The Digital Twin Organization (DTO) A Theory of Organizational State and a Computational Architecture for Predictive, Causally-Informed Change Management

Organizations change faster than the management systems built to guide them, yet most leaders still rely on retrospective dashboards and intuition to steer people through disruption. This proposal introduces the Digital Twin Organization (DTO): a continuously synchronized computational model of an organization's state, coupled to a temporal graph neural network that forecasts how a proposed intervention would propagate through the organization before that intervention is made. The central scientific contribution is not the simulation engine itself but a preceding theoretical object the engine is built to estimate: an organizational state vector S(t), a low-dimensional, empirically grounded, and interpretable representation of coordination capacity, communication health, workload balance, adaptive capacity, and structural cohesion, distinct from the raw interaction graph and explicitly defined in Section 4. The architecture (Section 8) specifies how S(t) is estimated, how forecasts propagate uncertainty rather than returning point estimates, how causal effects are identified from staggered intervention rollout rather than asserted from correlation, and how the recalibration loop absorbs both ordinary model drift and the reflexive drift that occurs once an organization's members know the model is informing decisions. The framework does not claim to predict individual behavior, model the meaning employees make of events, or replace human judgment; it is scoped explicitly as decision support that ranks candidate interventions by forecast outcome and uncertainty, leaving the choice to leadership. The study tests one confirmatory hypothesis, that DTO-informed decisions produce higher change-adoption rates and better-calibrated forecasts than three defined baselines (naive persistence, time-series regression, and blinded expert judgment), through a single deeply instrumented pilot followed by a confirmatory deployment. The result is intended as a validated architecture, an organizational state theory that is falsifiable independent of any specific model implementation, and founding empirical evidence for computational organizational design as a research area.