Okta Tackles AI Agent Chaos with Identity Controls

Companies are increasingly deploying autonomous software agents, but many lack a clear understanding of where these digital workers operate or what they are permitted to do. Okta’s President and COO, Eric Kelleher, argues that this uncertainty is the primary barrier to adoption. He describes the current market environment as one dominated by confusion, where organizations face a surge of vendor claims while trying to secure their new AI-driven workflows.

According to reporting by GN technics/ai (en-US), Okta is addressing this gap by extending its core identity security platform to cover these non-human actors. Rather than building a separate system for AI agents, the company is integrating them into the same infrastructure used to manage employee and customer accounts. This approach allows businesses to apply familiar controls, such as access permissions and audit trails, to their software agents.

Extending Human Identity Rules to Machines

Okta’s strategy relies on the premise that an AI agent, whether temporary or long-term, requires a trackable identity to be secure. The company has updated its Universal Directory to include these agentic identities, allowing them to be governed through standard provisioning and deprovisioning processes. This means that when an agent is no longer needed, its access rights can be revoked just as easily as an employee’s login credentials would be.

The practical implication for IT teams is a reduction in the complexity of managing a hybrid workforce. By treating agents as distinct entities within the existing directory, companies can enforce specific rules about which systems an agent can connect to and what actions it is authorized to perform. This centralized oversight helps prevent unauthorized access and provides a clear audit history for every digital action taken.

Rapid Growth in Agent Deployments

Demand for these controls appears to be growing quickly. Okta’s AI-agent product became generally available on April 30, and the company has already completed dozens of transactions. One notable example involves a Fortune 50 healthcare company that initially reported having only 50 deployed agents. When Okta’s team returned for a follow-up discussion just weeks later, that number had surged to 1,500.

This rapid expansion highlights the challenge of keeping pace with autonomous software that can be spun up and scaled down with little oversight. Kelleher noted that forward-looking security leaders are driving this adoption, but many organizations are discovering unexpected growth in agent activity only after it has already occurred. The ability to quickly identify and govern these new identities is becoming a critical operational need.

Pricing Strategy Reduces Friction

To encourage adoption, Okta is initially pricing its agentic identity services as a per-user uplift rather than charging separately for each agent. This pricing model is designed to lower the barrier to entry for companies that are still figuring out how many agents they need or how their usage might fluctuate. Most agreements are currently one-year terms, reflecting the experimental nature of the technology.

The trade-off for this flexible approach is that long-term costs may rise as agent counts stabilize and grow. However, by avoiding per-agent billing, Okta aims to remove a significant source of hesitation for customers. This strategy positions the company as a partner in solving the immediate problem of security confusion, rather than just a vendor selling additional software licenses.