Metabolic engineering is entering a new phase of programmability, evolving from mechanistic models toward AI-driven systems that can design, test, and refine biology with increasing autonomy. Early efforts to combine genome-scale modeling with machine learning began to improve genotype to phenotype prediction, hinting at a more predictive and designable biology. Today, that paradigm is advancing into a new layer. Agentic AI systems are beginning to orchestrate the full design, build, test, learn cycle. These platforms integrate experimental data, automation, and decision-making into continuous closed loop workflows, enabling faster iteration and more intelligent exploration of biological space. This session explores the next frontier of metabolic engineering, examining long standing bottlenecks such as limited data, combinatorial design complexity, and slow iteration cycles, and how AI native, end to end platforms are transforming pathway design, strain optimization, and scalable biomanufacturing.