Two especially important challenges that are quite often overlooked are interpretability and extendibility. Increasingly, AI is deeply woven into the fabric of modern manufacturing-automation efficiency and optimization of processes like predictive maintenance and quality control. Yet especially for industrial applications of AI, the challenges of interpretability and extendibility need to be sufficiently appreciated. For instance, in manufacturing, the primacy of transparent operations, occupational safety, and regulatory compliance would make the development of interpretable models of AI rather critical. Given the risk of a lack of interpretability, and specifically for complex AI models like deep learning, agents who make the decisions themselves will not be able to understand or justify their predictions based on the AI. This chapter attempts to show growing demands on manufacturers to produce an interpretable AI system by describing how there is increased ability to foster trust and collaboration between AI systems and human operators as computation improves decision-making processes. Extendibility: It describes the ability of AI models to adapt in the changing and evolving conditions of manufacturing. Production processes continually change themselves in terms of new product lines, changes in market demands, and modifications in operating conditions. The extendible AI models adapt the new tasks or datasets with minimal retraining and thus reduce the cost and time to update or modify AI systems. The techniques that will be discussed in detail are transfer learning, modular architectures, and continuous learning in an insight of how AI models might be at a good level of performance yet flexible enough for new challenges and environments. This chapter discusses representative techniques in constructing interpretable and extensible AI models, including decision trees, rule-based systems, LIME (Local Interpretable Model-Agnostic Explanations) and SHAP (SHapley Additive exPlanations), case studies referring to the automotive, electronics, and food manufacturing industries, showing how the model is used in actual applications, from predictive maintenance to quality control, for example. Thirdly, it addresses issues like the integration of AI systems with legacy manufacturing infrastructure, the balancing of interpretability with performance, and the scale of AI models for real-time high-volume data processing. The book concludes with future research directions, envisioned to enhance the role of AI further in transforming manufacturing processes, this time focusing on transparency, adaptability, and efficiency.