The Next Great Transformation Needs a Great Transition

From Polanyi to AGI and the Fight to Re-embed Intelligence in Society

The last Great Transformation taught us what happens when society reorganizes itself around a fiction.

For Karl Polanyi, that fiction was the self-regulating market. Markets had always existed, but market society was something different: a civilization in which the market ceased to be embedded within social, moral, political, and communal life, and instead became the organizing principle for society itself. The economy was no longer a set of institutions serving human purposes. Human purposes were increasingly reorganized to serve the economy.

To make that possible, society had to accept a dangerous abstraction. It had to treat labor, land, and money as commodities.

But labor is not a commodity. It is human activity, inseparable from human life. Land is not a commodity. It is nature, place, ecology, and the material ground of existence. Money is not a commodity in the ordinary sense. It is a social and political technology for coordinating value, credit, obligation, and time. None of these things was produced for sale in the way shoes, grain, or furniture are produced for sale. They were, in Polanyi’s famous phrase, “fictitious commodities.”

That fiction made industrial capitalism possible. It also made industrial capitalism socially explosive.

The nineteenth-century market order generated extraordinary productive energy. It helped unleash industrial growth, global trade, urbanization, and modern prosperity. But it also produced dislocation, insecurity, environmental damage, inequality, imperial extraction, and recurring crises of legitimacy. Eventually society responded. Labor law, trade unions, central banks, public education, social insurance, democratic regulation, welfare states, and public infrastructure were not incidental add-ons to capitalism. They were the countermovement that made market society survivable.

The transformation happened first. The transition came later.

But we do not have to return to the nineteenth century to see the consequences of an under-managed transformation. We have been living through a smaller version of it.

The age of globalization and outsourcing was not the Great Transformation in Polanyi’s sense. It did not reorganize the full foundations of society around a new economic fiction. But it did reorganize production, trade, labor markets, supply chains, communities, and politics on a vast scale. It created real aggregate gains. It lowered prices, expanded markets, lifted many people out of poverty globally, and made supply chains more efficient. But the transition was partial, technocratic, and politically under-imagined.

The consequences have been enormous.

In many advanced economies, the costs of globalization were concentrated while the benefits were diffuse. Some places gained. Others lost factories, bargaining power, tax bases, civic institutions, dignity. The damage was not only economic. It was felt in towns and families - as factory closures, community drift, humiliation, and the knowledge that decisions affecting their lives had been made somewhere else by people they would never meet.

The policy response was to let the pain do the work. Workers needed to retrain. Communities needed to adjust. Economists pointed to aggregate gains and politicians promised that everyone would benefit from the ensuing growth. But compensation was partial, bureaucratic, slow, and detached from the social importance of what had been lost. A person does not lose only income when a factory closes. A town does not lose only jobs when its economic base disappears. It loses a story about itself.

The political consequences are still ongoing and now impossible to ignore. Research on the “China shock,” including work by David Autor, David Dorn, Gordon Hanson, and Kaveh Majlesi on the electoral consequences of import competition, has shown that trade exposure contributed to political polarization. Dani Rodrik has argued that globalization shocks, working through both economics and culture, helped fuel populism. J. Lawrence Broz, Jeffry Frieden, and Stephen Weymouth have described the geographic foundations of the globalization backlash in declining regions of wealthy countries. Joseph Stiglitz’s later work makes the same point in broader form: globalization was not merely implemented; it was mismanaged, and the backlash reflected a crisis of legitimacy as much as a dispute about efficiency.

This very recent history matters for what happens next in AI. A transformation can be positive in the aggregate and still socially destabilizing if its costs are concentrated, its benefits are abstract, and its legitimacy is assumed rather than earned.

If a smaller transformation produced consequences this large, we should expect a larger transformation to produce consequences larger still.

That is why the next Great Transformation needs a Great Transition.

We are now entering a transformation whose central force is not only the market, but machine intelligence. And the question is whether we can build a Great Transition before the transformation overwhelms institutions designed for the last age.

Joseph Stiglitz saw this possibility early. In “The Coming Great Transformation,” he argued that advances in artificial intelligence and robotization could produce a transformation comparable to earlier industrial upheavals, but also insisted that societies are not helpless before it. The future is not determined by technology alone. It depends on policy, institutions, distribution, and democratic choice. That is the essential starting point: the AI transformation may be coming, but the transition remains open.

Anton Korinek has pushed the problem further toward the frontier. Where Stiglitz identifies the coming transformation, Korinek asks what happens if advanced AI changes the basic role of labor itself. In work on scenarios for the transition to artificial general intelligence, Korinek and Donghyun Suh model a stark uncertainty. If human tasks have an effectively infinite frontier of complexity, then there may always be new work for people to do. But if the range of economically valuable human tasks is bounded, and AI eventually performs all of them, wages may collapse. Even before full automation, wages may fall if the pace of automation outruns capital accumulation and makes human labor too abundant.

The question is not simply whether AI will create more jobs than it destroys. The deeper question is whether the economy will continue to need human labor as a scarce productive input. If labor scarcity weakens, then the institutions built around labor scarcity — wages, careers, education, taxation, bargaining power, status, and the moral centrality of work — all come under pressure.

That does not mean full labor substitution is inevitable or immediate. Human judgment, presence, trust, taste, responsibility, care, and social meaning may remain valuable in ways that are difficult to reduce to narrow productivity comparisons. Complementarity between humans and machines may persist for longer than some forecasts assume. But the possibility that labor scarcity weakens is serious enough that we should prepare for it before the shock arrives.

This is why the AI transition cannot be understood only as a technology story. It is a political-economic story and a story about how societies distribute security, agency, power, meaning, and voice when intelligence becomes increasingly detachable from human labor.

Recent Polanyian-inspired writing on AI has sharpened this point. Jeremy Shapiro has argued that the real AI race is not only technological or geopolitical, but social: which societies can embed AI in institutions that preserve legitimacy, status, income, and meaning? That is right. But we should go further. The challenge is not merely to absorb AI after disruption occurs. The challenge is to design a Great Transition before machine intelligence becomes another disembedded system to which society must adapt after the fact.

Artificial intelligence is usually discussed as a technology, a productivity tool, a labor-market shock, or a geopolitical race. It is all of those things. But if we take Polanyi seriously, and if we take the backlash to globalization seriously, we should also understand AI as a political-economic transformation: a change in the relationship between markets, institutions, knowledge, labor, power, and society.

AI will surely automate many jobs and even more tasks. It will also, we hope, significantly increase productivity. But what happens when more forms of cognition become detachable from human labor?

For most of economic history, human beings were the primary general-purpose intelligences in the system. We observed, interpreted, judged, coordinated, persuaded, cared, designed, managed, taught, repaired, imagined, and governed. Institutions were built around the scarcity and development of human cognition. Schools cultivated it. Professions certified it. Firms rented it. Bureaucracies organized it. Wages compensated it. Careers narrated it. Status rewarded it.

The transition from today’s AI to more general-purpose machine intelligence changes the institutional status of cognition.

Intelligence becomes something that can be generated, copied, rented, embedded. It becomes infrastructure. It becomes capital. If markets were the operating system of the first Great Transformation, machine intelligence may become the operating system of the next.

And this is where Polanyi’s analysis becomes newly useful.

The first Great Transformation depended on the fictitious commodification of labor, land, and money. The next Great Transformation risks the fictitious commodification of cognition, agency, trust, and social reality. As Bob Jessop argued in his work on knowledge as a fictitious commodity, capitalism repeatedly attempts to commodify capacities that are not naturally produced for exchange. AI extends that logic into domains once considered inseparable from human judgment and social life.

To call cognition, agency, trust, and social reality “fictitious commodities” is not to say they are identical to labor, land, and money. Cognition is cultivated through human development, education, language, culture, and experience. Agency is the basis of moral and political responsibility. Trust is a social relation. Social reality is the shared symbolic environment in which democratic life occurs. When these are rendered as services, signals, scores, outputs, agents, or optimizable assets, the market does not merely exchange them. It changes the conditions under which they can be reproduced.

Cognition is commodified when expertise, judgment, analysis, creativity, emotional interpretation, and persuasion can be generated on demand. Agency becomes a commodity when autonomous systems act, decide, negotiate, buy, sell, hire, fire, recommend, allocate, and govern. Trust becomes a commodity when credibility, intimacy, authority, and reputation can be synthesized, optimized, and scaled. Social reality becomes a commodity when the shared environment of public meaning can be generated, personalized, targeted, and manipulated by machines.

The danger is not simply that AI will replace workers, but that society reorganizes itself around machine intelligence before it has built the institutions needed to make that intelligence legitimate, accountable, and broadly accessible.

Globalization gave us a warning. People need agency, recognition, status and an understanding of how transformation serves them as well as compensation. They deserve institutions that arrive before the damage, not after.

Globalization was a warning shot. AI will be the main event.

What remains scarce?

If machine intelligence becomes abundant, what remains scarce?

That question runs through the emerging economics of AGI. It was also central to a recent discussion between Alex Imas, Phil Trammell, and Dwarkesh Patel about post-AGI economics. For centuries, labor’s share of income has remained surprisingly durable because capital and labor have been complementary. Machines still needed people somewhere in the production chain. But advanced AI raises the possibility that some supply chains could become fully automated, with ownership, capital, compute, energy, land, and other bottlenecks absorbing far more of the gains.

This does not mean all scarcity disappears. Scarcity shifts.

Some scarce factors may be physical: energy, chips, land, minerals, data centers, and access to supply chains. Some may be institutional: legitimacy, trust, safety, regulatory capacity. Some may be human: care, moral judgment, responsibility, meaning. Some may be positional: status, attention, authenticity, and the desire for experiences that are valuable precisely because they are human.

This has two implications for a Great Transition.

First, the transition cannot assume that productivity growth alone will solve distribution. If labor becomes less scarce while ownership of scarce complements remains concentrated, the gains from AI may flow toward capital, compute, energy, platforms, land, and institutions that control access. Shared prosperity will require deliberate mechanisms of participation, not passive faith in growth.

Second, the transition cannot reduce human value to whatever tasks machines cannot yet perform. That would leave human dignity hostage to the frontier of automation. The point is not to find the last defensible human job. The point is to build a society in which human beings remain secure, recognized, and free even as the economic demand for labor changes.

This is why post-labor thinking matters. David Shapiro’s Labor Zero and Post-Labor Economics work provides a useful vocabulary. Its most important distinction is between work and compulsion. A post-labor future need not mean a post-effort, post-purpose, or post-contribution future. It means a society in which survival is no longer conditional on selling labor into a market.

This is not a prediction that paid work simply vanishes. It is a way of asking what institutions should exist if the coercive force of labor-market dependence weakens, changes, or becomes less morally defensible in a world of abundant machine capability.

People may still build, learn, teach, repair, create, govern, explore, and more. But work becomes chosen and, in that sense, the transition is not from work to idleness, but from labor as economic necessity to contribution as human possibility.

The Great Transition

A Great Transition would begin by resisting three inadequate responses.

The first is technological fatalism: the belief that whatever can be automated will be automated, whatever can be optimized should be optimized, and whatever the market rewards must be accepted as progress. This is the old dream of the self-regulating market in computational form. It treats capability as destiny and political choice as friction.

The second is nostalgia: the hope that we can preserve the existing labor-market order by slowing or containing the technology until our familiar institutions feel safe again. Much work is meaningful, dignifying, and socially necessary. Much work is also coercive, degrading, badly paid, alienating, or organized around domination. The aim cannot be to preserve dependence on wage labor as such. The aim must be to preserve and expand human agency, security, dignity, belonging, and freedom as the economic basis of work changes.

The third is globalization déjà vu: the belief that society can let the transformation proceed, measure the aggregate gains, promise retraining and compensation, and trust that people will eventually adjust. That would be a profound mistake. The globalization era showed that people do not experience transformation only as workers or consumers. They experience it as members of places, families, communities, occupations, and nations. They ask not only whether prices are lower or GDP is higher, but whether their lives remain legible and respected inside the new order. AI will intensify this question because it reaches not only industrial labor, but cognitive work, creative work, institutional authority, public trust, and the meaning of contribution itself.

Polanyi teaches us that markets can be socially destructive when they are disembedded from social purpose. Globalization teaches us that efficiency without legitimacy produces backlash. We don’t have to learn the AI lesson the hard way.

The task is not to stop AI. The task is to re-embed it.

But re-embedding should not mean forcing machine intelligence into old institutional forms and calling the job done. This is where AI may be not only the object of governance but part of the means by which governance improves. If machine intelligence lowers the cost of analysis, coordination, translation, auditing, representation, bargaining, and participation, then a Great Transition should use those capabilities to create forms of democratic agency that were previously too expensive, too slow, or too difficult to sustain.

The goal is not simply to constrain AI from above. We can use AI to make bottom-up coordination, public deliberation, collective bargaining, institutional auditing, and democratic authorship more feasible.

That re-embedding has five dimensions: security, share, sovereignty, reality, and resilience and meaning. Four of them are moves a society can make, but reality is a condition all the others require: a shared world in which a society can know what is happening to it and decide together what to do.

The Great Transition

A Great Transition would begin by resisting three inadequate responses.

The first is technological fatalism: the belief that whatever can be automated will be automated, whatever can be optimized should be optimized, and whatever the market rewards must be accepted as progress. This is the old dream of the self-regulating market in computational form. It treats capability as destiny and political choice as friction.

The second is nostalgia: the hope that we can preserve the existing labor-market order by slowing or containing the technology until our familiar institutions feel safe again. Much work is meaningful, dignifying, and socially necessary. Much work is also coercive, degrading, badly paid, alienating, or organized around domination. The aim cannot be to preserve dependence on wage labor as such. The aim must be to preserve and expand human agency, security, dignity, belonging, and freedom as the economic basis of work changes.

The third is globalization déjà vu: the belief that society can let the transformation proceed, measure the aggregate gains, promise retraining and compensation, and trust that people will eventually adjust. That would be a profound mistake. The globalization era showed that people do not experience transformation only as workers or consumers. They experience it as members of places, families, communities, occupations, and nations. They ask not only whether prices are lower or GDP is higher, but whether their lives remain legible and respected inside the new order. AI will intensify this question because it reaches not only industrial labor, but cognitive work, creative work, institutional authority, public trust, and the meaning of contribution itself.

Polanyi teaches us that markets can be socially destructive when they are disembedded from social purpose. Globalization teaches us that efficiency without legitimacy produces backlash. We don’t have to learn the AI lesson the hard way.

The task is not to stop AI. The task is to re-embed it.

But re-embedding should not mean forcing machine intelligence into old institutional forms and calling the job done. This is where AI may be not only the object of governance but part of the means by which governance improves. If machine intelligence lowers the cost of analysis, coordination, translation, auditing, representation, bargaining, and participation, then a Great Transition should use those capabilities to create forms of democratic agency that were previously too expensive, too slow, or too difficult to sustain.

The goal is not simply to constrain AI from above. We can use AI to make bottom-up coordination, public deliberation, collective bargaining, institutional auditing, and democratic authorship more feasible.

That re-embedding has four dimensions: security, share, sovereignty, resilience and meaning. It is also conditional on establishing and maintaining a shared world in which our society knows what is happening and decide together what to do.

1. Security: economic stability beyond continuous wage labor

If intelligence becomes more abundant while income remains tied to scarce employment, the result will not be a frictionless age of creativity. It will be insecurity, humiliation, and political rage. The wage-labor economy does not merely distribute income. It distributes identity, routine, status, discipline, bargaining power, social insurance, social mobility, and political legitimacy. When people ask whether AI will “take jobs,” they are often asking a deeper question: what happens to a society whose moral and material order is built around work when work itself becomes less central to production?

This isn’t just hypothetical. Acemoglu and Restrepo show that technological change simultaneously displaces existing labor and creates new tasks, but the balance depends on institutional choices rather than technological inevitability. Recent estimates from the IMF and the International Labour Organization indicate that generative AI may affect a substantial share of cognitive and professional occupations, particularly in advanced economies. The IMF estimates that roughly 60% of jobs in advanced economies are exposed to AI — not necessarily doomed, but exposed to either significant augmentation or disruption. Anthropic’s Economic Index adds another kind of early evidence: AI usage is already visible across real tasks and occupations, showing both augmentation and automation.

Labor markets will change, but can societies shape that change to broaden opportunity rather than concentrate vulnerability?

A Great Transition must therefore partially decouple economic security from continuous labor-market participation. That does not mean one policy only. We should encourage experimentation. Solutions could include a mix of stronger social insurance, universal basic services, wage insurance, child allowances, portable benefits, public employment options, shorter working weeks, care credits, lifelong learning accounts, sovereign wealth funds, forms of universal capital ownership, or social dividends linked to AI-driven productivity gains. The design matters, but the principle that people should not be asked to absorb a transformation in the value of labor while bearing the risks individually matters more.

Globalization teaches us that compensation that arrives late, narrowly, and bureaucratically is not transition. It is apology. A genuine transition makes security visible early enough that people can experience change as something society is navigating with them, not something the economy is doing to them.

Security is not a retreat from ambition. It is the platform from which people can adapt and contribute without being crushed.

2. Share: visible participation in AI-driven gains

The gains from AI must be visibly shared.

This is not just a moral claim. It is a legitimacy claim. A transformation that produces extraordinary productivity while concentrating wealth, compute, data, and decision power in a handful of firms will not remain politically stable. If the public experiences AI as a system that absorbs human knowledge, automates human work, degrades public trust, and returns the gains primarily to capital, then the countermovement will come. The only question is whether it will be democratic, emancipatory, and constructive — or reactionary, authoritarian, and destructive.

The globalization backlash should make this obvious. Aggregate gains are not self-legitimating. If the winners are visible, the losses local, and the public story dismissive, people will not accept the transformation as progress simply because the spreadsheet says it is efficient.

A Great Transition therefore has to confront ownership. Who owns productive intelligence? Who controls the models, data centers, chips, interfaces, agents, datasets, deployment channels, and standards? Who receives the returns when cognitive work is automated? Who has bargaining power when AI systems mediate employment, credit, healthcare, education, insurance, public services, and public voice?

These are constitutional questions in disguise.

Earlier economic settlements built institutions around the general conditions of production. Roads, courts, schools, public health systems, central banks, communications networks, research universities, and public infrastructure did not eliminate markets. They made complex market societies possible. If machine intelligence becomes a general condition of production, then some form of public AI capacity will become equally necessary.

As Blayne Haggart has argued in the context of the data-driven economy, markets for information-intensive systems are politically constructed. The question is not whether governments shape AI ecosystems, but how.

That might mean public-interest foundation models, publicly governed data trusts, compute access for researchers and civic institutions and small businesses and local governments, democratic procurement standards for public-sector AI, shared-prosperity targets for developers and deployers, or treating certain AI capabilities as public capabilities. Recent proposals for public AI, including those advanced by Nathan Sanders, Bruce Schneier, and Norman Eisen, point toward models that combine innovation with democratic accountability.

But public AI should not be imagined only as state AI. A society with machine intelligence needs plural public capacity — not one government model but an ecology of them: civic and open models, private systems, user-owned agents, public-interest intermediaries, and procurement rules that align private markets with public ends. What matters is the ecology, not the state.

The point is not that the state should build everything but that a society mediated by AI should not rely exclusively on privately owned intelligence systems any more than industrial society could rely exclusively on privately built roads, privately enforced contracts, privately issued money, or privately educated citizens.

3. Sovereignty: democratic agency over intelligent systems

A Great Transition should also be democratic.

One of the weaknesses of many AI governance regimes is the promise of protection without participation. Experts classify risks. Firms publish principles. Regulators define compliance processes. Institutions speak in the language of fundamental rights. And citizens are subjects to be protected rather than authors of the system they must live under.

That’s not enough. The democratic opportunity of the AI transition is not merely to reduce harm but to increase the meaningful involvement and responsibility of individuals in the systems that represent them. AI could become another layer of technocratic distance between people and power or it could become a tool for making democratic participation broader, more informed, more continuous, and more legitimate.

Nancy Fraser’s extension of Polanyi is useful here. Polanyi described a “double movement” between marketization and social protection. Fraser argued that modern societies also contain a third force: emancipation. Protection can be necessary, but it can also become paternalistic. A humane AI transition should not merely protect people from disruption, but expand their freedom and deepen their agency.

The same point appears in the language of resilience. The Imagining the Digital Future Center’s report, Building a Human Resilience Infrastructure for the AI Age, distinguishes between adaptive resilience and agency-based resilience. Adaptive resilience helps people cope with change. Agency-based resilience helps people evaluate, contest, and reshape the conditions they are adapting to. A society in which people calmly adapt to systems they did not choose is not free. It is merely well-managed.

Democratic sovereignty in the AI age cannot mean individual self-sufficiency. No person can inspect every model, negotiate every data relationship, or audit every algorithmic system that touches their life. It must mean institutionalized agency: rights to explanation; democratic oversight of public-sector systems; liability for autonomous harms; limits on behavioral manipulation; and public power sufficient to shape the infrastructures through which people increasingly act.

It can also mean new forms of participation, through a hybrid deliberative model for the transition: expert assemblies to clarify the choices, citizen assemblies to legitimate the direction, and digital civic infrastructure to keep participation open, iterative, and consequential. Communities could deliberate over AI in schools, policing, benefits, zoning, healthcare, and public services. Nationally, citizens’ assemblies and transition observatories could shape social insurance, labor-market policy, public AI capacity, and high-impact AI rules. Globally, expert and citizen processes could give countries outside the frontier-model race a voice in the rules of AI diffusion, infrastructure, safety, and access, rather than leaving them to absorb decisions made elsewhere.

Here, AI itself has the power to change the institutional possibility space — while also containing the threat of being turned against itself.

In the past, broad democratic participation has been limited by the cost of coordination. People do not have time to read policies, attend hearings, understand tradeoffs, or negotiate the systems that affect their lives. Agentic representation could make those costs lower.

The optimistic version is concrete enough to picture. Personal agents that help citizens understand policy tradeoffs, compare proposals, and monitor what their representatives actually do. Worker agents that help employees see how AI is being deployed around them and hold a bargaining position. Civic agents that let a community interrogate a data-center proposal or a policing tool. Public-interest agents that audit official claims and translate technical documents back into democratic language.

Of course, every one of those capabilities is potentially dual-use. An agent that can summarize a hearing can fabricate a thousand fake submissions to it. An agent that helps a worker bargain can be the instrument an employer uses to model and defeat the bargain. The deeper trap is the asymmetry: powerful institutions will field these agents whether or not citizens can, so the real question is not whether agentic representation exists but whether it is distributed — and on that question the optimistic case is a hope, not yet a result. A Great Transition should ask what forms of agent-mediated representation, bargaining, auditing, and collective action are compatible with democracy — and then build them deliberately, or watch them get built against the public by default.

That requires rules and institutions: verified civic identity without mass surveillance; disclosure of automated participation; limits on agentic lobbying and astroturfing; public-interest agent standards; democratic audit trails; collective representation for groups using agents; and rights for citizens to use agents when dealing with powerful institutions. If companies and governments will use AI agents to act strategically, citizens and workers will need agents of their own.

A Great Transition should therefore be understood not only as a policy agenda but as a democratic capacity-building project. The task is to use AI to make people more capable citizens, not less necessary ones.

4. Resilience and meaning: living well with machine intelligence

The hardest part of the transition may be cultural.

Modern societies do not simply depend on work economically. They depend on work morally. Work is how people prove discipline, earn recognition, structure time, form adult identity, and justify claims on social resources. Even those who dislike their jobs often live within a culture that treats employment as the primary evidence of contribution. This is why a purely redistributive response to AI will be insufficient. Money matters enormously, but people need more than income. They need roles, obligations, recognition, community, responsibility, and a story about why their lives matter.

The globalization era showed this vividly. When work left communities, people did not only lose wages. They lost routine, informal institutions, local and personal pride, intergenerational continuity, and confidence that the future had a place for them. The resulting politics cannot be understood only through income. It was also a politics of meaning.

AI may produce a wider version of the same problem. If cognitive labor, professional identity, creative production, and institutional judgment are all transformed at once, the challenge will not be limited to displaced workers. It will reach teachers, lawyers, designers, managers, journalists, analysts, researchers, public servants, and artists — anyone whose sense of contribution has been organized around forms of judgment that machines may increasingly perform.

The Human Resilience Infrastructure report is useful here because it shifts attention from individual adaptation to social scaffolding. Its contributors emphasize agency, autonomy, institutional restructuring, existential literacy, shared reality, human connection, and the risk that people will passively adapt to algorithmic systems. The core insight is that resilience cannot be left to individuals alone. It has to be built into institutions, education, governance, communities, and norms.

This strengthens the Great Transition frame. Economic security is resilience against labor-market shock. Shared ownership is resilience against concentrated power. Democratic sovereignty is resilience against dependency on opaque systems. Epistemic institutions are resilience against synthetic reality. Post-work cultures of contribution are resilience against meaning collapse. AI literacy and existential literacy are resilience against manipulation, over-dependence, and the erosion of judgment.

A Great Transition must ask what comes after the work ethic without collapsing into emptiness or indulgence. The answer is not a society without effort, ambition, mastery, or contribution. It is a society with more forms of contribution than paid employment alone can recognize: cultivation of the human experience and excellence, democratic participation and civic service, learning, art, science, mentoring, local and cultural stewardship, ecological repair, and spiritual life.

The point of an AGI-informed economy should not be to make human beings economically unnecessary. It should be to make more of human life possible.

5. Shared Reality: Creating a world in which a transition can be chosen

The first market society strained the material foundations of life: labor and land. The globalization era strained the political foundations of legitimacy: trust that decision makers understood the consequences of the system they were building, and trust that democratic institutions could make the gains and losses of transformation visible, fair, and contestable. The AI society may strain the epistemic and symbolic foundations: trust, attention, evidence, reputation, and the shared sense of what is true.

Synthetic media, automated persuasion, personalized propaganda, agentic lobbying, machine-generated intimacy, and algorithmic reputation systems all point toward a future in which social reality itself becomes programmable.

This may be the least developed part of the current policy debate. It is also the part on which everything else depends.

A society cannot govern itself if it cannot establish shared facts. It cannot sustain public reason if every institution is flooded by machine-generated speech calibrated for advantage. To date, the idea that digital platforms could reliably predict and control human behavior has often been overstated. Human beings remain stubborn, social, contextual, contradictory, and challenging to model - but generative and agentic AI could make a version of the concern much more real.

The danger is not behavioral control but the further atomization of the environments in which people form preferences, judgments, and trust. The thing most at risk is not any particular fact, but society’s capacity to hold a single hard question in common view long enough to decide it together. The risk is not that everyone believes the same falsehood but that the most important question never receives collective focus until it is too late.

Agentic systems intensify this problem. If organizations can deploy agents to persuade, negotiate, lobby, or personalize political pressure at scale, then democratic institutions will need new defenses. The problem is not only fake content. It is fake publicness: the appearance of collective attention, collective pressure, or collective consent where no real public has formed.

That is why reality is the keystone of the transition rather than one item on its agenda. Public life requires institutions capable of preserving the conditions of shared judgment: provenance standards, public-interest media, civic education, trusted archives, model transparency for high-impact uses, protections against synthetic impersonation, contestability for high-impact AI-mediated decisions, and norms that preserve human accountability in public life.

Reality is not simply an information problem. It is an infrastructure problem. A society mediated by machine-generated content needs public systems for verification, provenance, civic interpretation, and institutional credibility.

A non-event or democratic authorship?

We should also consider that none of this requires a villain.

I believe the companies building frontier intelligence are sincere and that governments are acting in what they understand as the public interest, even if that interest is contested. Engineers can be genuinely motivated by safety and human benefit. The argument does not depend on good faith or bad faith. It depends on structure.

Whoever builds the infrastructure, standards, interfaces, and dependencies first shapes the terrain on which later politics must operate. Once schools, hospitals, firms, militaries, public agencies, and households depend on particular AI systems, the range of realistic democratic choices narrows. The question is no longer “What settlement should we build?” It becomes “How do we manage the settlement we have already become dependent on?”

A Great Transition would move choice back upstream.

This is Polanyi’s oldest lesson in new clothing. The self-regulating market was not a conspiracy. It was a system that overrode the intentions of the people inside it. Globalization was not a conspiracy either. It was a system of incentives, rules, efficiencies, and elite assumptions that outran the legitimacy of the communities it transformed.

So what leverage does society hold against a transformation with no one clearly at the wheel?

Not labor alone. That leverage has already been weakening for generations. Not voting alone; you cannot vote a vision into power that no one is discussing.

What is necessary is collective attention: the shared act of a society turning to face a question together. What is necessary is public engagement and democratic authorship.

Everyone deserves to know the next Great Transformation is happening — not as a narrow risk filed under future-of-work, not as a proxy fight over data centers, not as a compliance exercise, not as a national-security race, but as the defining settlement of the century, being drafted now in infrastructure, markets, institutions, and code. And they deserve to know it does not have to go badly: that there are positive visions for the future and institutions capable of making the transition more secure, more democratic, more humane, and more free.

For sure, frontier AI companies have a vital role to continue playing on this front, as do researchers, civil society, workers, educators, communities, and users. But the burden cannot remain with the companies alone. Labs are inviting governance. Governments and democratic institutions must respond with a transition plan.

I am inherently suspicious of the idea that broad awareness should always be the first, best answer. But in this case, it is morally necessary.

Our task is not only to wake people up but to give them institutions through they can participate in shaping the future contestable before it hardens into infrastructure.

We should not let the greatest opportunity to advance humanity polarize into a standoff between people and AI or between people and AI companies. Machines will continue to become more intelligent, but will intelligence become another disembedded system to which society must adapt after the fact? Or will it be a public capability embedded in human flourishing from the start?

The first Great Transformation created today’s market-based society. The globalization era showed what happens when transformation is treated as an efficiency problem that leaves the social consequences as something for politics to deal with later. The next Great Transformation will test whether society can govern machine intelligence before machine intelligence governs society.