The most important economic question in the AI era is not how many jobs will be replaced. It is whether AI is expanding human capacity: what we can imagine, do, and want. Youngjin Yoo argues that the demand frontier of human capacity — the outer limit of what people want, attempt, and are willing to pay for — will expand.
Much of today’s AI debate assumes that productive capacity can expand indefinitely while demand stays largely static. If that assumption holds, even spectacular AI-led growth can lead to a broad-based economic crisis: output rises, but household income and broad-based consumption collapse.
My research on generative innovation over three decades points consistently to one finding: the most consequential economic effects of new technology come not from productivity gains within existing demand but from the discovery of new demand — as people use new tools to attempt problems they could not previously approach. Recent experiment evidence from my team shows this: AI produces its most creative outcomes when users push beyond what they originally set out to do. The question is not whether to use AI, but how. Generative AI is likely to amplify this dynamic at the population scale — putting that creative capacity within reach of anyone with a sufficiently urgent personal problem.
That is why the same AI capability can produce radically different futures.
On 22 February, the Wall Street research firm Citrini circulated a fictional memo dated June 2028. In that scenario, the S&P 500 had fallen 38 per cent and unemployment had reached 10.2 per cent, including white-collar professionals once seen as protected. As firms replaced knowledge workers with AI and reinvested savings into more AI, household purchasing power deteriorated. Citrini called this “Ghost GDP”: growth visible in national accounts but absent from the lived economy.
A few weeks later, Sydney data scientist Paul Conyngham made headlines for a very different reason. Trying to save his rescue dog, he used ChatGPT and AlphaFold to help design a custom mRNA vaccine and worked with a university lab to produce it. Within a month, the dog’s tumour had shrunk dramatically.
The technology is the same. The economic logic is not.
Over two centuries ago, Thomas Malthus argued that population grows geometrically while food production grows arithmetically. His logic was coherent, but it assumed a fixed production frontier. He could not foresee advances such as the Haber-Bosch process that transformed agricultural capacity.
Citrini makes the same mistake, inverted. It assumes that while AI expands productive capacity, the demand frontier cannot move.
History argues otherwise. The Model T did not satisfy a pre-existing mass demand for cars; it created one. The iPhone did not respond to a clearly articulated demand for smartphones; it revealed possibilities consumers had not yet imagined.
In the industrial era, this market-creating innovation spark primarily came from firms. With generative AI, it can come from anyone. Conyngham was not testing a market thesis. He was solving an urgent personal problem.
And he is not alone. A colleague of mine, an amateur stargazer in Germany, connected Claude to a smart telescope. The system began sequencing observations and uncovered a Bluetooth control path that the manufacturer’s own community had not discovered. Millions of non-programmers are now building useful tools through “vibe coding,” often to solve their own immediate needs.
These experiments may look like side projects or hobbies, as most of these are not “work” in the conventional sense. Economically, however, they are powerful alternative ways to discover new market opportunities. Intrinsic innovation begins with a lived personal need — specific, passionate, and authentic. When many people build from lived need, they reveal latent demand that formal organisations have not yet seen.
This is why the standard debate on AI and the future of work, asking whether machines will replace or augment workers in a fixed task set, is wrong-headed. That framing misses the larger shift. The most important source of new demand may come from outside traditional work, through intrinsic innovation that starts with personal necessity and scales through shared relevance.
The Citrini scenario is not automatically wrong. It has to be made wrong. Demand expands only if AI does more than automate existing labour; it must expand human imaginative capacity.
That raises the real question for investors, policymakers, and AI firms. Instead of asking how many jobs AI will eliminate, we should ask whether we are building systems that help people discover new goals, pursue meaningful problems, and create value others did not yet know they wanted. Because if we do, Ghost GDP remains a warning, not a destiny. If we do not, the demand frontier will stall because we chose efficiency over imagination.
- This blog post represents the views of its author(s), not the position of the London School of Economics and Political Science Department of Management.
- Photo by Caleb Jack on Unsplash.
- Explore our other latest blogs on AI.
The post The real risk of AI is not job loss — it is a failure of imagination first appeared on LSE Management.