From Prompt to Performance in Tourism SMEs: Human-AI Orchestration and the Productivity-Capability Paradox

Generative artificial intelligence (GenAI) gives tourism small and medium-sized enterprises scalable cognitive and creative support, but access alone does not ensure sustained advantage. This conceptual article asks how firms can realize productivity gains while preserving the human and place-based capabilities required for future adaptation. An integrative, theory-building review connects tourism research with resource and asset orchestration, AI readiness, human-AI complementarity, organizational routines, and dynamic capabilities. Human-AI orchestration is proposed as a workflow-level configuration of task decomposition, contextualization, verification, and learning. Its contribution lies in coordinating established mechanisms rather than claiming that its dimensions are individually new. The prompt-to-performance framework distinguishes a capability-accumulation loop from a delegation-driven erosion loop. A proposed stock-and-flow mechanism explains switching between them: erosion dominates when losses from displaced practice and knowledge discontinuity exceed learning and replenishment; verification becomes vulnerable when retained capability falls below task-dependent requirements. Seven proposition families specify untested mediating, moderating, nonlinear, multilevel, and temporal relationships, with disconfirming observations. Sector-specific boundary conditions concern perishable service commitments, situated experience, partner interdependence, and concentrated expertise. Four configurations and a foresight matrix support contingent organizational choices. The contribution is a temporally explicit extension of automation-augmentation theory that separates immediate assisted performance from the evolving capacity to verify, recover, and adapt.