Artificial intelligence does not introduce a simple reorganization of tasks, but acts as a principle of stratification that profoundly redefines the architecture of professional inequalities. It does not merely replace human labor with automation, but restructures the way value is attributed to different forms of productive contribution. In this sense, AI operates as a systemic factor that progressively separates function, recognition, and decision-making power, generating a new morphology of work that cuts across firms, markets, and institutions.
Occupational polarization is the first observable structural effect. OECD data show that, in advanced economies, employment growth over the past decade has been concentrated mainly at the high and low ends of the skill spectrum, while intermediate professions have experienced a significant contraction. Artificial intelligence accelerates this dynamic because it directly affects those activities of cognitive mediation that for decades constituted the core of the professional middle class: standardized analysis, operational coordination, procedural evaluation, and mid-level planning. These functions, once sustained by widely distributed human skills, are progressively absorbed by predictive systems and decision support tools.
The erosion of intermediate professions is not a side effect of innovation, but a logical consequence of cognitive automation. These roles performed an organizational stabilization function, translating strategic visions into executable processes and ensuring continuity between top management and the operational base. When AI assumes this role of translation and coordination, organizations become leaner and faster, but also more exposed to systemic shocks. In the corporate world, numerous cases show how an excessive reduction of intermediate layers increases dependence on models and reduces the ability to detect weak signals emerging from day-to-day operations.
At the opposite end of the stratification, new cognitive elites emerge. They are not defined solely by level of education or hierarchical position, but by privileged access to decision-making systems, critical data, and algorithmic architectures. According to estimates by the World Economic Forum, less than 5 percent of the global workforce is currently directly involved in the design, training, and governance of artificial intelligence systems, yet this minority exercises a disproportionate influence over strategic decisions and the allocation of value. This power is often opaque, because it is embedded in parameters, models, and optimization criteria rather than in explicit decision acts.
Labor marginalization represents the other side of this process. It does not coincide exclusively with unemployment, but with the expansion of a residual zone of work characterized by low protection, limited mobility prospects, and direct competition with automated systems. Work remains necessary, but progressively loses its capacity to guarantee social integration. According to the International Labour Organization, a growing share of workers in the service sectors perform tasks that contribute little to income growth and are highly exposed to forms of partial automation, without benefiting from the productivity gains generated by AI.
These dynamics cannot be understood exclusively in economic terms. The thought of Pierre Bourdieu offers a particularly effective interpretative key, showing how inequalities are reproduced through the distribution of cultural and symbolic capital. In an economy driven by artificial intelligence, cognitive capital becomes the primary factor of distinction. The new elites do not merely possess technical skills, but control the interpretative codes that define what is considered value, risk, and efficiency. Those who remain excluded from these codes lose not only income, but the very possibility of influencing the rules of the professional game.
Within firms, this stratification is directly reflected in leadership and governance models. The widespread adoption of intelligent systems reduces certain human cognitive limits, but amplifies others, especially those linked to uncritical delegation. When cognitive power concentrates in a few technical nodes, the risk is a form of leadership that merely validates algorithmic outputs, progressively losing autonomous judgment capacity. More mature organizations have begun to recognize this risk, introducing functions of model control and audit precisely to prevent the concentration of expertise from translating into a deficit of responsibility.
The macroeconomic effects of occupational polarization are equally significant. The compression of intermediate professions weakens aggregate demand and accentuates the fragility of welfare systems, historically built around a broad base of stable and qualified work. The International Monetary Fund has repeatedly highlighted how rising occupational inequalities reduce long-term growth potential, creating economies that are more efficient at the micro level but less resilient at the systemic level.
The professional hierarchies that emerge in the artificial intelligence economy are therefore not an inevitable outcome of technological progress, but the result of implicit choices about how to distribute power, value, and recognition. Making them visible is a necessary condition for governing them. Firms and institutions that are able to confront this transformation not by artificially reconstructing the past, but by rethinking mechanisms of integration and responsibility, will gain an advantage that goes beyond immediate productivity.
Work, in the era of AI, thus becomes a sensitive indicator of the quality of collective decisions. The stratifications that are forming clearly reveal which forms of intelligence a society chooses to reward and which it is willing to marginalize. In this still unstable space, the future of work remains open not due to a lack of technical solutions, but because it implies a political and institutional decision about the type of social order one intends to make sustainable.
Global AI Observatory