Measuring AI Productivity Wrong: Why Intermediate Inputs Miss the Real Gains

Описание: We sometimes measure “AI productivity” by looking in the wrong place: the intermediate inputs. Faster code, cheaper compute, more electricity, better internet, higher programming speed — these are all upstream. But the real welfare and growth effects show up downstream, in the final goods and services that consumers actually experience: better products, faster delivery, higher quality, and entirely new offerings.

In this video, I argue that Schumpeterian theory gives us a better lens. AI should be treated as an intermediate good in the production process, just like earlier general-purpose technologies. Its true impact shows up when it spills over downstream—changing the speed, cost, variety, and quality of final output. Focusing only on intermediate metrics (like “developer speed”) risks underestimating both the gains and the disruptions.

To make this concrete, I’ll discuss Zack Shapiro’s article “A Claude-Native Law Firm,” which is not theory or hypothesis but a real-world case: a two-lawyer firm using general Claude to achieve productivity levels that normally require hundreds of lawyers. This is exactly the kind of downstream, firm-level transformation that simple task-timing studies miss.

If you’re interested in going beyond the hype and looking at AI through a serious growth and innovation lens, this video is for you.

#AIProductivity #Schumpeter #AIEconomics #IntermediateGoods #Automation #AIInnovation #FutureOfWork #ClaudeAI #LegalTech #Productivity #GeneralPurposeTechnology