An agent identity is not enough.
Someone has to own what the agent is allowed to do.
I have seen organizations accumulate service accounts nobody really owns. AI agents make that pattern far more dangerous.
Registering an agent as an identity is only the first step. The harder question is ownership. Who approves its tools? Who reviews its permissions? Who gets paged when it causes damage? Who decides when its access expires? Who checks whether it is still solving the right problem? The work that platforms like Okta's agent-identity offering point toward is the right direction — agents as first-class identities, with named human owners and scoped, short-lived credentials instead of hardcoded ones.
But identity is the easy half. The hard half is accountability. An agent can authenticate perfectly and still take an action nobody authorized, because authentication answers "who is this" while ownership answers "who answers for what it does." A service account with no owner is a latent incident. An agent with no owner is the same thing, moving faster and making more decisions.
Without ownership, an agent is just automated ambiguity with credentials.
The question is not whether the agent can authenticate. It is who is accountable after it acts.