AI in AEC is moving fast, but most teams are still stuck at the chat window. This session maps the three stages of AI maturity in BIM workflows: conversational chat, saved commands, and agents. Using real deployments inside Autodesk Revit, we'll show how firms turn one-off AI prompts into reusable commands their whole team can run (think add-ins created in minutes, not months), and how agent-based systems go further: planning multi-step work, using tools, and even creating new tools when a workflow demands it. Drawing on lessons from enterprise teams automating modelling, documentation, and QA/QC, we'll cover what worked, what failed, and how to judge where your organization sits on the chat-to-agent curve. You'll leave with a practical framework for scaling AI beyond individual experiments to team-wide automation and how agents can provide adoptive solutions to workflows.