StewAI Blog · Economics of AI · 13 min read
The New Solow Paradox: AI Is Everywhere Except in the Productivity Statistics
Frontier models match human experts on nearly half of real professional deliverables, yet credible forecasts put AI's total productivity gain near half a percent per decade. Both numbers are right. The 25x difference lives in the software layer between the model and the work.
Frequently asked questions
What is the new Solow paradox in AI?
Frontier models match human experts on nearly half of real professional deliverables, yet credible forecasts put AI's total productivity gain near half a percent per decade. The gap lives in the software and workflow layer between the raw model and the actual work.
Why is AI not showing up in the productivity statistics yet?
For the same reason the original Solow paradox held in the 1980s: measured value arrives only after organizations reorganize their processes around the technology. Capability is not the bottleneck; the workflow layer that turns capability into reorganized work is.
How do you close the gap between AI capability and AI value?
By building the software and workflow layer on top of the model. Encoding work as structured, reusable processes is what converts a capable model into measurable output.
Read the full article on StewAI