Imagine falling asleep for three decades and waking up to today’s economics seminar anywhere in Europe. You would probably not recognise much. Globalisation has changed economic research profoundly: it has made the field more comparable, more competitive, and thoroughly quantitative. Research is now conducted almost entirely in English, by Asian and European scholars as much as American ones, all competing to publish in the same small set of top journals. Promotion runs on those publications, and the community that judges them crosses borders as easily as the papers do.

Before this, the European landscape was fragmented, largely along language lines. Theory existed everywhere, but in many fields, method sat closer to history and law than to statistics. German Ordnungspolitik is the clearest case: a tradition that argued from first principles, grounded in the constitutional order, and reasoned in absolutes rather than trade-offs. It was convinced, and convincing, in a way that modern economics, with its endless “it depends”, rarely manages. The shift towards marginal thinking and cost-benefit reasoning was progress away from that certainty, but it also cost economics some of its ability to speak plainly.

AI may unsettle this new order as thoroughly as globalisation unsettled the old one. Two mechanisms are worth separating. The first is technical: AI is already good at economic theory and at executing empirical method, and it drafts and edits well, which means the supply of competent papers is rising fast while gatekeeping capacity – three to four hundred ‘top five’ slots a year for a global pool of between 8,000 and 11,000 professors at PhD-granting universities – stays fixed. Competition that started as a healthy discipline against parochialism risks curdling into something closer to what the Chinese call ‘involution’, effort consumed by the contest itself rather than converted into knowledge. Something has to give: more journals, or journals mattering less for promotion, closer to the norm in the natural sciences or medicine.

The second mechanism is geopolitical, not technological: fragmentation along geoeconomic lines, with economists outside the US gradually becoming less US-oriented than they have been for the last three decades. Anecdotal evidence from European admissions committees points to fewer of their strongest doctoral candidates applying to US programmes than in the past. The result need not be a step backwards for scientific progress, any more than the old fragmentation by language was purely a loss. But it means new, possibly regional, publications and quality standards have to emerge to sustain competition. A loss of the global community, though, would be a true loss. Much of economics is inherently cross-border – think of international trade or exchange rates.

A third, more radical possibility is that AI becomes good enough at modelling, empirical method, and self-improvement that it substitutes for a meaningful share of current researchers rather than merely assisting them. Currently, the profession does not seem to anticipate this. A recent CEPR survey of its own Fellows on the likely effects of AI on the profession returned results that were, on balance, surprisingly benign: almost 90% of respondents thought it likely that not much would change in the research and publication model.  However, most of CEPR’s Fellowship is tenured, and a tenured economist has limited reason to expect the coming disruption will touch them personally. Benign expectations may say less about the technology than about who was asked.

What is already becoming visible, independent of anyone’s forecasts, is that AI is making researchers more productive, and that the signal contained in a paper’s apparent quality is getting noisier, harder to distinguish from a well-drafted, machine-assisted product. As the public signal – the paper itself – becomes harder to read, the private signal – who the author is and who vouches for them – becomes more valuable: knowing your gatekeeper, or ensuring your gatekeeper knows your co-author, matters more, not less. That may reinforce clubby tendencies, and since most of the global gatekeeping still sits in the US, it may be a development that works against researchers in the rest of the world.

Economists as policy advisors

In an earlier Europe, many professors saw their job as consultant and advisor to their own government, local or national, and had an advantage now easy to under-rate: they spoke, literally, the same language as the policymaker. They also had the advantage of strong convictions. An Ordnungspolitiker did not hedge.

The modern economist looks less like that and more like a statue of Vishnu, many hands, each holding a different piece of the argument: this dataset, that caveat, this alternative specification. The advantage is that the advice is evidence-based. The disadvantage is that economists increasingly speak in tongues to the people who need to understand them most, and are not infrequently declared, by the same policymakers, to be useless.

How will AI change this? One outcome is that AI becomes the preferred advisor outright. It has a real advantage here: prompted appropriately, it can be made one-handed, decisive, uncaveated – everything the modern economist has trained itself out of being. A minister who wants an answer rather than a distribution of possible answers may simply prefer to ask the machine.

The more likely outcome runs the other way. As AI increasingly generates the technical analysis, competent trade-off calculation becomes cheap and common, and what stays scarce is the thing AI cannot fabricate: a human relationship built on trust, accountability, and the willingness to be blamed when advice is wrong. Human experience and interaction, rather than becoming obsolete, become more valuable precisely because everything else has become easier to automate. 

One prediction can be made with high confidence: a Sleeping Beauty who wakes up thirty years from now will recognise economics as little as someone who fell asleep thirty years ago and just woke up today.