E-COMMERCE ENTERPRISE RESILIENCE: MANAGEMENT MISALIGNMENTS IN GENERATIVE AI IMPLEMENTATION

Abstract

The article examines the managerial conditions under which generative artificial intelligence may strengthen or weaken the resilience of e-commerce enterprises. Access to generative AI is shown to be a resource rather than a ready-made capability, and its consequences depend on the management readiness of a particular process. A preliminary set of five organisational conditions of readiness is proposed: owners of AI output review, resources for review, a process recovery procedure, limits on actions AI may take without human involvement, and defined subject-matter boundaries with responsibility for the relevance and traceability of knowledge; each condition may be defined, resourced or tested in practice, which separates declared from actual readiness. The study is conceptual and applies comparative analysis, the typological method, structural and logical analysis, and graphical modelling. Three management misalignments distinguished by the subject of the managerial decision are identified: misalignments of expectations, of rules, and of work processes, the latter caused by the duality of their users, humans and models. For each, the mechanism weakening the capacity to absorb disruptions or recover and the conditions under which it does not operate are traced. When AI is added on top of an existing process, accelerating one stage merely shifts the process constraint to review stages, which calls for formally embedding the requirements of both users. A matrix of generative AI implementation combining process criticality, managerial readiness and a conditional risk appetite boundary is developed; only developing readiness, unlike changing the risk appetite, alters a process's capacity to withstand disruptions. The results help managers set the order of actions: in critical processes with low readiness, rules for AI use should precede expected acceleration.