Primary prostate cancer (PCa) cells can disseminate early and stay dormant in distant organs but reactivate later, causing lethal metastasis and recurrence. The mechanisms of PCa cell dissemination, dormancy, and reactivation are overlooked, especially whether and how various treatments affect these processes. Using C4-2B PCa subcutaneous xenograft mouse model, we found that in mice with xenograft tumor removed, the disseminated tumor cells (DTC) were exclusively enriched in the bones, more specifically, in the bone cortex instead of the bone marrow. We hypothesize that DTC are dormant in the bone cortex and proliferative in the bone marrow and developed an in vitro mixed co-culture model to determine the bone stromal cell type inducing the dormancy. We found that osteoblasts, but not other stromal cells induced C4-2B PCa cell into a dormancy-like state, i.e., slowed proliferation without increased apoptosis, G0/G1 arrest, increased expression of dormancy marker, and decreased expression of proliferation markers. Through RNA-sequencing, gene ontology, and public dataset analyses, we found a reverse correlation between mitochondrial gene expressions and enrichment of mitochondria-related functions and PCa progression. We further predicted dormancy-mimicking drugs using a novel artificial intelligence (AI) platform. One of the candidates, PF-562271, a focal adhesion kinase (FAK) inhibitor, was validated as dormancy-mimicking in vitro. Altogether, our study revealed for the first time the effects of primary tumor removal on PCa cell dissemination, profiled the osteoblasts-induced C4-2B cell dormancy, and induced PCa cell dormancy in vitro using AI predicted drugs, which could potentially inhibit PCa bone metastasis progression and recurrence clinically.