Tag: Wright’s Law

OM in the News: AI and the Learning Curve

Scale drives efficiency—for almost a century, industrial planners have relied on this simple principle. In 1936 aeronautical engineer Theodore Wright discovered that costs fell in a predictable way every time production doubled. The more you produce, the cheaper things become, in part because the learning cost per unit declines. This is the topic of Module E in your text.

Artificial intelligence has accelerated this principle, writes The Wall Street Journal (Oct. 23, 2025). It is rewriting Wright’s Law, which assumes that experience follows production: You make mistakes, learn from them and improve. AI makes it possible for experience to come before production. Simulation can happen millions of times before a single box is shipped. Experience scales almost instantly at no real cost. The learning curve doesn’t only steepen. It collapses.

That means knowledge that once took decades of human trial and error can emerge in weeks, days, even hours. In a supply chain, this is a profound shift. Decisions about capacity, warehouse space, routing, technology adoption and risk management can be modeled, tested and optimized in advance. The costs of imprecise planning shrink dramatically.

AI is breaking Wright’s Law because the learning cycle is no longer physical but computational. Models can test, fail and improve millions of times faster than any team of human engineers. Experience can be generated in advance, and at negligible cost.

The implications for logistics are extraordinary. AI agents will negotiate, reroute and optimize flows of goods in real time. Traditional ownership models, fleets, warehouses and even labor could be replaced by dynamic orchestration of perfectly used assets.

This new golden age of logistics will unveil solutions to problems we may not even know exist. Wright’s Law still matters, but perhaps AI has broken it.  The challenge will be not building the tools but surviving the pace of their consequences.

Classroom discussion questions:

  1. Why can AI have this impact on learning curves?
  2. Besides logistics, which is mentioned in this article, can AI impact operations?

OM in the News: Learning Curves and Solar Power

Recent history shows that climate policies such as taxes, subsidies and mandates matter most by catalyzing a virtuous cycle of higher demand that leads to more innovation, learning-by-doing and economies of scale that lower costs and further boost demand. In solar power, the results have been spectacular. Between 1980 and 2012, the cost of a photovoltaic module made from crystalline silicon fell 96%. Roughly 30 percentage points of this is attributed to public and private research and development, which among other things, led to more efficient modules and larger, thinner silicon wafers. Another 60 points came from “learning-by-doing”—improvements to the manufacturing process, such as less waste, that came with experience—and economies of scale, reports The Wall Street Journal (Aug. 25, 2022). The average plant capacity grew roughly 200-fold.

Similar, though less dramatic, dynamics have been at work in wind power and battery storage. They all hewed to “Wright’s Law,” named for the 1930s aeronautical engineer Theodore Wright. In Module E, Learning Curves, we explain that each doubling of production is accompanied by a roughly constant percentage decline in cost, known as the learning rate. “Over the long term these learning rates appear to be the best way to predict the future cost of technology that we know of,” said one industry expert.

One implication is that as a technology matures, production takes longer to double and so costs fall more slowly. In solar power, for instance, PV module factories are now so large and the manufacturing process so efficient that incremental improvements are much harder to come by. Sure enough, the cost of solar-generated power has fallen an average of 6% annually from 2018 through 2021, compared with 21% in the previous nine years.

Greater potential lies in replicating the experience of solar in other technologies that are currently too costly for widespread adoption. Emission reductions in the coming decade are the low hanging fruit, achievable with technologies that are already competitive or nearly so with fossil fuels such as wind, solar and batteries. Getting the rest of the way to net zero depends on hard-to-decarbonize sectors such as aviation, industrial processes and agriculture for which commercially viable technology to eliminate emissions doesn’t yet exist.

Classroom discussion questions:

  1. Where were learning curves first applied? (See Table E.1 in your Heizer/Render/Munson text)
  2. Why do costs fall more slowly as a product matures?