Log transportation accounts for up to 40% of operating costs in the Canadian forestry sector Audy et al. (2023), yet optimization tools for transportation planning have seen little adoption in practice. Planners set them aside because they omit critical operational constraints, cannot evolve as business constraints change, and offer no way to recover when disruptions invalidate the plan. Natural Resources Canada (NRCan) designed and validated an end-to-end decision support system (DSS) for large-scale log-truck routing and scheduling. At its core, the system brings together three analytics and optimization contributions. The first is an exact and comprehensive formulation of the multi-period routing and scheduling problem on a time-space network, solved at industrial scale through a tailored relax-and-fix/fix-and-optimize decomposition. The second is an implicit constraint learning framework that embeds trained machine learning predictors into the model as linear constraints, so that plans self-adjust to evolving planner preferences without manual re-engineering. The third is a dynamic reoptimization algorithm that combines local branching with reinforcement learning to recover near-optimal schedules under disruptions such as road closures, truck breakdowns, or travel delays, while staying close to the original plan. Deployed as a web-based platform within NRCan, the system is estimated, in its first industrial validation with a major Canadian forest company, to generate approximately $8 million in annual savings and to avoid 4 800 tonnes of CO\(_2\) per year. National adoption is projected to yield up to $377 million in annual fuel savings and to avoid more than 500 000 tonnes of CO\(_2\) per year, directly supporting Canada's net-zero objectives.