Environmental uncertainties and hyperconnectivity force techno-socio-economic systems to introspect and adapt to succeed and survive. Current practice is chiefly intuition-driven which is inconsistent with the need for precision and rigor. We propose that this can be addressed through the use of digital twins by combining results from Modelling & Simulation, Artificial Intelligence, and Control Theory to create a risk free in silico experimentation aid to help: (i) understand why system is the way it is, (ii) be prepared for possible outlier conditions, and (iii) identify plausible solutions for mitigating the outlier conditions in an evidence-backed manner. We use reinforcement learning to systematically explore the digital twin solution space. Our proposal is significant because it advances the effective use of digital twin to new problem domains that have greater impact potential. Our novel approach contributes a meta model for simulatable digital twin of industry scale techno-socio-economic systems, agent-based implementation of the digital twin, and an architecture that serves as a risk-free experimentation aid to support simulation-based evidence-backed decision-making. We also discuss validation of this approach, associated technology infrastructure, and architecture through a representative sample of industry-scale real-world use cases.