ABSTRACT:
Background: The pharmaceutical formulation development process has been traditionally based on sequential screening, expert judgment and iterative experimentation. This approach is costly and hard to scale up because of the complex relationships between the drug physicochemistry, excipient properties, processing conditions and critical quality attributes. Objective: This review critically discusses the role Artificial Intelligence (AI), Machine Learning (ML) and integrated experimental design can play in relation to data driven formulation development. Method: A narrative synthesis of peer-reviewed literature and pharmaceutical quality guidance using a structured approach was conducted, focusing on solid oral dosage forms, lipid delivery systems, polymeric delivery systems, manufacturing and model validation. Conclusions: Supervised learning for predicting dissolution, particle size, encapsulation efficiency and manufacturability and stability outcomes; Bayesian optimization and active learning for selection of informative experiments; and interpretability by hybrid mechanistic–data-driven models. However, data leakage, lack of external validation and poorly documented negative experiments, and changes between laboratory and commercial scales are still significant challenges. In Conclusion: AI can be considered an evidence-generating decision support tool within the Quality by Design framework, rather than a substitute for physicochemical understanding, confirmatory experiments and regulatory quality systems.