AI agents spark an identity cleanup

AI agents are turning identity and access management back into a boardroom problem. As companies move agents out of pilots and into production, the old sprawl of stitched-together tools suddenly looks less like harmless complexity and more like a risk they can’t keep ignoring.

That was Jon Addison’s read at Okta’s Oktane event, where he said customers now see agent security as close to the identity work they’ve already been doing for years. The pressure is straightforward: business teams want speed, while security teams want visibility and control over what these agents can touch.

Addison said many enterprises built their identity stacks piece by piece and still pay heavily to keep point products connected. Agents are raising the cost of that approach. Before they scale up, some organizations are choosing reference architectures. Others already have shadow agents running and need help putting access controls around them.

The result is a cleanup phase, not a brand-new category. Addison framed it as housekeeping: getting the fundamentals right, cutting back on point solutions that create vulnerabilities, and consolidating where they can. And underneath that sits a bigger shift — today’s use cases are about efficiency and augmentation for people, but autonomous agents will force companies to think about access at a much larger scale.