The Capacitated Vehicle Routing Problem (CVRP) is a significant variant of the vehicle routing problem that incorporates constraints related to customer demand and vehicle capacity. Owing to its extensive applications in logistics and transportation, CVRP has attracted substantial research attention, with numerous algorithms proposed from the perspective of intelligent search. A common solution strategy involves two phases: first, assigning customers to different vehicles to form feasible routes, and second, optimizing these routes. This paper presents a two-phase CVRP solution framework through the clustering concept with intelligent search to improve route planning. In the first phase, a set of clustering methods - fuzzy c-means, k-means, and k-medoids - combined with a nearest neighbor heuristic search, are applied to generate feasible routes for each vehicle. In the second phase, these routes are iteratively optimized using the Simulated Annealing (SA) algorithm. The process yields three distinct solution pathways: fuzzy c-means with SA, k-means with SA, and k-medoids with SA. For performance evaluation, 46 benchmark CVRP datasets from a publicly available library are used. Simulation results demonstrate that k-means with SA performs the best, surpassing the other two approaches and outperforming other clustering-based two-phase state-of-the-art algorithms in terms of solution quality.