Onboard the New Hire
Eighty-eight percent of companies have had an agent incident — fewer than one in four can tell you which agent did it.
Wrote a piece today about AI agent security and the onboarding process behind it.
Okta shipped an agent identity platform. Gravitee published a report showing 88% of organizations have had a suspected or confirmed agent security incident. Only 22% treat agents as identities. Those two numbers sit side by side and tell the whole story.
Companies are deploying agents the way they’d deploy a script. Connect to API, hand it a long-lived token, move on. No directory entry. No scoped permissions. No human owner. When something breaks, the first question is always the same: “Who set this up?”
That’s a security failure rooted in onboarding. We know how to onboard workers. Credentials, scoped access, an owner in the org chart, a kill switch if things go sideways. It’s established process — applied to a category of worker most companies haven’t acknowledged exists.
Missing process contributes to the 88% incident rate. Forty-four percent of organizations have zero governance for their agents. Governance is entirely absent.
The companies getting this right started with three questions: Where are my agents? What can they access? Who owns them?
Most orgs can’t answer any of the three.