DeepMind economists say AI teamwork, not skill, drives change
A Google DeepMind essay argues that AI agents' ability to coordinate, more than raw competence, decides when AI reshapes the economy.
Researchers at Google DeepMind argue that the big economic shift from AI will depend less on how skilled individual AI agents become and more on how well they work together. Their essay, The Economics of Intelligence, was published on October 29, 2026 by Alex Imas, DeepMind’s director of AGI economics, research scientist Andrew Koh, and co-founder and chief AGI scientist Shane Legg. It draws on a longer paper, Organizing Intelligence, which sets out the formal model.
The core argument
The authors split AI capability into two parts. Competence is skill at individual tasks such as maths, writing or coding. Coordination is the ability of AI agents to work together inside a firm. In their model, gains in competence shift wages, prices and the mix of people and AI agents in a company gradually, while leaving the structure of industries largely intact. Gains in coordination change how organisations scale, and once coordination passes a critical threshold, the essay says, it can cause large, abrupt jumps in market concentration, employment and wages.
Four regimes
Mapping AI along both dimensions, the essay names four regimes: minimal AGI, supercompetent AI, supercooperative AI, and artificial superintelligence. The authors argue that the most visible transformation may come from better coordination, a capability they say most current benchmarks do not measure.
Why it matters (analysis)
The essay offers one explanation for a common puzzle: AI tools keep getting more capable, yet the wider economy looks much the same. It is a theoretical model, and its scenarios are the outcomes of the model’s assumptions rather than forecasts. The essay appears on the DeepMind Institute, a site for essays by DeepMind researchers, and should be read as the authors’ work rather than as a Google product plan. Its practical suggestion is that anyone tracking AI’s economic impact should watch how well agents cooperate, not only how they score on individual tests. More on industry and business.
This report is not legal, financial, or health advice. See the disclaimer.