Knowledge Operations: A Capability Model for AI Systems

Two AI systems can both be called "RAG" while performing fundamentally different kinds of knowledge work; conversely, two systems can use different substrates—vectors, SQL, graphs, or tools—while satisfying the same workload requirement. This mismatch makes architecture-centered labels a poor guide to capability. This paper proposes a capability model for knowledge-augmented AI systems based on the operations they perform reliably, rather than on the data formats, stores, or orchestration patterns they use. We define Knowledge-Augmented Systems (KAS) as AI systems that combine language models with external knowledge, structure, computation, or execution in order to answer questions, support decisions, or perform workflows grounded in authoritative information. The model profiles systems by knowledge operations: retrieving, scoping, interpreting, combining, computing, traversing, planning and acting, governing, and evaluating. It introduces seven operation archetypes (K0–K6), a governance scale (G0–G5), and an evaluation discipline applied across every operation. The result is a multi-dimensional capability profile rather than a single maturity level. A system may be strong at scoped retrieval and cross-source synthesis while having no need for relational traversal or exact computation; another may be computation-dominant with little need for synthesis. The framework retains prescriptive force through workload fitness: a profile is insufficient when the task demands operations the system cannot perform, when governance falls below the workload's risk requirements, or when the system introduces unnecessary operations whose latency, cost, or complexity the task cannot absorb. KAS profiles therefore support evaluation of systems per workload, not per industry, vendor stack, or architectural fashion.