How the Job Actually Changed: A Field Snapshot
Before the argument, the week. The same PM, the same five days, three years apart, and you can see in the calendar what changed.
The preface made a claim. This chapter shows it, because the claim is the kind that sounds like consultant noise until you watch it land on an actual calendar. So here is a week, drawn from a composite of PMs I have worked with and been, close enough to real that you will recognize your own in it.
The week that used to be
Three years ago the PM’s week had a shape, and most of the shape was production. Monday went to the status deck for the steering review and the half-day of pulling numbers to fill it. Tuesday was requirements: writing the PRD for the next feature, then the meetings to walk three teams through it because the document never quite carried on its own. Wednesday was the backlog, grooming tickets, writing acceptance criteria, answering the same clarifying questions in five Slack threads. Thursday was the customer calls and the competitive summary nobody else had time to write. Friday was catch-up, which meant the spec that did not get finished Tuesday and the deck that needed reworking before Monday.
Inside that week, the actual product judgment, what should we build, for whom, why this and not that, happened in the gaps. An hour here, a hung-over-from-a-good-conversation insight there. The judgment was the part everyone agreed mattered most, and it was the part that got whatever was left after the production work was done. Which was usually not much.
The week now
The production work did not disappear, but it stopped owning the week. The status deck is a first draft in twenty minutes from the data the PM points an agent at; the hour that remains is spent deciding what the deck should argue, not assembling it. The PRD drafts itself from a conversation and a few documents; the work is now reading it like an adversary, finding the three places it is confidently wrong about a clinical workflow. The backlog grooming and the clarifying answers and the competitive summary are mostly handled, mostly fine, needing a check rather than a creation.
What fills the space is not free time. It is the work that used to live in the gaps, now with room to breathe, plus a category of work that did not exist before. The PM spends Tuesday designing how an agent should behave when it is uncertain, which is a harder question than any PRD ever posed. Wednesday goes to a problem nobody had three years ago: the senior engineer whose code review the agent now does, and what that has done to how the team makes decisions. The customer calls are still there, because no agent has them for you, and they matter more, because they are now the main place raw reality enters the building.
What moved, what vanished, and what got harder
The three are different.
What vanished is the least valuable layer: the assembling, the formatting, the re-explaining, the first draft of everything. Little of it is mourned, and none of it is missed for long.
What moved is the judgment work, from the gaps to the center. It did not get easier. It got more exposed, because it is now what the week is actually for, and there is nowhere to hide a weak product decision behind a busy calendar.
What got harder is the new work, the part with no playbook: specifying behavior under uncertainty, supervising a system that acts faster than you can watch, carrying accountability for a decision no human in your building actually made. This is most of the rest of the book.
And here is the number the preface promised not to inflate. For most PMs this shift is real but partial. The measured reallocation for a heavy AI user today is a few hours a week, two to four for the typical case, with the very top tier of adopters reclaiming something closer to a full day. Routine has not collapsed to a tenth of the week for the median PM; it has fallen from perhaps half the week to perhaps a third, and a good deal of what was freed gets quietly eaten by the new work of checking what the agent produced. The direction is unmistakable and the destination is visible. We are not there yet. The book is written for the journey, not the arrival.
If you sketched your own week the way I have sketched this one, two columns, three years ago and now, the interesting line would not be how much an agent now does. It would be where the freed hours actually went, and how much of them disappeared into checking the agent rather than into the judgment work you keep meaning to get to. The gap between where the hours went and where you wish they went is the real subject of this book.