Selecting an optimal alternative in multi-criteria decision-making often involves maximizing benefit criteria and minimizing cost criteria. However, in many real-world scenarios, decision-makers have specific reference values in mind rather than seeking extreme optimization. This paper introduces a proximity-based MCDA approach that extends traditional distance-based methods by allowing certain criteria to be pointing at preferred reference values while still optimizing others. Unlike conventional techniques that construct ideal and anti-ideal solutions, the proposed method prioritizes alternatives that deviate the least from a user-defined reference model. The approach was tested in an empirical office space selection study, demonstrating its effectiveness and stability. A comparison with TOPSIS confirmed its reliability, while an AI-based weight evaluation experiment highlighted both the potential and risks of using large language models (LLMs) in MCDA applications.