Agency is not causality or causal power in the modern sense. Acting and doing are distinct from mere occurrence of an event by virtue of their meaning or interpretation.
The modern notion of causal power is something blind, mechanical, taken over from mathematical physics. Events follow one another in accordance with mathematical law. Even physical causality, though — insofar as we seek to understand it — is more about mathematical form than about the operation of raw power.
Talk about agency belongs in the register of semantics rather than physics in the modern sense. This takes us even further away from considerations of raw power.
Greek mathematics was really just getting started in Aristotle’s time. Aristotle correctly judged that that mathematics did not begin to address interesting problems of the intelligibility of becoming. Accordingly, he developed a discourse about the intelligibility of becoming in other — we could say, more semantic — ways. For Aristotle, even physics is fundamentally a semantic investigation. As Leibniz saw, there need be no contradiction between Aristotle’s semantic physics and modern mathematical physics. They simply investigate different things.
Aristotle originally introduced talk about agents, and did so in the context of his semantically oriented physics. From early modern times, physics moved to a mathematical approach that was immensely fruitful. At the same time, the Aristotelian semantic-physical notion of agents was (badly) translated into a register of mathematical-physical causes. (See also Aristotelian Matter.)
Because of its fundamentally semantic character, the semantic-physical notion of agency was well suited to be extended to social and ethical contexts. Its mathematical-physical analogue is not nearly so well suited. It took the monumental achievements of Kant and Hegel to begin to restore a semantic notion of agency. (See also Expansive Agency.)