One line every time you override an AI agent. In three weeks, it shows you which catches only you can make, and which ones you’ve stopped making.
You are never going to check every output. That ship sailed the moment the work started arriving faster than you could read it.
So stop trying to verify the work. Start recording your judgment.
When you fix, rewrite or reject something an agent produced, you write one line: what it got wrong, and how you knew. Ten seconds.
Three weeks of those lines sorts your corrections into three piles: the failures only you can catch, the ones you catch out of habit (write the preference down once and stop re-imposing it), and the ones you’ve quietly stopped catching. The first pile is your job description now. The third is the one that ends up in front of a client.
Where it comes from: McKinsey’s August 2026 paper on agentic adoption recommends that leaders note when human judgment overrode an agent, and say why. The log is my extension of that, for your own work. Credit and source are inside.
Two pages: the log itself, and what three weeks of it will show you.
Get the free two-page log
It opens on the next page. You’ll also get five short lessons by email (they start in November), then the occasional note from Bob. Unsubscribe anytime.
Dr. Bob Nolley is the author of Leadership for Entrepreneurs: Rise of the New Leaderati, Part 1, out now on Kindle.