The work rewrite has started, and the window is shorter than it looks

Last month, a CFO at a Brazilian industrial company called me for a conversation I had been expecting for a while. He had just signed off on the first structured round of layoffs the company had done in nearly ten years. Almost a hundred and twenty people. The rationale was explicit. Automatable functions. Layers of mid-tier operations that AI already handled better with less friction. He wanted to walk me through the decision and ask whether he was moving too early or too late.

The answer is not simple, and he knew that. That is why he had called. Sit with the numbers in front of him, look across at an operator who had already seen three similar cycles, and try to separate the part of the decision that was strategic from the part that was fashion. His question turned into the basis of this piece. Not the org chart question. The deeper one. Are jobs going to disappear?

The short answer is no. The long answer is what matters.

History has already told this chapter three times

Agricultural mechanization did not end human labor. It rewrote who worked at what. At the beginning of the last century, more than half of Brazil's population was on the land. Today it is a small percentage, and we eat much more and much better than back then. The people did not vanish. They moved into industry, into services, into professions that did not exist when the tractor first showed up in the field.

Industrial automation did the same thing in the second half of the twentieth century. Assembly line workers became maintenance technicians, CNC programmers, quality supervisors, logistics specialists. Halfway through the curve, the picture was ugly. Industrial cities in the United States and in Europe lived through a hard decade. Whole families got caught in the rearrangement. What happened, by the end of the arc, was rewriting, not extinction.

The internet added another layer. Typesetter, telephone operator, travel agent, photo film executive. Whole professions got compressed or rewritten over twenty years. In exchange, the digital economy created hundreds of thousands of roles no career analyst in nineteen ninety-five would have predicted.

The pattern repeats. Shock. Friction. Rewrite. Expansion.

Why it feels worse this time

AI is following exactly that script, with one important difference for whoever is inside it.

The speed.

Agricultural mechanization took decades. Industrial automation, roughly a generation. The internet, twenty years. AI is pushing the same transition into a compact window. The models did not stop improving. The systems that orchestrate models went from demo to production in less than two years. The technology curve is moving faster than any labor market can absorb with calm.

That mismatch is the source of the pain. The technology is arriving at a pace the organizational economy cannot yet digest, and that is calendar friction, not function friction. People, in the aggregate, will reposition. They always have. The time between shock and repositioning is shorter now, and that is why short-term suffering shows up more concentrated, more visible, more painful. Anyone who looks at this phase and sees the end of human usefulness is reading the symptom as if it were the phenomenon. Anyone who sees the phase as the predictable friction of a compressed transition is reading the phenomenon for what it actually is.

No one in nineteen fifty had a way to predict the role of product manager. Anyone making decisions back then based only on what was visible in nineteen fifty was deciding for an economy that never arrived. Anyone who read that rewriting was coming sat in a different position, even without knowing the names of the roles that would emerge.

The model taking shape

On the other side of the friction, a clear model is taking shape. You can already see its edges in the people furthest along.

The human does not disappear from the operation. Roles change. The human moves out of the role of task executor and into the role of orchestrator. Sets the intent, picks the agents, supervises execution, audits the result, steps in when the system slips. The work stops being doing, and turns into thinking about how to do, with the AI agent doing.

It is a shift in what counts as production inside the company. Before, what produced value was the professional doing the task. Now, what produces value is the professional commanding a set of agents while they do the task. Whoever is good at this multiplies their own capacity. Whoever does not learn falls behind inside their own function, even when the function still exists. That is why the company that comes out of this transition in better shape will produce much more with fewer people, and the people who stay will be doing denser work, more strategic work, and, in most cases, better paid work. It is the script from a hundred and fifty years ago, compressed into a few.

The short window between ready and adopted

The point I wanted to make with the industrial CFO is this. The deeper question is not whether jobs are going to disappear. They are not. The question is sharper. There is an interval, right now, where the technology is already ready to deliver much of what it promises, and the market, in the aggregate, has not yet digested how to organize work, teams, processes, and governance around that.

This interval is short. It will not last ten years. It will probably not last three. Sometime between now and the end of this decade, the market normalizes. Companies will have agent orchestration structures in place. Careers will have clear operator-orchestrator tracks. Boards will know which metrics to ask for. Compensation models will reflect the new shape of who produces value. The advantage of having arrived first will become table stakes.

In the middle of all this, during that short interval, sits the asset. A company already running work in orchestrated mode while its competitor is still arguing about whether to train people in AI is, in practice, compounding productivity while the competitor compounds lunch conversations. A professional already doing dense work alongside agents is accumulating experience that no later course can speed up. The compound advantage does not come from perfect timing. It comes from the fact that the friction of rewriting gets paid first by an internal learning curve, instead of being paid later by an abrupt market correction.

In a short transformation, prudence wins. In a long transformation, prudence loses. AI is a long transformation, with a short capture window. Those two facts together say more about your company's next quarter than any macro reading published this week.

The right question for your next quarter

Stop asking whether the jobs in your company are going to vanish. History already answered. They will be rewritten.

Start asking who in your company, today, is already doing their work with an agent at their side, and who is still doing the work exactly the way they did it in two thousand twenty-three. Ask what percentage of the operation has already been designed with humans orchestrating agents, and what percentage still depends on humans doing it alone. Ask whether the next ten hires will come in to fill existing boxes, or to lead squads that mix people and agents in a new design.

Those questions are worth more than any long-range macro projection. They measure where your company stands inside the window. And the window is the asset.

AI is rewriting work while the news cycle insists on saying it is ending it. Whoever rewrites first will write the rule the rest will inherit.