When an AI can draft an answer in seconds, it is tempting to ask whether we will need to think at all. A more useful question may be what people can do with easier access to cognitive tools—and what happens to those whose work changes along the way.
A historical analogy, not a forecast
In “Cognitive Mercantilism”, published on 25 October 2026, writer and strategist Venkatesh Rao compares some AI anxieties with a historical tendency to treat wealth as a scarce stock to be controlled. Mercantilism was a family of economic ideas associated with state control of trade and the pursuit of wealth and power. Rao’s analogy asks whether intelligence, too, is being treated as something valuable only when held by a select few.
His alternative is that cheaper access to cognitive assistance could generate demand for education, research and projects that people previously could not pursue. He points to the long shift away from agricultural labour as an analogy: rising productivity helped make different kinds of work possible. The analogy suggests a possibility, not proof that AI will produce the same outcome. Rao also speculates about AI systems generating demand of their own; that is a further theoretical extension, not an established finding.
Original essay: https://contraptions.venkateshrao.com/p/cognitive-mercantilism
What has actually been measured?
One study of customer-support workers found that access to an AI assistant increased issues resolved per hour by roughly 14% on average. The gain was around 34% for less experienced and lower-skilled workers and much smaller for the most experienced. This is evidence about one tool and workplace, not a prediction for every occupation or a guarantee of new jobs.
Research paper: https://www.nber.org/papers/w31161
A separate 2025 survey of 319 knowledge workers by Microsoft and Carnegie Mellon researchers found an association between greater confidence in AI and less self-reported critical-thinking effort. Respondents also described a shift toward checking outputs and managing the task. The study relies on self-report and does not establish that AI causes long-term cognitive decline.
The missing questions: access and transition
Imagine a small business owner who can use a tool to explore several ways to explain a service. Producing options becomes easier; deciding which claims are accurate and useful still requires judgment. Price, training, available time and control over the tool will shape who benefits.
Our assessment: Rao broadens the debate beyond the tasks that disappear to the ambitions that might become practical. But an expanding set of possibilities does not automatically replace lost income or give everyone equal access. Education, transition support and the distribution of gains deserve as much attention as the technology’s potential.
