaced to a user. The AI cannot produce a recommendation that violates a contract constraint without explicitly flagging the violation and citing its regulatory source. This is not a post-hoc filter applied to AI outputs. It is a structural evaluation that occurs inside the inference architecture before any output is generated.