This thesis addresses the research gap in the adaptive maintenance of timber structures affected by climate change, specifically focusing on decay from fungi. Traditional inspection methods are inadequate for such dynamic environmental impacts. The main aim is to enhance the precision and efficiency of maintenance plans using a novel approach: reinforcement learning (RL). This methodology allows for predictive maintenance strategies responsive to ongoing changes in timber's mechanical properties. The research employs RL to develop a predictive model that adapts dynamically to evolving conditions, primarily decay caused by fungi. The novelty lies in integrating machine learning with traditional timber engineering practices, providing a methodologically innovative approach to maintenance. The main outcome is a validated RL model that efficiently predicts maintenance needs, facilitating timely interventions that ensure the longevity and safety of timber structures, with potential scalability for global application in diverse climatic conditions.