Okta Targets AI Agent Confusion With Identity Security Push

Okta is positioning identity security as a solution to AI-agent uncertainty, focusing on where agents operate, which systems they can access, and what actions they are authorized to perform.

The company says agent identities can be governed through existing identity tools, including provisioning, deprovisioning, auditing and access controls.

Okta’s AI-agent offering has completed dozens of transactions, including a multimillion-dollar deal with a Fortune 50 healthcare company whose agent count reportedly grew from 50 to 1,500 in weeks.

The company is initially pricing the service as a per-user uplift rather than charging by agent, with mostly one-year agreements.

Beyond AI agents, Okta reported strength across its workforce and customer identity businesses, with new products contributing 30% of quarterly bookings.

Its strategy emphasizes platform independence, broad integrations, ecosystem partnerships and continued product expansion following the Permiso acquisition.

Okta NASDAQ: OKTA President and COO Eric Kelleher said enterprise customers are increasingly seeking guidance on securing AI agents, with uncertainty around the emerging technology representing a central challenge for the identity-security market. Speaking at the Goldman Sachs Communacopia + Technology Conference, Kelleher said Okta is positioning its AI-agent offerings around three core questions: where agents are deployed, what systems they can connect to, and what actions they are authorized to perform.

“The biggest competitor we really have right now is just confusion,” Kelleher said. He described a security environment in which enterprises face a growing number of vendor claims while attempting to understand the risks posed by autonomous software.

Kelleher said Okta’s Blueprint for the Secure Agentic Enterprise is intended to simplify the process for customers exploring agent deployments. The company’s Okta for AI Agents offering maps to the blueprint’s focus on discovery, connectivity and authorization, he said.

He pushed back on the view that managing agent identities requires an architecture fundamentally separate from human identity systems. While some agents may be temporary, Kelleher said not all agents are ephemeral and each still requires an identity that an enterprise can track and govern.

Okta’s Universal Directory, which has historically supported workforce, customer and service-account identities, is expanding to include agentic identities, according to Kelleher. He said customers can use the company’s recently launched Agent SSO, now generally available on its workforce platform, to add agent identities to their directories at no additional charge.

The same governance capabilities used for human identities can be applied to agents, including just-in-time provisioning and deprovisioning, auditability, and controls over access permissions, he said.

Okta’s AI-agent product became generally available April 30, and Kelleher said the company has completed dozens of transactions so far. He characterized the sample size as small but cited several drivers behind early customer adoption, including forward-looking chief information officers and chief information security officers, as well as organizations that have identified unexpected growth in agent activity. He pointed to a multimillion-dollar transaction with a Fortune 50 healthcare company. At an initial discovery call, the customer repor...