Spec-Driven Operations (SDOps): A Specification-Centric Operating Model for Human, Software, and Agentic Organizations

Organizations increasingly execute work through hybrid systems composed of people, deterministic software, workflowplatforms, and artificial intelligence agents. Operational intent, however, is commonly fragmented across procedures,policies, tickets, dashboards, meetings, spreadsheets, application logic, and tacit knowledge. Business-process modeling andlifecycle management are mature fields [1,2,5]; artifact-centric research established operational specification and declarativelifecycles [21,22]; process-mining research established conformance checking, including multi-perspective, data-aware, andresource-aware variants [3,4,23,26,28,29]; workflow research established authorization and compliance constraints [30,31];and recent work addresses agentic business-process management and agent-behavior governance [17,18,25,27]. This paperdoes not claim novelty for those mechanisms or for the phrase 'spec-driven operations'. It proposes Spec-Driven Operations(SDOps) as a conceptual integration for one bounded business workflow. The theory separates five objects that arefrequently conflated in practice: a versioned Operational Specification S_v, an Executor Binding B_t, execution EvidenceEv_t, a contract-scoped Conformance profile C_t, and Economics K_t. The specification describes invariant operationalintent and constraints; a binding identifies a particular human, software, agent, or hybrid executor configuration and is validonly when its capability, authority, evidence, and risk properties satisfy the specification. Execution evidence is provenanceaware and distinguishes RECORDED, OBSERVED, DECLARED, and INFERRED claims; inferred claims cannot becomeauthoritative without validation and release authority. The running specification is continuously reconciled with observedreality and economic outcomes, yielding drift signals and governed proposals for S_(v+1). Four core invariants - executorseparation, epistemic non-promotion, historical governance reconstructability, and conformance-value separation - definethe proposed irreducible mechanism. The paper remains foundational rather than confirmatory: whether this integrationimproves traceability, governability, portability, adaptability, or economic performance remains an empirical question