Abstract
Organisations have begun to purchase reasoning the way they long purchased electricity: metered, priced per unit and elastic on demand. Large language models sold through per-token interfaces let a firm dial the depth of machine deliberation applied to any problem, from a shallow single pass to prolonged multi-step reasoning, at a price that scales with depth. Management theory is unprepared for this. Its foundational treatments assume cognition is scarce because human attention is scarce, and its resource theories treat computational capacity as background infrastructure rather than as a governed input to advantage. This conceptual paper introduces the construct of governed cognitive capacity, defined as an organisation's metered stock of machine deliberation together with the routines that shape its demand, allocate its depth and assure its outputs. Drawing from resource orchestration, the attention-based view, bounded rationality and the emerging literatures on delegation to agentic artefacts, six propositions are developed linking capacity governance to value capture, including an inverted-U relation between deliberation depth and net value conditional on output verifiability, an allocation-not-access locus of competitive advantage and a transaction-cost boundary for external agent services. A three-plane governance framework - demand shaping, allocation and assurance - organises the propositions for practice. Contributions to theory, implications for managers and an operationalisation agenda are discussed.