AI agents are writing more of your code every day, and securing that code now means putting AI to work on security too. This session explores how to think about security across the agentic SDLC, from governing the coding agents your developers already use to orchestrating security agents of your own. Real-world examples help you weigh what to build or buy and what makes sense for your software factory. We’ll explore why a triage agent might earn unattended autonomy long before a patch agent should, and how to keep token spend predictable as you scale from one repository to thousands. You’ll leave with a framework for understanding where your program sits today and deciding where to go next.