Three years of surveys, stable headcount across the economy, and companies still making decisions based on a forecast that hasn't materialized. Here's what's actually happening.
The headcount data is the real tell, but it undersells how this feels from the hiring side. I'm on the buying end of engineering talent in LATAM, and the inbound now describes a role that AI already compressed, the CV still maps to the title, not to what the work became. The gap surfaces on the first depth question: the "senior" who owns a system on paper freezes the moment you ask what breaks under load. Stable headcount isn't stability. It's role-definition debt, and most orgs are paying interest on it without having named it.
We needed some good news about not being fully replaced by AI, so thank you for that! There is a great example articulated by Chamath Palihapitiya in radiology. The gist of it is that radiology doctors were supposed to be completely replaced by AI. Contrary to that belief, while AI analyzes images much better than doctors, there is now increased demand for that profession because we need those brilliant humans to interpret what AI has discovered from the imaging and provide the diagnosis.
Just like you, I would like to stay optimistic and see it as a net positive for humanity, job creation, and our overall happiness in the future!
The 22-25 age cohort chart buried near the bottom is the one that deserves top billing. Total headcount holds stable while entry-level hiring collapses. That's not stable employment. That's a drawbridge being pulled up. The experienced workers keep their seats because they already own the judgment layer. The entry-level pipeline that was supposed to train the next generation of judgment owners gets cut because AI handles the tasks those junior roles were built around. Five years from now the org chart looks the same but the bench is empty.
The Jevons framing is right at the economy level and wrong at the individual level. More work gets created in aggregate. The person whose specific skill got automated doesn't get the new work. Someone else does, often someone cheaper, often in a different country, often a person who never held the old job title to begin with. The lump-of-labor fallacy is a fallacy across the economy. It's a lived reality inside any single career.
The headcount data is the real tell, but it undersells how this feels from the hiring side. I'm on the buying end of engineering talent in LATAM, and the inbound now describes a role that AI already compressed, the CV still maps to the title, not to what the work became. The gap surfaces on the first depth question: the "senior" who owns a system on paper freezes the moment you ask what breaks under load. Stable headcount isn't stability. It's role-definition debt, and most orgs are paying interest on it without having named it.
This is well put. Good operators turn judgment into routines so the business does not depend on one person remembering everything.
We needed some good news about not being fully replaced by AI, so thank you for that! There is a great example articulated by Chamath Palihapitiya in radiology. The gist of it is that radiology doctors were supposed to be completely replaced by AI. Contrary to that belief, while AI analyzes images much better than doctors, there is now increased demand for that profession because we need those brilliant humans to interpret what AI has discovered from the imaging and provide the diagnosis.
Just like you, I would like to stay optimistic and see it as a net positive for humanity, job creation, and our overall happiness in the future!
The 22-25 age cohort chart buried near the bottom is the one that deserves top billing. Total headcount holds stable while entry-level hiring collapses. That's not stable employment. That's a drawbridge being pulled up. The experienced workers keep their seats because they already own the judgment layer. The entry-level pipeline that was supposed to train the next generation of judgment owners gets cut because AI handles the tasks those junior roles were built around. Five years from now the org chart looks the same but the bench is empty.
The Jevons framing is right at the economy level and wrong at the individual level. More work gets created in aggregate. The person whose specific skill got automated doesn't get the new work. Someone else does, often someone cheaper, often in a different country, often a person who never held the old job title to begin with. The lump-of-labor fallacy is a fallacy across the economy. It's a lived reality inside any single career.
nice work.