Intelligent Memory Is All You Need: The Age of Cyborg Organizations is beginning

AI systems can recommend, coordinate, and execute organizational action, but technical capability alone does not determine who should hold the rights to act. We develop an account of persistent organizational memory as action infrastructure: retained experience can shape the allocation of present authority, making human–AI agency state-dependent. Agency is represented as six separable rights: Decide, Execute, Supervise, Approve, Intervene/Escalate, and Override. A randomized vignette experiment with 800 participants varies documented organizational history while holding the current task, formal policy, and described AI capability constant. Average AI-involved rights allocation is: 65.2% under positive relevant history; 21.3% under adverse relevant history; 37.2% without relevant history; 44.6% under verified but irrelevant history. Archival Wikimedia governance episodes provide complementary evidence that prior operating history and institutional precedent are explicitly invoked in real automation decisions, with task novelty and risk qualifying their use. A controlled procurement experiment produces history-sensitive rights profiles across two open-source model families, although the specific reallocations differ between models. Process-log evidence supports the structural decomposition of authority across organizational roles. Together, these results support a temporal account of distributed agency in which persistent history informs who may act now. A theoretical extension reserves ultimate authority over delegation rules to the human hierarchy while treating AI operational agency as conditional and revocable. The evidence directly concerns memory-to-rights allocation; the recursive extension through action, outcomes, accountability, and memory update remains a proposition for longitudinal research.