The Thought Experiment

Citrini Research, a New York–based investment research firm founded in 2023 by James van Geelen and known for its “guerrilla” thematic and macro research work, took a stab at answering this question last February. The result was a thought experiment, “The 2028 Global Intelligence Crisis,” written as a fictional postmortem from June 2028. It details a systemic economic collapse triggered by the sudden unwinding of the scarcity of human intelligence. What this means is AI eviscerates service industries, causing mass white-collar job losses and crushing consumer demand.

Innumerable economists leaped to AI’s defense, all more or less saying the same thing — even if jobs were destroyed, even in currently high-paying roles, capitalism would create others just as it always had. Citrini’s provocation, while scary, was unlikely to come to pass.  

I don’t want to debate the finer points of the thought experiment. What I do want to do is set out how Citrini opens up three specific critiques of any AI “success,” and how any such win for AI capitalists would be a loss for capitalism, ultimately further eroding the foundations of the Global North’s economies.

Stagnation and the Role of Frontier Industries

To understand how everything going right would ultimately mean everything going wrong, it’s important to see AI as an answer to an economic question: how to solve the problem of secular stagnation.

Secular stagnation is a concept that describes the persistently low rates of productivity and demand growth throughout the Global North. There are a range of both orthodox and heterodox theories that account for it, but as the economic historian Aaron Benanav contends, it has become more or less the consensus view across the political spectrum.

Within this context, AI represents hope: it is a frontier industry promising the revival of economic growth.

Frontier industries are industries that are not yet “mature,” meaning there are both economic and technological gains to be made, promising high returns to business as well as innovative advances from intellectual property to new monopolies to productivity gains to high share prices. Frontier industries include all of the green industries and those in the so-called fourth industrial revolution — AI, biotech, automation, as well as other cutting-edge fields.

The bet on frontier technologies is they will enable new growth — new markets, more productive labor, and new sites of investment.

Citrini’s provocation is that AI will in fact worsen the problem of stagnation even if it delivers on productivity gains and investment returns (initially).

While Citrini stays relatively close to mainstream economics, we can read across its fictional postmortem. In doing so, we find three primary drivers of AI’s destructive future history that map onto specific theories of secular stagnation and economic decline that deserve further scrutiny: the impact of the shift to service-dominated economies on productivity; the rise of services overcapacity; and the impacts on rentierism and intermediation (generating income from the ownership and control of assets and the business of mediating economic activities, such as accounting or digital platforms, respectively) within the neoliberal economy.

AI and My Boy Baumol

Much of the orthodox response to Citrini focused on the question of job destruction — that AI would not augment jobs but replace them. But in that debate the nuance of what was being posed was lost.

Citrini claims AI will enable a rapid expansion of digital Taylorism into service work. Services have been historically hard to industrialize, as they tend to be limited not only by the speed at which people can work but by being more variable and “social.” But already with chatbots and AI agents we are seeing an erosion of the “humanness” of services. This may lead to two things — job destruction and a surge in productivity.

This won’t be an even process. What will likely occur is a bifurcation of services into high-productivity, highly automated service sectors and low-productivity sectors, with the workforce similarly splitting between a small, high-waged workforce and mass of low-waged service workers.

This is a version of what the economist William Baumol called the “cost disease.” Baumol and economist William G. Bowen developed the thesis when commissioned to study the economic performance of the performing arts. They found that the labor output of performing arts is generally fixed — it takes the same amount of time to perform a Shakespearean play today as it did hundreds of years ago. Conversely, workers in industry had increased their productivity many times over. While a factory worker could produce ten times more car parts because of the introduction of machines, a violinist could not “speed up” their performance without ruining the product. The thesis has since been applied to the divide between capital-intensive and labor-intensive sectors — manufacturing and services, broadly speaking.

The divide between labor-intensive services and increasingly capital-intensive manufacture creates a specific economic problem — the costs of services increase relatively while dragging down the growth rate of a broader economy. This happens as in the manufacturing sector, technological innovation drives high productivity, allowing wages to rise while the relative cost of goods falls. Conversely, in labor-dependent services, wages rise despite flat productivity, causing the relative cost of these services to climb. The cost climbs as wages in labor-intensive sectors like health care and education rise to keep pace with the rest of the economy, despite these sectors lacking the productivity gains seen in manufacturing.

The impact of this is that while goods get cheaper thanks to technological innovation, services get more expensive. Additionally, manufacturing sheds jobs as productivity increases, shifting employment to the lower-productivity services sector, exacerbating the problem.

Baumol’s cost disease leads to a low-growth economy where essential services such as health care become unaffordable while TVs get cheaper every year.

Citrini’s argument is that AI enables the automation of some (eventually most) services, recreating Baumol’s cost disease within the service sector. Services that can be broken down into discreet tasks (“Taylorized”) and can make use of an increasingly data rich environment, such as call center work, basic accounting, legal discovery, graphic design, much sales work, or routine diagnostics and coding, will be automated, reducing the total labor force employed and increasing productivity. At the same time, there will remain a labor-intensive service subsector with low productivity growth. This labor-intensive subsector will itself be under immense pressure as AI and robotics advance.

This bifurcation recreates Baumol’s cost disease within the service sector, destroying many of the well-paid positions that have retained some degree of workplace autonomy in the process. The result of this transformation in work would be the emergence of an economy shaped by very few highly paid service workers, and an army of low-skill, low-paid workers. All of this would take place against the backdrop of a collapse in the total mass of service employment due to productivity gains.

The lesson from technological innovation in the manufacturing sector is that increased productivity means that firms require fewer workers. While new markets may develop along with new services, these new services will not escape the division between a small number of well-paid workers and a dwindling mass of their low-paid peers. The worst-case scenario would be one in which even this low-waged work disappears thanks to service automation.

Finally, the remaining low-productivity services like education would also face pressure because of their rising cost. The effect of this downward pressure would be felt more severely by public services than private companies, as an economy increasingly dominated by secular stagnation will impose ever stricter budgetary constraints on governments.

Brenner and Overcapacity

Part of the dynamic Citrini describes involves AI leading to a vast excess of service capacity as competition leads to companies locking in rather than exiting the market.

This is the dynamic that the economist Robert Brenner argues was the structural cause of the global economic crisis of the late 1960s and early 1970s — a global overcapacity in manufacturing. The global post–World War II build-out of manufacturing capacity squeezed profit margins for all manufacturing businesses. Growing global competition in turn drove down margins, and, in response, industry looked to raise productivity to increase revenues rather than exiting the sectors in which they had already made investments, ultimately worsening the profit crisis.

Classical economics would suggest that, in this situation, what would occur is a “clearing out” of the lower performing businesses. Investment would move into other sectors where there is growth to be found, while underperforming companies would close down or sell up to competitors.

In contrast to a “healthy” capitalist dynamic, where poorly performing companies give way to high-performing ones, what Brenner tracks is how, when challenged by more productive rivals, companies refused to give way and abandon fixed assets. Instead, they doubled down on chasing market share, creating a persistent tendency toward excess manufacturing capacity, reducing overall profit rates and capacity utilization.

What Brenner doesn’t consider is the role of nation-states in maintaining manufacturing overcapacity — something already underway within AI. Specific industries have long enjoyed political support, either for military purposes or for far more explicitly political ends, be it to ensure voter support or just as an aspect of the everyday corruption of political elites.

The Citrini narrative suggests both aspects of overcapacity will come into play. Rather than compelling firms to pack up and move into some other sector of the economy, Citrini suggests that AI will engender a similar escalatory dynamic, where competition drives adoption while at the same time pushing companies to “stay and fight” for market share.

As competition intensifies, the drive to industrialize service work and to adopt labor-replacing AI will further reduce workforces while making services paradoxically less attractive as investments (due to falling margins and lower growth prospects). At the same time, governments, caught in a vision of international relations preoccupied with great power competition, will be unwilling to cede AI dominance to their rivals and will instead shore up national AI companies and infrastructure, worsening global AI and service-sector overcapacity.

Ultimately this will lead to a persistent tendency toward overcapacity in services, mirroring the tendency within manufacturing, eroding profit margins and tempering the appetite for investment in additional businesses or even entire market sectors.

The Final Euthanasia of the Rentier?

Rentierism is not an aberration, but a central aspect of the economies of the Global North. The most complete accounting of this aspect of contemporary capitalism has been undertaken by the economist Brett Christophers. Christophers brings two accounts of rent together in his work. The first is income due to the ownership and control of scarce resources, while the second is due to monopoly or oligopoly power. In both, rentierism constitutes the ability to generate revenue above “average or expected” normal returns through the ability to limit or prevent economic competition.

Much of what constitutes the service economy could be described as rentierism, including most digital services and platform businesses that generate revenue from their occupation of critical nodes mediating economic exchange.

Citrini describes this intermediary work as “friction” — it adds to the costs business customers and consumers pay for a service. It also adds to the internal costs of business operations insofar as some specific operations, such as legal compliance or accounting, rely on hiring either certified staff or consultants. Much of the white-collar work threatened by AI is precisely this kind of intermediary work. As AI automates it, it puts not only specific roles at risk but huge swaths of the service economy as well. And while rentierism may theoretically constitute a parasitic form of accumulation, one that adds “friction” to economic processes and higher costs to consumers, it is also a huge source of employment and site of investment.

The key vehicles for rentierism are investment funds such as Blackrock and Blackstone, making rentier capitalism a system run by and through asset managers. If we bring these institutional asset managers together with those businesses that are rentiers, such as Google and Microsoft, the vast bulk of the US stock market is owned by, and dependent on, rentierism as a foundation. And while we could clearly say around one-third of the US economy comprises rentierist businesses at a minimum, including those businesses and jobs that are fictive in Citrini’s reading would lead to a much higher percentage.

We can understand rents and the drive to rentierism as a response to secular stagnation — as a means of securing certain and well-defined future revenues and of escaping the destructive effects of market competition. To eradicate rents would be to destroy a primary site of capitalist investment, alongside whole subsectors of the economy and millions of jobs. It would also fatally undermine stock market investments, based as they are on perpetual rents.

While the talk of eradicating friction or even rents suggests a “freeing up” of capital for more productive investment, given services would follow manufacturing into a realm of hyperproductive overcapacity, there would seem to be no upside to the euthanasia of the rentier in this instance.

Rather than “free up” business, this development would destroy it. Capital may well be a parasite, but in the absence of revolutionary pressure it is still work-producing. Our jobs might be bullsh-t, but without them there is only unemployment and (even more) poverty.

Whoever Wins, We Lose

Not all frontiers lead to expansion or growth. Exhaustion is just as much a possibility.

There is much to doubt about the utility and sustainability (economic and environmental) of AI. We are also increasingly seeing labor and social conflict over the new technology, from the relentless build-out of water-hungry data centers to the labor process itself.

Yet while we should organize against the further industrialization of our labor and exploitation of our sociality and natural world, we should also be clear-eyed as to the possibility that AI capitalists will manage to push ahead with their agenda.

Should they do so, it may well be a moment of singularity, just not the one Sam Altman and company have in mind. As Citrini suggests, it could very well lead to a vast collapse of business and consumer demand, while making whole aspects of the contemporary economy unviable. A profound deepening of stagnation, not its overcoming, would result. The tepid plans for universal basic incomes pushed by Silicon Valley tech bros would be laughably inadequate when faced with such an event.

The three aspects outlined above do not even constitute the totality of the challenge AI could pose to economic growth. What made the Citrini think piece so provocative was not its AI doomerism, but its recognition of the threat posed by the technology’s success.