Published as a Sagetap perspective: as AI agents and automation take on real work, they do it as non-human identities — and most organizations have spent far less time governing those than they have governing employees.
Agents are identities
Every agent, service account and automation is a principal with credentials and permissions. If you can't answer who owns it, what it can reach, and when its access expires, you don't have an AI problem — you have an identity problem wearing a new hat.
The gaps you already had, multiplied
Over-broad scopes, credentials that never rotate, and no offboarding path scale badly when software can create workloads faster than people can review them. Autonomy raises the cost of every existing IAM shortcut.
Fundamentals still win
Least privilege, short-lived credentials, clear ownership and auditability aren't new ideas — they're just newly urgent. Getting them right for non-human identities is the unsexy prerequisite for deploying autonomous systems safely.