What Is Agentic Identity?
Agentic identity is the identity an autonomous AI agent authenticates and acts through as it executes tasks across systems on its own, without a human directing or approving each individual step. It differs from a conventional AI application, where a human typically reviews the output before anything happens: an agentic system can read a support ticket, query a database, and issue a refund in one uninterrupted sequence.
Two identity problems stack on top of each other. First, an agent typically runs under a service account or API credential it inherited, which means it can reach anything that account could already reach, often far more than its specific task requires. Second, an agent’s work happens in a session that can start and finish in seconds or minutes, faster than most identity systems, built around persistent, slowly-changing accounts, were designed to capture.
The result is that an agent’s real privilege footprint is usually invisible until someone maps what it inherited from its underlying credential and traces what it actually did during a specific session. Knowing a "bot" exists in the account inventory answers almost nothing about what it’s capable of doing, or what it already did.