allow an organization to establish governance, prove reliability and build institutional confidence before extending the same capability into production environments, where data is more complex, equipment more varied and decisions carry physical consequences. For energy leaders, this raises a practical question: how can agents contribute more directly to the work of operating assets? The opportunity is to connect insight with execution. Identifying a potential equipment fault is one task. Preparing an inspection plan, assembling the relevant records and coordinating an approved response can extend that capability. The ambition is moving beyond simply another recommendation and towards automated actions, always with human oversight. Autonomous operations require explicit choices about authority. Policies must define what an agent may read, recommend, prepare and execute—and where a person must approve the next step. Access to asset information for example should not automatically confer permission to change equipment settings. Checks have to match the consequences of an action, with an accountable operational owner and a clear fallback when information is incomplete or conditions change. This is both a leadership priority and a technical one. Seventy percent of leaders say AI-integration work took up a substantial amount of their professional time over the past twelve months.Sixty-eight percent now spend a quarter or more of their working day on tasks substantially involving AI.Fifty-six percent say AI influenced a significant number of their professional decisions in the same period.Intelligence has reached the executive layer of organizations; extending it to the operating asset at scale is the work now in front of the sector. Ultimately, the end goal for greater use of agentic AI in energy should focus on expanded capability, not autonomy for its own sake. Each and every workflow should answer a practical question: what can the organization do better because intelligence can act within an agreed scope? For energy businesses, the ambition is faster coordination, better-prepared responses and greater continuity of operations. How the market reads the sector Market expectations are running ahead of deployments. Investors describe a more automated industry than operators report. On grid operations, investors place agentic deployment at 20%, against the 7% operators report for their own organizations. On generation and production, investors say 13%, against 8%.While the two measures are not directly comparable — investors are assessing the sector, operators are reporting their own operations, and the bases differ — it is a gap that the sector has both the opportunity and the evidence to close. • UNCHANGED: 49% of organizations ran AI across several business functions in 2025, 50% today. • NARROWED: The pilot stage. A third of organizations were piloting isolated use cases a year ago; today it is one in six. That ground has gone to integrated agentic workflows, which rose from 15% to 33%. • NEW: Where the agents actually run. Digital and technology functions lead at 55%, while generation and production stands at 8% and grid operations at 7%. The 2026 perspective “The question is no longer  what AI can do, but how we deploy it effectively in our operations. In Subsurface, by combining decades of geoscience knowledge and know-how, petaflops of data with next-generation AI models, we can support our geosciences experts in the exploration, characterization and development of oil and gas reservoirs. In industry, performance goes hand in hand with reliability, safety and clear accountability. AI can help us operate our assets better and make better decisions. But technology must remain at the service of people. The real value will come from combining the power of AI with human expertise. It is a matter of augmenting the people capacities at their service.” Patrick Pouyanné Chairman & CEO, TotalEnergies Powering Possible 2026 19