CONCEPT
The AI Measurement Gap
Diane Coyle’s diagnosis: not that the AI transition is failing to generate productivity gains, but that the statistical infrastructure designed to measure an industrial economy is constitutionally blind to the gains the AI transition produces and the costs it incurs.
The AI measurement gap is not the Solow
productivity paradox of the 1980s, when computer investments failed to appear in the productivity statistics because the organizational transformation that converts technology into output took decades to complete. In that case the gains were real but delayed; waiting was the correct response. The AI transition presents a different structure: the gains are appearing now, spectacularly, in the productivity statistics that can see them—and the most consequential effects of the transition are invisible to every metric the policy apparatus currently consults.
Diane Coyle’s argument, developed in her October 2025 essay “Measuring AI’s Economic Impact” and the Stanford Digital Economy Lab white paper on measuring AI, is that the measurement infrastructure was designed for an economy in which output was physical, labor was hourly, and quality was approximated by price; that AI’s primary impact operates through decision quality, organizational transformation, and
invisible surplus production outside markets; and