# Okta Wants Identity to Be the Control Plane for AI Agents > Source: > Published: 2026-09-22 09:22:27+00:00 As builders and infrastructure operators gather at [Oktane 2026](https://www.okta.com/newsroom/press-releases/showcase-2026/) in Las Vegas this week, the conversation has shifted from the novelty of AI agents to the mechanics of their governance. With the event underway at Caesars Forum, the industry is confronting a reality where agent deployment has outpaced the infrastructure designed to secure it. According to the [Okta AI Agents at Work 2026 report](https://www.okta.com/newsroom/articles/ai-agents-at-work-2026-agentic-enterprise-security/), only 34% of organizations apply the same security controls to AI agents as they do to human workers. This gap is the primary challenge for those building the next generation of enterprise workflows. ## Identity as the Primary Control Plane To scale agentic workflows, identity must move beyond simple authentication. Instead, it is becoming the primary control plane for the agent economy. This shift requires a structural change in how we manage agent interactions. The [Agent Gateway](https://www.okta.com/newsroom/articles/okta-july-2026-product-innovations/), released in July 2026, exemplifies this by functioning as a Policy Enforcement Point (PEP), while the identity platform acts as the Policy Decision Point (PDP). By intercepting requests and validating them against real-time identity data, this architecture ensures that an agent’s scope is continuously verified rather than relying on static, one-time configurations. “Enterprises are deploying AI agents in higher-value, but more complex workflows that require deeper access to sensitive resources and collaboration with other agents,” said Ely Kahn, Chief Product Officer at Okta. “With these new innovations within Okta for AI Agents, every agent connection can be evaluated and authorized in real time, then continuously validated as projects evolve or an agent’s scope changes.” For developers, this means integrating agents as first-class identities. A critical component is [Agent SSO](https://www.okta.com/newsroom/press-releases/okta-brings-first-class-identity-to-ai-agents-with-agent-sso/), which reached general availability on August 24, 2026. By bundling this capability into core Okta SSO at no additional cost for over 20,000 customers, the barrier to entry for securing agent access has been significantly lowered. “Okta is an undisputed leader in SSO, and now we’re bringing SSO for your AI agents,” said Ric Smith, President of Products and Technology at Okta. ## Maintaining Lineage in Multi-Hop Workflows As agents pass tasks to one another, maintaining security context becomes increasingly complex. The Identity Assertion Authorization Grant (ID-JAG) addresses this by maintaining identity lineage across multi-hop workflows. By embedding both the human user (the `sub` claim) and the acting agent (the `act` claim) within the token, systems can track the chain of command. This ensures that every action taken by an agent remains tethered to its original authorization — the accountable human travels with the request. “Our AI Fabric is built on the principle that every AI agent should have a trusted identity, just like every employee,” said Shawn Fogarty, Director of IT at LogicMonitor. “Identity is the foundation for secure AI at enterprise scale.” ## Standardization at the Protocol Layer The infrastructure for AI agents is increasingly defined by protocol-level decisions. A significant development is the incorporation of Cross App Access (XAA) — an open protocol Okta initially led — as the official Enterprise-Managed Authorization extension for the Model Context Protocol (MCP). By shaping the protocol layer, the industry is moving toward a standard that allows for consistent governance across diverse agent frameworks. This matters because 76% of AI applications are purchased rather than built internally, according to Menlo Ventures, requiring a unified way to manage access across disparate systems like Claude Code, GitHub Copilot, and Salesforce Agentforce. ## The Gating Function for Adoption The deeper question for builders is whether identity will become the primary gating function for enterprise agent adoption. Currently, 84% of organizations doubt they could pass a compliance audit focused on agent behavior or access controls, according to the Cloud Security Alliance. This uncertainty acts as a friction point for scaling agentic workflows. If identity platforms can provide the visibility and control required for these audits — through mechanisms like automated resource access certifications — they may transition from background utility to the central infrastructure that enables enterprise-scale AI. The path forward involves moving away from treating agents as external scripts and toward integrating them as first-class identities. This requires a shift in mindset: from managing static credentials to managing the full lifecycle of an agent — from discovery and onboarding to continuous authorization and eventual deactivation. As the ecosystem matures, the ability to enforce granular, identity-based policies at the edge will likely define the boundary between secure, scalable agent deployments and those that remain trapped in the experimental phase.