Geyto Research · Mariano D'Angelo · Market note · · 2 min read
The productivity paradox. Why technology takes time to show up in results
From Solow to the J-curve: why technology investments take time to show results, and the role played by changes in work and management.
In brief: technology takes time to show up in results because its benefit depends on changes that do not come with it: new processes, new capabilities, new ways of managing. Economics has described this pattern for almost forty years. Today it is relevant again with artificial intelligence.
"Everywhere but in the productivity statistics"
In 1987, the economist Robert Solow wrote a line that became famous: you can see the computer age everywhere but in the productivity statistics. Companies were investing heavily in computing, and aggregate productivity did not reflect that investment.
The explanation that took shape over the following decades is not that the technology was useless. It is that its effect depended on something else.
The J-curve
Erik Brynjolfsson, Daniel Rock and Chad Syverson formalized that "something else" in 2021. General purpose technologies, such as computing or AI, require complementary investments in intangible assets: redesigning processes, developing capabilities, reorganizing work. Those investments are hard to measure as capital, but they consume resources.
The result is a J-shaped curve. At first, measured productivity can even fall, because resources are going into things that cannot be seen. Later, as those intangibles mature, the benefits appear.
Management makes the difference
Nicholas Bloom, Raffaella Sadun and John Van Reenen compared companies investing in information technology in Europe. They found that the subsidiaries of US multinationals got more productivity out of the same investment, and they linked that difference to their people management practices.
The reading is direct: the same technology yields different results depending on how the work around it is organized.
What it means for AI
Generative AI is at a similar stage. Organizations buy licenses, launch pilots and announce initiatives. The question left by the history of the paradox is whether they are also investing in the complements: redesigning the work, the judgement to use the tool, the handoffs between people and systems.
Without those complements, AI can end up installed like so many other technologies: available, half-used and with no visible change in results.
What to do with this
- Budget for the complements, not only the technology: redesign time, learning on the job, support.
- Expect a curve, not a jump. Measure the intermediate changes in the work as well.
- Review management and the organization of work as part of the initiative.
More on AI as one more technological change in our approach.
Geyto Research publishes essays, evidence reviews and market notes on what happens after go-live.
Frequently asked questions
What is the productivity paradox?
It is the observation that large investments in technology do not show up right away in productivity statistics. Robert Solow summed it up in 1987: you can see the computer age everywhere but in the productivity statistics.
What is the productivity J-curve?
It is the pattern described by Brynjolfsson, Rock and Syverson (2021): general purpose technologies require intangible investments, such as new processes and capabilities, that are not measured at first and depress apparent productivity before the benefits appear.
Why does it matter for AI?
Because AI also requires changes in work, management and capabilities. Without those changes, the investment can stay installed without turning into results.
Sources
- Solow, R. M. (1987, July 12). We'd better watch out. New York Times Book Review, p. 36.
- Brynjolfsson, E., Rock, D. and Syverson, C. (2021). The productivity J-curve: How intangibles complement general purpose technologies. American Economic Journal: Macroeconomics, 13(1), 333–372.
- Bloom, N., Sadun, R. and Van Reenen, J. (2012). Americans do IT better: US multinationals and the productivity miracle. American Economic Review, 102(1), 167–201.
Share