INSIGHT RECOMMENDATION Agentic AI is moving from experimentation to enterprise deployment. Adoption has more than doubled since 2025, but remains concentrated in digital and technology functions. The next opportunity is extending agents into operational environments as governance, oversight and trust mature. Build for scale, not just deployment. Establish trusted data foundations, governance and operating models that allow agents to be deployed consistently across business and operational environments. The top two barriers are one problem. Cybersecurity and data quality lead the list for a second year. Only 8% of organizations describe most of their data as AI-ready, and leaders prioritizing data integration put security and compliance ahead of accuracy. Run data readiness as a governance program, not an IT upgrade. Name owners for the data agents will act on, define who may move it across which boundaries, and resolve permission before pipelines. Capital and policy both point at the network. Grid modernization is the leading infrastructure priority, and grid investment is leaders’ first policy ask. Annual global investment in grids remains far below what flows into generation11. Invest for system resilience, not single-point optimization. Meeting future demand will require balanced investment in generation, networks, storage and digital infrastructure to improve capacity, reliability and flexibility across the energy system. Physical intelligence is arriving, and its value is protective. Thirty percent of operators are piloting autonomous and robotic systems and 24% use them routinely. Leaders expect the greatest value in efficiency and predictive action: protecting critical assets before they fail and keeping people away from harm. Start where it removes risk. Deploy physical AI first in hazardous, remote and hard-to-reach work, and manage it as an extension of asset integrity and safety as much as a robotics program. The sector is funding its own capability. Seventy-seven percent of organizations increased upskilling spend, leaders are spending more time learning about AI, and skills programs rank last among the policy actions leaders ask for. Invest deeply in in-house capability and make it measurable. Agree expected levels of AI fluency by function, track them as a safety competency, and train people to work across energy and intelligence systems. Accountability is agreed; liability is not. Forty-four percent of leaders want AI decisions to be transparent and accountable. Only 18% ask for clear lines of liability when a system fails, the second-lowest of six governance priorities. Settle liability before autonomy scales. Write responsibility for automated decisions into contracts, partnership agreements and operating procedures before AI decisions become more consequential. 11 World Energy Investment 2025, IEA, June 2025 Powering Possible 2026 37