Four enterprise software vendors shipped standalone AI agent governance products between August 24 and early September 2026. The window between the first launch and the last was roughly two weeks. None of them waited for the agents to figure out their own rules first.
Okta went first on August 24, making Agent SSO generally available to the 20,000-plus customers already using its core single sign-on product. The pitch is straightforward: replace static API keys with short-lived, identity-governed tokens and treat every AI agent as a first-class identity alongside human employees. Ric Smith, Okta’s president of products and technology, framed it as infrastructure necessity: “AI agents are fast becoming a primary interface for how work gets done, but granting them access to enterprise systems shouldn’t require trading away security or visibility.”
Seven days later, on August 31, two more landed. IBM’s watsonx Orchestrate shipped its AgentOps Agent, which approaches governance from the orchestration layer. Its AI Gateway discovers agents built on Amazon Bedrock (with Azure AI Foundry and Google Vertex AI support coming by end of September) and imports them into a single control plane. A Trace Inspector lets teams follow an agent’s full execution path step by step, and a custom evaluation system lets companies define what “good” means on their own terms rather than relying on generic quality scores.
That same day, Broadcom unveiled AgentMinder at VMware Explore in Las Vegas, taking the security and runtime angle. AgentMinder acts as what Broadcom calls a “traffic controller” for autonomous agents—it verifies each agent’s identity and intended mission, then authorizes every action against that mission before the action reaches enterprise resources. Broadcom’s CIO Alan Davidson claimed the platform has delivered “massive global scale paired with zero downtime,” though that reflects internal metrics rather than independent verification.
Dataiku’s Agent Management rounds out the sequence—launching in September with general availability planned for October. It is a standalone product that does not require the Dataiku platform. Where Okta focuses on who an agent is and Broadcom on what it is allowed to do, Dataiku focuses on whether it is actually delivering results. The product connects to agents built on Microsoft Copilot Studio, Salesforce Agentforce, AWS Bedrock, Google Vertex, LangChain, ServiceNow, and a dozen other platforms, then measures them against business key performance indicators rather than uptime. “Technical monitoring tells you agents are running,” the company notes. “It does not tell you they are doing their jobs.”
The timing matters less than the pattern. These four products come from different infrastructure layers—identity, orchestration, security, and observability—but they share a structural move: selling governance as standalone infrastructure rather than a feature bolted onto an agent platform. Vendors are building a compliance and control layer that sits above the fragmented landscape of LLM frameworks and cloud providers, standardizing oversight from the outside in.
The urgency is real. Okta’s 2026 AI Agents at Work report found that only 34% of organizations apply the same security controls to AI agents as they do to human workers. Gartner’s June 2025 CIO survey put agent deployment at 17% with another 42% planning to deploy within a year, but the firm also predicted that 40% of enterprises will have autonomous AI efforts partially derailed by governance gaps discovered only after production incidents. That is the market these vendors are responding to—agents are proliferating faster than the controls to manage them.
For enterprise buyers, the practical question is whether to layer specialized tools—identity from one vendor, runtime controls from another, observability from a third—or wait for a more integrated approach. There is no settled answer yet. These products overlap in some capabilities but their primary governance approaches differ enough that they may function as complementary components rather than direct competitors. The risk is governance sprawl: managing the governance tools themselves becoming as complex as managing the agents they are meant to control.
What is clear is that the governance layer is forming before agents have standardized. That is either a sign of a maturing market getting ahead of its own complexity, or a sign that the complexity has already outpaced the ability to manage it. The answer probably depends on which side of a production incident you are sitting on.