Competence and Coherence: The Structural Foundations of Enterprise AI Readiness

Abstract

Large language models (LLMs) possess the competencies required for enterprise knowledge work: writing, analysis, reasoning, and the ability to operate tools programmatically. Knowledge work, however, requires more than competence. It requires context: knowledge of who, what, when, why, and in what state. Competence is a property of the model, context is a property of the enterprise's data architecture. This paper presents a formal decomposition of that architecture. We reduce enterprise operational data to a small set of ontological primitives, typed entities participating in defined state machines. We show how the enterprise software stack fragments this ontology across proprietary systems and demonstrate that human workers have historically served as an implicit integration layer bridging the resulting contextual gaps. Building on enterprise architecture research establishing that architectural choices constrain execution capability, we introduce a two-layer AI readiness framework: state readiness (is the current ontological state complete, coherent, legible, and operable?) and transition readiness (can the enterprise trace decisions through state mutations over time?). We identify coherence, the standardization of the enterprise's ontological representations, as the load-bearing dimension, and define AI readiness as a function of that coherence.

Read the full paper on SSRN