Who Is an AI Agent Acting For?

Software agents are beginning to communicate, decide and act on behalf of people and businesses. The systems they interact with were designed around a simpler assumption: that a human being sits behind every request.

That assumption is breaking. When an agent books, disputes, refunds, escalates or transacts, the counterparty needs to know three things—who the agent is, who it acts for and what it was permitted to do.

This note will examine how agent identity, delegated authorization and accountability may develop, and where durable companies could be built as those primitives take shape.