Generated by Codex with GPT 5.6 Sol XHigh

This summary covers The Economist’s June 20th, 2026 Finance & economics article listed in the contents as AI-pilled economists and published under the headline The new-seekers.

Artificial intelligence may reshape work, productivity and economic growth, yet many academic economists have been slow to study it. The article argues that the centre of AI economics is consequently moving away from universities and toward government agencies, think-tanks and the AI labs themselves. That shift gives researchers better data and closer contact with the technology, but it also risks putting important public questions inside institutions with commercial interests and limited incentives to share what they learn.

Economics Lags The Technology

Economists have responded quickly to earlier shocks. After the financial crisis of 2008, research on bank runs and credit crunches surged. Within two months of the covid-19 outbreak, nearly a third of working papers published by America’s National Bureau of Economic Research addressed the pandemic. More than three years after ChatGPT’s launch, research on AI remains comparatively scarce, and even some prominent work rests on debatable assumptions.

One reason is that AI has not produced the kind of clean, immediate shock that economists can easily measure. Unemployment across rich countries is roughly where it was when ChatGPT appeared, while official GDP data do not identify AI investment or output with much precision. The technology is changing processes inside companies, often before its effects become visible in headline statistics.

A second reason is intellectual caution. Economic history shows that even powerful inventions can take decades to lift national growth because finance, institutions and culture slow adoption. Many economists therefore expect AI’s effects to be gradual. That scepticism may be prudent, but it can also discourage research into the possibility that AI will create entirely new products and activities rather than merely make existing work more efficient.

Research Moves Closer To The Models

AI-curious economists are finding two alternatives to academia. Statistical offices and central banks are building the basic evidence that future research will need. Surveys from American and Canadian agencies track business adoption, the Bank of England studies firms’ expectations, and Britain’s new AI Economics Institute examines the technology’s effects on productivity and labour markets.

The more consequential migration is into frontier AI companies. Anthropic, OpenAI and Google DeepMind have recruited economists to study how increasingly capable models may change the economy. These labs offer proprietary data, direct access to model developers, influence with policymakers and salaries that can exceed \$300,000 even for relatively junior roles. Researchers outside universities are also devising experimental measures such as “AI GDP” and real-time productivity trackers, helping fill gaps that conventional statistics leave open.

This work is improving, but its quality is uneven. Some corporate reports amount to little more than descriptions of chatbot usage dressed up as economic analysis. More fundamentally, industry researchers may be encouraged to study narrow commercial questions or publish findings that portray AI as useful and safe. Research that would challenge a lab’s strategy, expose risks or reduce demand may be harder to pursue openly.

Better Access, Less Independence

The trade-off is familiar from the earlier migration of computer scientists into technology firms. Industry can supply better tools and turn research into practical innovation, but it tends to produce more patents and fewer openly published papers. Knowledge that could inform society’s response to AI may become proprietary just when policymakers most need impartial evidence.

The article’s warning is not that economists should avoid AI companies. The labs possess information that outside scholars cannot reproduce, and collaboration with them can reveal how the technology actually works. The danger is allowing the field’s most important questions to be defined almost entirely by those companies. AI may transform the economy slowly or suddenly; either way, universities need independent researchers capable of testing corporate claims, identifying harms and pursuing questions whose answers may be inconvenient. Academic economics has ground to make up before the institutions building AI also become the main authorities on what it means.