Okta turns its own Dex AI agent into customer zero for identity controls

At Okta's Oktane event, Jenna Cline, senior vice president of technology, data and intelligence at Okta, described how the company uses its own Dex AI agent both to streamline employee support and as a proving ground for the identity controls it offers customers. She spoke with theCUBE Research's Krista Case and co-host Rebecca Knight.

Okta has raised its internal efficiency target sharply. After surpassing last year's goal of saving 250,000 employee hours, the company set this year's target at 1 million. Cline framed that as roughly a month of work for every employee, a mix of AI and the deterministic automation that remains in its systems.

Architecturally, Dex routes work to specialized sub-agents rather than trying to be one agent that handles everything. Cline said Okta wanted to avoid a swivel-chair problem as individual business teams brought in their own agents, so Dex provides a common interface instead.

In her account, Dex calls on the right sub-agent, and each sub-agent is scoped to only certain work, which keeps it secure and efficient because it does one thing well. The same pattern extends to the expense team and the IT help desk.

Identity is the other half of the story. Okta registers Dex in Universal Directory and applies capabilities from Okta for AI Agents, so the internal agent is governed like any other identity in the environment.

Using Dex as customer zero also shortens the feedback loop. Cline said the internal team worked with the product team, whose members took feedback from what Okta was trying to do in order to make the product stronger, giving the company an enterprise environment to evaluate features before they become generally available.

For enterprises preparing to put agents to work at scale, the sequencing matters: Okta runs its own agent under the same identity, permission and audit constraints it sells, then markets that capability to customers. An internally proven control model is itself part of the product's credibility.

One caveat: the discussion came from a paid media partnership interview produced by SiliconANGLE's theCUBE at Oktane, so the figures reflect Okta's own account. The next things to watch are when capabilities such as Okta for AI Agents move from internal trials to broad availability, and whether the 1 million-hour target is actually reached.