For two decades, the offshore engineering pitch was a spreadsheet: the same work, at a lower hourly rate. That model is quietly dying, and AI is what killed it.
When a capable model can generate a working implementation in minutes, the value of an engineer who simply types code faster and cheaper collapses. Keystrokes are now close to free. What did not get cheaper — what got more valuable — is judgment: knowing what to build, what not to ship, where the system will break under load, and which AI-generated change is quietly wrong.
From cost arbitrage to judgment arbitrage
The old model sold capacity. The new one sells decisions. A small, elite team that directs AI with discipline will out-build a large team that uses AI to type faster — and it will do it for less, because you are paying for judgment, not headcount.
This is the part most “AI-augmented” claims miss. Handing a mediocre team better tools mostly amplifies whatever was already there. AI is a multiplier, and a multiplier works in both directions.
What this means for how you staff
If judgment is the scarce input, then team design is the whole game: who you hire, how you grow decision-quality, and how you keep context from evaporating across people and timezones. That is a different discipline from “add seats.” It is the one we have spent three decades building — and the reason a team half the size can ship twice the product.