The integration of large language models and agentic AI systems in smart grids has gained significant attention, with applications in forecasting, optimization, and control. These AI systems leverage external tools to plan and act in technical domains, enabling the automation of complex workflows. By wrapping trusted solvers behind language interfaces, agentic schemes can orchestrate multi-step workflows, enhancing the efficiency and reliability of smart grids. The use of LLMs and agentic AI systems in smart grids has the potential to revolutionize the way energy is managed and distributed, with implications for policy, security, and workforce dynamics1. As the smart grid ecosystem continues to evolve, the development of unified architectures and applications for LLMs and agentic AI systems will be crucial for realizing their full potential. The application of these AI systems in smart grids matters to practitioners because it can significantly impact the security and reliability of energy distribution, making it essential to prioritize the development of robust and secure AI architectures.