Rules Changed

Rules Changed: How AI Can Break Institutions' Historical Role

Douglas North stated that institutions are the "rules of the game" - the formal and informal laws that structure human interaction. Since the publication of his landmark book Institutions, Institutional Change and Economic Performance (1990) that framing has held arbitrarily true across disciplines, from law to economics, from political science to sociology. Governments legislate, courts adjudicate, central banks calibrate, and markets price - each playing a prescribed role within a web of understood constraints. However, the new agentic and generative artificial intelligence age might disrupt the game set, the players, and - most fundamentally - the rules themselves. We are not facing a mere technological upgrade; we may be confronting a structural rupture in how authority, information, and accountability are distributed across modern societies, and, as you may know, uncertain expectations and imperfect information usually lead to rushed decision making and societal welfare losses.

Therefore, the question is no longer whether AI will reshape institutions. It already is. The question is whether institutions, built on centuries of human agency, can adapt swiftly enough to govern what they did not see coming, or whether AI will rewrite the rules before anyone agrees on what the new game should be.

Classical economic theory rests on the premise that actors - firms, consumers, governments - respond to incentives within known institutional constraints. But generative AI introduces a new type of actor: one that can simulate, advise, and increasingly execute economic decisions at machine speed, at negligible marginal cost, and with an opacity that traditional regulatory frameworks were never designed to handle.

The International Monetary Fund warned in 2024 that AI could affect up to 60% of jobs in advanced economies, with effects ranging from productivity augmentation to full displacement. Unlike previous waves of automation, which primarily replaced routine manual tasks, AI now encroaches on cognitive work: drafting contracts, reading medical scans, coding software, generating financial models. According to McKinsey Global Institute’s 2023 report on the economic potential of generative AI, the automation potential of tasks requiring expertise - previously considered safe from automation - jumped 34 percentage points with generative AI, fundamentally redefining which tasks are considered to require human intelligence at all.

This creates a fundamental tension within the institutional architecture of market economies. The labour market is perhaps where the institutional disruption will be most socially felt. Trade unions, minimum wage legislation, employment contracts, and social insurance systems were all designed around a core assumption: that human labour is the primary input in economic production, and that the returns to that labour must be negotiated and protected. Such a rapid change can mean a total disruption in labour institutions’ role and power.

Generative AI disrupts this assumption not by eliminating jobs wholesale - at least not yet - but by restructuring the value of skills in ways that outpace institutional adaptation. The OECD's 2023 Employment Outlook found that AI exposure is highest among high-education, high-wage workers - the inverse of previous automation waves, which disproportionately affected low-skill, routine occupations. Lawyers, analysts, consultants, and journalists face higher displacement risk than, say, plumbers or care workers. This creates a paradox for labour institutions: the workers most likely to be displaced are also the least represented by traditional collective bargaining structures.

The gig economy had already weakened the standard employment relationship. AI threatens to accelerate this fragmentation by enabling "micro-tasking": the decomposition of professional work into discrete, AI-assisted sub-tasks that can be outsourced globally at minimal cost. If a law firm can use an AI system to conduct initial case research, draft standard pleadings, and review contracts - leaving only the highest-order judgment calls to human lawyers - the institutional bargaining power of the legal profession as a labour force diminishes accordingly. As Microsoft AI CEO Mustafa Suleyman told the Financial Times in February 2026, most tasks performed by lawyers, accountants, project managers, and marketing professionals will be “fully automated by an AI within the next 12 to 18 months” - a prediction that pushes every professional institution to clarify whether human judgement is truly irreplaceable.

Nonetheless, it would be intellectually dishonest to present these developments as certainties. The history of transformative technology is littered with predictions of institutional collapse that did not materialise, and with complacency about changes that did. The printing press was expected to destroy the Church's authority over knowledge; it did, but it also created conditions for the Reformation, new literacy, and eventually modern democratic institutions. The internet was expected to abolish privacy; it largely has, but it also created new forms of civic coordination and economic value.

AI may follow a similar pattern: not elimination of existing institutions, but their radical reconfiguration. What is clear is that the reconfiguration is already underway, and that institutions are lagging behind the pace of change. The EU AI act (2024), took years to negotiate and will take years to implement in full - during which AI capabilities will have evolved several generations further. The institutions of law, finance, labour, and governance were built for a world in which humans were the only entities capable of complex, consequential judgment. That world is ending.

Douglas North also observed that the gap between formal rules and actual behaviour - what he called "informal constraints" - is where significant institutional change happens, often invisibly, before it is formally recognised. The challenge for policymakers, legal scholars, economists, and citizens alike is to close it before the distance becomes unbridgeable - before the rules of the game are rewritten by actors who were never invited to the table.