Agentic AI as an Emerging Economic Actor in Sustainable Energy Systems

Abstract

 Artificial intelligence is moving from tools that people invoke toward autonomous agents that perceive, decide, and act with limited human oversight. This shift introduces a new economic actor into energy systems: software that draws electricity through computation, can sense and participate in markets, and may either create or destroy value relative to the energy and carbon it consumes. Work on computing demand, data centres, demand response, autonomous agents, and carbon markets has
grown in parallel but remains disconnected, leaving the underlying economics unexamined. This review consolidates these strands into one framework and proposes the economics of autonomous energy actors as a subfield of energy economics. The evidence is organised around five themes: agents as a distinct demand sector; their interaction with wholesale and retail electricity markets; agent-mediated demand response; the value, productivity, and carbon performance of energy-aware agent economies; and autonomous participation in carbon markets. Two research streams are named, agentic computing demand and agentic demand-response economics, and three measurement concepts are defined, energy return on intelligence, carbon return on intelligence, and agent energy productivity, each with units, a worked illustration, and explicit limitations. A coverage analysis shows that prior reviews treat these themes partially and from a technical rather than an economic standpoint. The review maps
research methods to questions, catalogues citable data and benchmarks, weighs competing policy options, and sets out ten open questions. It is written for energy economists, electricity- and carbon-market designers, and policymakers who will confront autonomous demand sooner than the literature implies.