As it continues to evolve with new and improved abilities, AI has the potential to play an even more valuable role in the energy system of the future. Beyond the development and deployment of AI’s current abilities, entirely new abilities are on the horizon. One example is agentic operation: autonomous or semiautonomous systems designed to make decisions and accomplish goals—on behalf of an individual— within specific environments.111 Technical research in the field of agentic systems has advanced greatly in recent years, and they are significantly more sophisticated than assistive systems, such as ChatGPT, that rely on human guidance to retrieve and process information. With the help of agentic AI, autonomous systems are rapidly advancing in areas like self-driving vehicles, complex games, and research assistance.112, 113, 114 With appropriate safeguards in place, agentic AI has the potential to unlock value in the vastly complex future energy system, particularly in grid management and demand response. In these areas, decisions must be While agentic AI has the potential to provide many opportunities, the potential applications in the energy sector are still in their infancy. More research and learning is needed to understand and refine its potential. For agentic AI systems to scale, especially in complex and technical fields like energy, they must undergo robust testing and be equipped with strong safeguards. made rapidly and accurately based on vast amounts of dynamic and interconnected data at a scale not possible today, even with assistance from AI. For example, in grid management, today’s AI can predict demand across various loads and forecast power availability based on factors like weather and grid congestion. Agentic systems could enable going one step further, assisting humans to automatically activate alternative power sources or storing surplus energy for future use. When it comes to demand response, today’s AI can gather data from devices like EVs, appliances, and smart meters to help human operators allocate power more effectively. Agentic systems, by contrast, could be instructed by humans on when to directly control devices, curbing energy use during demand spikes or charging batteries when energy is abundant. The EU Artificial Intelligence Act, which regulates AI models and systems, sets forth specific requirements for “high risk” AI systems, which include those used in the management and operation of safety components of critical infrastructure, like the energy sector. These requirements include obligations around risk assessments, data governance, transparency, and human oversight.115 In the energy sector, therefore, it is possible that certain agentic AI systems may need some level of human oversight, making them semi-autonomous rather than fully autonomous. 68 Powering Possible 67 AI has the potential to develop entirely new abilities that—if realized—would position it to play an even greater role in the energy system of the future. AI's potential to evolve more advanced abilities