Transcript
But in enterprises, that's not always the reality. The innovation is moving faster than security. In our first two videos, we looked at why AI agents need an enterprise security model and how Okta for AI agents is designed to help deliver it. Now, in the final video in this series, we ask, what does it actually take to secure an agentic enterprise? We got answers from Accenture, a systems integrator deploying Okta solutions in their customers' environments. The biggest hurdle and really step one is discovery. Using a tool like Okta to discover the agents in your organization and the shadow AI that might exist. You can't govern what you can't see or what you don't know about. You can't authenticate and authorize things that you can't see. Shadow AI is already becoming a real enterprise problem, with studies showing that over 80% of workers use unapproved AI tools, often introducing unmanaged non-human identities, data exposure and compliance vulnerabilities. So how do you give employees the freedom to build with AI without losing visibility and control? What we're suggesting that organizations start with is a secure framework for how to build agents. Because if you have a secure framework for how your developers are going to build and create agents within your organization, then every agent starts its lifecycle already registered, already in your identity governance system. In other words, the goal isn't just to find rogue agents after the fact, it's to make secure deployment the default. That's the thinking behind what Okta calls a blueprint for the secure agentic enterprise. Know what agents exist, control what they can connect to, and govern what they're allowed to do. At a minimum, do you have a list of all the agents that are running around in your organization and can you control what those agents are connecting to? It sounds simple and it sounds like plumbing, but once you have that in place, everything builds on top of that. And for Accenture, that model only works if it fits the way enterprises already operate. We bring the accelerators and the playbooks that help our clients integrate tools like Okta for AI agents. It treats agents as first-class identities and keeps them in the same universal directory and not a separate governance silo. And that mirrors how we tell our clients to think about the problem. Don't build a parallel universe for agents. Use your existing infrastructure and your existing cybersecurity tools and extend your identity fabric to include agents. But if enterprises want this model to scale and to stay viable as AI agents evolve, it has to work seamlessly across environments, partner ecosystems, and more apps, systems, and vendors. That's why standards are quickly becoming part of the conversation. A few years ago, we built out a cross-app access capability with the goal of making it really simple to connect applications to applications. But now it's also applications to agents or agents to applications. Accenture is using Okta's cross-app access in our own platforms and tools that we're building to help our clients understand and use AI. So we're developing a set of frameworks and capabilities in our cyber.ai platform that will help clients piece together their existing infrastructure and cybersecurity tools in order to be ready for agents. And we see Okta's technology as critical in that enablement. For enterprise leaders, a takeaway is simple. AI adoption isn't slowing down, so security can stay a blocker. It has to become part of how these systems are built and scaled. You need to use AI to secure AI. You need to continue to move at the speed of AI. Because in the age of AI agents, speed alone won't be enough. The real advantage will come from building a model that can scale securely. Signing off for Okta's Newsroom, I'm Diana Blass.
