What if your coding agent didn't just fix the repository in front of it, but became an automated modernization worker operating across your entire engineering estate? In this hands-on session, we'll build an agentic modernization pipeline using GitHub Agentic Workflows and Microsoft Agent Package Manager, then connect the agent to deterministic code-transformation capabilities so the agent can discover work, reason about what needs to change, select the right tool, and produce reviewable changes at scale. We'll explore the critical boundary between probabilistic reasoning and deterministic execution: let the agent decide what should happen, but use specialized tooling to ensure how the code changes happen is repeatable, explainable, and safe. Along the way we'll cover repository discovery, context engineering, agent skills, MCP, guardrails, validation, pull-request generation, and fleet-wide rollout. The goal is to turn AI coding from a developer-side assistant into an operational system for continuously paying down technical debt across hundreds or thousands of repositories.