Introduction: The Last Human Pass
At 4:42 on a Thursday afternoon, you have a decision to make. A client is waiting for a summary. Your manager wants it before the end of the day. You have six pages of notes, two spreadsheets, a transcript, and just enough mental energy left to realize that doing the work properly will take longer than you have.
So you give it to AI. Within seconds, you have a polished draft. The headings are clean. The language sounds professional. The recommendations are arranged into neat categories. You skim it, fix one awkward sentence, and send it. For you, the task is finished. For the person who receives it, the work may just be starting.
You saved forty minutes. They lost ninety. That leads to a more uncomfortable question than “How much time can AI save us?”: Whose time did it save—and who paid for the shortcut?
The book’s central standard is simple: AI can help search, draft, compare, translate, brainstorm, calculate, question and revise. But AI doing part of the work is not the same as you being finished with the work. If your name is attached to the result, the judgment, context, decision and responsibility still belong to you.
The Last Human Pass
The last human pass is the moment after the machine has helped and before the work leaves your hands. It is where you ask: Is this actually useful? Is it true? Is it complete enough for the decision being made? Did I understand what I am sending? Did I protect information I was trusted with? Did I leave hidden work for the next person? Do I need to explain how AI contributed?
Chapter 1: You Saved Time. Who Paid for It?
AI changes the economics of workplace quality because low-quality work can now hide inside high-quality presentation. A polished document may still force the recipient to open the raw data, trace claims, ask follow-up questions and rebuild the thinking that should already have been done.
Time saved is not the same as work eliminated. Sometimes AI eliminates work. Sometimes it simply moves the work.
The Next-Person Test
Before sending AI-assisted work, ask: What does the recipient need to do after receiving this? What questions will they immediately have? What will they need to verify? What context have I assumed instead of explaining? What decision was I supposed to make that I have quietly left for them?
A better definition of “done” is this: The work is ready when the next person can use it without doing the thinking you were responsible for.