Not sped up. Killed. Off your plate permanently. Running while you sleep.
I ask that in every workshop I run in San Francisco, and I watch the hands go down. For almost everyone, the honest answer is zero.
You open a chat window. You ask it something. You get a decent answer, copy it out, move on. And it feels like progress. But the inbox is still yours. The weekly report you rebuild by hand is still yours. The outreach, the research, the follow-ups — still yours.
Here’s the part nobody tells you: that’s not because you picked the wrong tool. It’s because you never picked the right task.
Stanford ran the study — 1,500 workers, 844 tasks. Most of the money going into AI agents right now is aimed at work people didn’t want automated in the first place, while the work they were desperate to hand off sat there untouched.
Those aren’t technical failures. They’re selection failures. Wrong task, not wrong model. Which means the first skill isn’t prompting — it’s knowing which hour of your week to point this at. That is session one: the HAS Audit, an hour of arithmetic on your own calendar that sorts every task into Green Light, R&D Opportunity, or Red Light — and shows you your Human Moat, the part of your week that stays yours.