I use AI considerably more than I did a year ago, yet I have become more deliberate about where and how I allow it to participate in my work. Staying in control now extends beyond protecting data or checking outputs. It requires deciding when AI should enter the work, understanding the access and authority it receives, ensuring human judgment participates early enough to influence direction, and preserving enough cognitive effort to retain genuine understanding. This post examines four areas where my thinking and AI practices have evolved.