Addressing the industry bottlenecks of low efficiency, strong subjectivity, and difficulty in integrating multi-source data in the preliminary assessment of old residential community renovations, this paper proposes a fuzzy comprehensive evaluation model based on multi-modal data fusion and AI-driven technology. The model establishes an evaluation system covering 17 indicators across seven dimensions: safety, livability, sustainability, economy, sociality, technology, and resilience. The Analytic Hierarchy Process (AHP) is used to determine indicator weights, and a multi-modal feature extraction module is designed to process image data, text data, and structured data separately. Building on this, the evaluation output is linked to renovation cost estimation, forming an integrated “evaluation-cost” output. This paper elaborates on the model’s overall architecture, the construction of the indicator system and weight calculation methods, multi-source data fusion strategies, and cost association mechanisms. Typical computational examples validate the feasibility of the model process and the rationality of its decision-making logic. The results indicate that the model can effectively integrate multi-source information, output comprehensive scores and renovation grades, and provide corresponding cost range recommendations, offering an engineering-ready intelligent tool for scientific decision-making in old residential community renovations.