The Work Reshapes

Every craft on a software team was organized around a “how.” The product manager’s how was translation: turning a fuzzy business need into a written requirement, a story, the meeting where you explained the requirement a fourth time until everyone built the same thing. The engineer’s how was producing correct, efficient code. The designer’s how was making the screens, the comps, the polished thing the user would touch. The team’s how was coordination, the sprint and the standup and the ritual that kept a dozen people building the same product in the same direction. Four crafts, four hows, and for thirty years the boundaries between them were drawn exactly along those hows. You were a different discipline because you owned a different how. There was a fifth role, the architect, who barely had a how at all, whose whole job was the judgment about whether the thing would hold together; hold that one aside for now, because she turns out to be the key to the whole pattern.

The agent ate the hows. Not all at once and not completely, but the direction is unmistakable and it is the same direction in every craft. The translation that was the PM’s daily labor is increasingly something a model does from a few sentences of intent. The code that was the engineer’s craft is generated, fast, and mostly plausible. The screens that were the designer’s output are produced from a prompt. The coordination that the sprint manufactured is being run by agents that draft the plan and write the standup. The specific skill that made each discipline a discipline, the thing it produced that the others could not, is the thing that got cheap.

What did not get cheap is the judgment. Someone still has to decide whether the requirement is the right requirement, whether this code fits a system it cannot see, whether the agent should have been allowed to take that action, whether the team is building the right thing at all. That work, deciding what should exist and whether it is safe to ship, was always the senior, scarce, slow part of every craft, the part that sat above the how. And it is now the only part left that is reliably human. So the crafts, which used to be separate because each owned a different how, are converging on the same scarce thing from different starting depths. The PM, the engineer, the designer, the team lead are all being asked to stop being the person who produces the artifact and start being the person who owns whether the artifact should exist and whether it can be trusted. That used to be several jobs. It is becoming one posture, held by people who still carry different expertise. And the architect, the fifth role, was already standing at that destination before any of the others started walking toward it, because the judgment about whether the whole thing holds was the only thing she ever did.

This is the section where that happens, watched across the crafts. Each chapter takes one and follows the same arc, because the arc is the same: the how gets commoditized, the production work collapses toward zero, and what remains is a supervisory job the discipline was never primarily trained for and now has to make its center.

There is a hard edge to this convergence, and it runs through all four chapters, so it is worth naming once here. The convergence is a gift to the people who already have deep judgment in a craft and a trap for the people who do not yet. A senior who spent fifteen years earning the judgment can now apply it at the speed of the agent and look like a force multiplier. A junior who has not earned it cannot tell good agent output from bad, and the routine production work that used to be how they earned it is exactly what the agent now does. The same shift that elevates the experienced may be quietly closing the door behind them, removing the apprenticeship that produced the judgment the whole system now depends on. Every chapter in this section hits this wall from a different side. None of them has a clean answer, because the field does not have one yet.

The convergence is hard to see from inside any one craft, because each discipline watches only its own road. It is visible from the seam, the position that sits across all the crafts at once, which is the product manager’s, and which is where the last chapter of this section stands to look back at the others. The crafts are not just each changing. They are becoming, at the level of the work that matters, the same craft. The rest of this section shows the roads in.

One thing to carry into these chapters and out the other side, because it is the hinge of the whole book. Everything in this part is the first of the two things the foundations distinguished: how AI changes the way a team builds, any team, any product, the feature and the dashboard and the agent alike. It is real and it is happening and it is most of what the industry means when it says AI is reshaping work. But it is only half the story, and it is the half that conventional products also get. The convergence makes the build leaner and the builder more senior; it says nothing about the second product, the supervisory layer, because a conventional product does not have one. So read this part as the story of the workshop, not the story of the thing on the workbench. When the thing on the workbench is itself an agent, that begins in the part after this one, where the work stops being about how the team builds and starts being about what it has to watch once the building is done.