Technically

AI, Tractors, and the Productivity Paradox

The article argues that AI’s lack of appearance in productivity statistics mirrors past “productivity paradox” episodes, where early technology adoption—exemplified by the kit era of steam engines, automobiles, and early computers—didn’t translate into measurable output until firms reorganized. It explains that today’s “kit stage” is the widespread, experimental use of large language models by individuals and small teams, generating real output but remaining invisible to macro data. While LLMs lower external transaction costs, they raise internal integration costs, creating a lag as firms build the coordination machinery needed to capture AI’s gains. Ultimately, a handful of AI‑integrated giants will grow, the mid‑market will contract, and productivity numbers will only catch up after this organizational shift.