Advanced automation and artificial intelligence are affecting the functioning of the economy at a level deeper than simple productivity. Their primary impact concerns the distribution of economic agency, that is, the effective capacity to decide, orient, and assume consequences within processes of value creation. In every economic order, agency is never evenly distributed: it is the result of institutional arrangements, dominant technologies, and organizational conventions. The systemic adoption of intelligent systems introduces a discontinuity in this balance, progressively shifting the locus of decision from visible action to the invisible structure of models.
The first structural effect is the displacement of decision-making power toward technical architecture. Decisions that previously required situated evaluation, comparison among alternatives, and direct assumption of risk are increasingly delegated to automated decision support systems, capable of processing volumes of information and combinations of variables beyond human reach. According to the OECD, more than 60 percent of large enterprises in advanced economies already use algorithmic systems to guide operational and strategic decisions, from inventory management to personnel selection. In this transition, decision-making does not disappear, but is anticipated and channeled: what appears as choice is often the outcome of a perimeter of possibilities defined upstream by the model.
This relocation of agency has direct effects on work. Labor autonomy is not simply reduced at the operational level, but restructured at the cognitive level. Human action is increasingly positioned as assisted execution, monitored and corrected in real time by systems that establish performance thresholds, behavioral standards, and criteria of acceptable error. Recent studies by the MIT Sloan Management Review show that the intensive use of decision support systems increases short-term efficiency, but tends to reduce operators’ capacity to exercise independent judgment in the medium term. Individual agency is not eliminated, but subordinated to a rationality embedded in the system.
At the organizational level, this transformation produces more coherent and centralized arrangements, but also ones more dependent on opaque logics. Decision chains formally shorten, while implicitly lengthening along the system design phase. Leadership thus finds itself governing not only people and processes, but also algorithmic assumptions that orient collective behavior. This condition is well known in large digital platforms, where effective power resides less in hierarchical roles and more in control over parameters that regulate ranking, allocation, and visibility.
In this framework, reference to Gilbert Simondon makes it possible to grasp the non-instrumental nature of technology. For Simondon, technical objects are not mere tools, but forms of mediation that structure the relationship between individual and collectivity. Advanced automation, read from this perspective, does not simply subtract action from the human being, but redefines the criteria of legitimacy of action itself. When what is considered rational coincides with what the system optimizes, human agency risks being reduced to a function of adaptation to decisions already implicitly taken.
The social and economic implications of this redistribution of agency are significant. Responsibility tends to fragment along the decision chain, making it more difficult to attribute outcomes to identifiable subjects. According to the World Economic Forum, one of the main risks associated with the adoption of AI in decision-making processes is precisely the dilution of accountability, with direct effects on institutional and organizational trust. When a decision is perceived as “taken by the system”, the possibility of contestation and collective learning is reduced, even in the presence of suboptimal outcomes.
In the business world, this tension manifests itself in the relationship between performance and governance. Enterprises that intensively use automated systems often achieve significant competitive advantages, but face the paradox of decision-making power that is more effective and at the same time less visible. Traditional governance, based on clearly attributable responsibilities, struggles to intercept decisions emerging from the interaction between data, models, and operational context. Some recent cases in the financial and logistics sectors show how systemic errors can propagate rapidly precisely because no actor perceives themselves as fully responsible for the initial choice.
Human cognitive limits, frequently invoked to justify automation, are not eliminated, but displaced. The human being renounces deciding on the individual case, but must decide on the overall system architecture, a task that requires abstract competencies, forecasting capacity, and broader responsibility. This asymmetry risks concentrating economic agency in a few technical and organizational nodes, removing it from the diffuse control that has historically supported the legitimacy of economic decisions.
The reduction of labor autonomy, if not accompanied by new forms of participation and understanding, tends to produce corrosive effects on the organizational fabric. Work loses its dimension of choice and is reduced to procedural conformity, widening the distance between those who design systems and those who use them daily. This fracture, already observable in several highly automated sectors, is not only an issue of individual well-being, but a problem of systemic resilience.
Automation as a new watershed of economic agency therefore does not represent a simple technical advancement, but a cultural and institutional passage of broad scope. It redefines who acts, who decides, and who is accountable, shifting the economic center of gravity toward less visible but more pervasive forms of power. Understanding this transformation means recognizing that artificial intelligence is not neutral with respect to the distribution of economic action, but becomes one of its main factors of reorganization.
In this open scenario, the challenge for enterprises and institutions does not consist in recovering an idealized autonomy of the past, but in reconstructing the conditions of responsible action in a system-mediated context. Economic agency is not annulled, but redistributed along new lines of force. The quality of the economic future will depend on the capacity to make these lines legible and governable, avoiding the illusion that decision-making can ever be fully automated without political and social consequences.
Global AI Observatory