Implementing a clinical pharmacy support system with a centralized drug database poses some challenges. One challenge is parameterizing drug records and their attributes, mainly when the data originates from different systems with multiple data standardizations. In this context, it is essential to correlate a substance for each drug to alert potential drug interactions since it enables the identification of prescription problems such as drug duplications. Manually indicating the drug-related substance is error-prone and time-consuming, leading to security issues with respect to patient care. This work aims to develop and evaluate AI-based classifiers, including Machine Learning, Language Models, and Large Language Models, to automate and optimize the correlation between drugs and their active substances, enhancing the efficiency of clinical pharmacy processes and improving patient safety by accelerating drug interaction verification. We identified that the BERTimbau Base LM model presented the best F1 score result: 97.02, although it was the largest and second most expensive model. Thus, we selected the ML Logistic Regression model to experiment with an end user. With our model, the average time spent by the specialist decreased by 2/3. Our results maximize the specialist’s efficiency in performing the substance indication and improve patient safety.