Every fortified town had a moment that revealed its real constitution: the approach of a lean season, when someone decided who stayed inside the walls and who was left outside them. The AI transition is producing that moment for tech employment, and the early evidence is specific about who stands where when the drawbridge rises.
The most careful data so far comes from Stanford's Digital Economy Lab, using large-scale ADP payroll records. Brynjolfsson, Chandar and Chen (2025) found that workers aged 22 to 25 in the occupations most exposed to AI experienced a relative employment decline of roughly 16 percent, while experienced workers in the same occupations, and workers in less exposed fields, held steady or grew. Two further findings sharpen the picture. The adjustment is happening through hiring and headcount, not through wages. And the losses concentrate where AI automates the work rather than where it augments it; the automation-versus-augmentation line, not the technology itself, predicts who gets hurt.
This is the drawbridge pattern. The people dropped first are those with the weakest claims on the inside: the newest, the youngest, the not-yet-hired. Incumbents are protected not because they are irreplaceable but because they are already in. An entire cohort's on-ramp narrows while the townspeople's daily life continues largely unchanged, which is why the harm is easy to understate; it arrives as an absence of offers, not a wave of firings.
Stated intentions match the data
Employer surveys point the same direction. In the World Economic Forum's Future of Jobs survey of over 1,000 companies across 55 economies, 41 percent of employers said they plan to reduce headcount as AI automates tasks, even as 77 percent plan to upskill existing staff (WEF, 2025). Read those two numbers together: the plan, in aggregate, is to invest in the people already inside and shrink the gate for those outside. Workers sense it. Gallup found 18 percent of U.S. employees consider it likely their job will be eliminated within five years, rising to 23 percent where AI has already been implemented (Gallup, 2025).
None of this settles the long-run question; the same WEF survey projects more jobs created than displaced by 2030 across the whole economy. Aggregates, though, are cold comfort in a lean year, because aggregates do not decide who eats. Institutions do.
Scarcity is a governance test
Here is the part the feudal frame clarifies. A layoff is usually discussed as an economic event, but its distribution is a governance event. When the harvest is thin, someone chooses: cut the newest, cut the cheapest, cut the farthest from headquarters, cut the department the CEO understands least. Those choices are rarely put to the people affected, and rarely even visible to them. The drawbridge goes up at night.
As Ricardo Semler has told it, Semco faced its own lean season in the Brazilian crisis of 1990 and made the opposite choice. Instead of a layoff designed in a conference room, the numbers went up on the wall and the workers decided: a salary cut for everyone, in exchange for control over spending and a larger share of profit in the recovery. The recovery came. His summary was about information, not kindness: families do not open the books to their members; companies that respect you do.
The point of the story is not that layoffs are always avoidable. It is that who decides and who knows are design variables, even in a crisis, and that a workforce given real numbers and real choices can act on them like adults. The modern norm inverts every element: decisions concentrated, information sealed until the morning the badge stops working, and the cut framed afterward as a force of nature rather than a choice among alternatives.
Reading a company before the lean year
For anyone inside the walls now, the practical question is not whether AI will thin some harvest somewhere; the payroll data says it already is, for the youngest first. The question is what your particular town does with scarcity, and that can be read in advance. Does the company publish enough real information that employees could evaluate a downturn themselves? Has it ever presented a trade-off (pay, hours, spending, headcount) as a genuine choice? Does its history show cuts as a last resort explained in numbers, or a first resort explained in euphemism?
A company's layoff style is not improvised in the bad quarter. It is rehearsed every day in how it shares information and distributes decisions when things are fine. The drawbridge only reveals the constitution. It was written long before.
Sources
- Stanford Digital Economy Lab — Brynjolfsson, Chandar & Chen (2025). Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of AI. Workers 22-25 in the most AI-exposed occupations saw ~16% relative employment decline; adjustment runs through headcount, not wages, and concentrates where AI automates rather than augments. https://digitaleconomy.stanford.edu/publication/canaries-in-the-coal-mine-six-facts-about-the-recent-employment-effects-of-artificial-intelligence/
- World Economic Forum (2025). Future of Jobs Report 2025. 41% of employers plan headcount reductions as AI automates tasks; 77% plan upskilling; net job projections to 2030 remain positive in aggregate. https://www.weforum.org/publications/the-future-of-jobs-report-2025/
- Gallup (2025). State of the Global Workplace. 18% of U.S. employees consider job elimination within 5 years likely, 23% where AI is already implemented. https://www.gallup.com/workplace/659610/state-global-workplace.aspx