Autonomous AI Agents and the Reorganization of Power: A Critical Sociology of Management, Tourism, and Technology in 2025

Author: Miguel López

Affiliation: Independent researcher

Received 5 August 2025; Revised 10 Sep 2025; Accepted 05 Oct 2025; Available online 30 Oct 2025; Version of Record 30 Oct 2025; Post-Publication Update 20 Jun 2026.

https://doi.org/10.65326/u7y566820-2

Volume 2, December 2025, (10020-2)

10020 - 2.pdf

Download PDF • 446KB

Abstract

Autonomous AI agents—software systems that can plan, decide, and act toward goals with limited human supervision—moved from isolated pilots to enterprise deployment over 2024 and 2025. This article develops a critical sociology of that shift across corporate management, tourism, and the technology supply base. Integrating Pierre Bourdieu’s theory of capital, field, and habitus; world-systems analysis; and institutional isomorphism, it advances a conceptual framework for how agentic AI reorganizes power within firms and across the global division of digital labor. The analysis yields seven propositions: agent adoption relocates authority toward actors who control data, orchestration, and governance; it gives rise to a distinct field of agent governance in which demonstrable control becomes a form of capital; it reconfigures the managerial habitus around the supervision of software actors; it tends to reproduce core–periphery asymmetries anchored in the concentration of models, compute, and standards; in intermediated sectors such as tourism it concentrates value capture while externalizing local social and environmental costs; coercive, mimetic, and normative pressures push organizations toward convergent agent architectures; and the legitimacy of deployment increasingly rests on symbolic markers of responsible control. A seven-layer governance model translates these claims into design guidance. The argument is conceptual and interpretive, and is intended to frame an agenda for empirical research on the political economy and organizational sociology of autonomous agents.

Keywords: autonomous AI agents; organizational power; Bourdieu; world-systems analysis; institutional isomorphism; AI governance; tourism; political economy of AI

1. Introduction

Over the course of 2024 and 2025, organizations began to delegate multistep work—procurement checks, scheduling and routing, content production, risk monitoring, and the orchestration of customer service—to autonomous AI agents that operate across human and digital environments within defined constraints. Industry analysts marked the turn explicitly: agentic AI was named the leading strategic technology trend for 2025, with the expectation that a measurable share of routine work decisions would, within a few years, be taken autonomously rather than by people (Gartner, 2024). The technical basis for this shift is the use of large language models as planning and tool-using cores, which allow software to interpret goals, call services, monitor outcomes, and adapt with limited supervision (Wang et al., 2024). The phenomenon is therefore neither a marketing label nor a simple extension of earlier automation; it is a change in the locus of organizational decision-making.

Existing scholarship has examined automation and artificial intelligence primarily through two channels. One estimates the labor-market consequences of automating tasks, documenting displacement and wage effects of robots and, more recently, of AI as expressed in hiring patterns (Acemoglu & Restrepo, 2020; Acemoglu et al., 2022). The other interrogates the political economy of digital infrastructures—platform capitalism, surveillance capitalism, and data colonialism—showing how value is captured through the appropriation of data and the control of connective infrastructure (Srnicek, 2016; Zuboff, 2019; Couldry & Mejias, 2019; Crawford, 2021). Both literatures are indispensable, yet neither was written for systems that act. Task-displacement models treat technology as a substitute for labor inputs rather than as an actor that exercises delegated authority; political-economy accounts analyze data extraction and platform power but say little about how the internal authority structure of the firm is reorganized when software, not staff, executes consequential decisions. The result is a gap at the meeting point of organizational sociology and the global political economy of computation: how the delegation of action to autonomous agents redistributes power inside organizations and reproduces or unsettles asymmetries between them.

This article addresses that gap. It asks who gains and loses different forms of capital as agents are adopted; how fields of practice and professional dispositions adapt; whether agentic AI deepens or disturbs core–periphery relations in the world-system of data, compute, and standards; and why organizations operating in very different contexts converge on similar governance arrangements. To answer these questions it integrates three sociological frameworks—Bourdieu’s theory of capital, field, and habitus (Bourdieu, 1977, 1986, 1990); world-systems analysis (Wallerstein, 2004); and the theory of institutional isomorphism (DiMaggio & Powell, 1983)—and applies them to three settings chosen for analytical contrast: corporate management, tourism, and the technology supply base.

The contribution is threefold. Conceptually, the article extends field theory to the supervision of autonomous software, proposing that demonstrable control—what we term governability—becomes an emergent form of capital around which a new organizational field forms. Theoretically, it connects organizational power to the global political economy of AI, specifying the mechanisms—model control, compute concentration, and standard-setting—through which agentic AI can reproduce core–periphery asymmetries. Practically, it offers a seven-layer governance model that translates the analysis into design guidance and reframes adoption as a question of power and legitimacy rather than efficiency alone. The remainder of the article defines agentic AI and its diffusion, presents the theoretical framework and research design, develops the analysis through seven propositions across the firm, sectoral, and global levels, and discusses the contribution before noting limitations and a research agenda.

2. Autonomous AI Agents in Organizations: Concept and Diffusion

2.1 Defining autonomous AI agents

An autonomous AI agent is a system that interprets a goal, plans a sequence of actions, invokes tools or services, observes the results, and adjusts its behavior to reach the objective with limited human intervention (Wang et al., 2024). Two features distinguish agents from earlier automation. First, agents reason iteratively rather than execute fixed scripts, which allows them to handle tasks whose steps are not specified in advance. Second, they act—they place orders, adjust prices, open tickets, or reconfigure schedules—rather than merely producing outputs for a person to act upon. In enterprise settings these capabilities are embedded in resource-planning, customer-relationship, supply-chain, and analytics platforms, and are increasingly coordinated through orchestration layers that route tasks among multiple agents and escalate to humans when constraints or uncertainty require it. The salient sociological fact is not the technique but the delegation: a measure of organizational authority is transferred to a non-human actor (Latour, 2005).

2.2 The diffusion of agentic AI

Adoption is uneven but broad enough to alter managerial routines and labor processes. It proceeds along three reinforcing pathways. The infrastructural pathway lowers integration costs through cloud services, connectors, and data stores, so that capability can be assembled rather than built. The organizational pathway creates new roles—policy design, safety evaluation, monitoring, and red-teaming—and new bodies, such as cross-functional councils that adjudicate escalation rules and acceptable use. The cultural pathway changes work itself: managers learn to supervise software, frontline staff move from execution toward exception handling, and performance metrics expand from throughput toward quality, alignment, and accountability. These pathways echo earlier accounts of how digital technologies reorganize firms and economies (Brynjolfsson & McAfee, 2014; Castells, 1996; Davenport & Ronanki, 2018) and of how AI reshapes whole institutions rather than discrete tasks (Katsamakas et al., 2024). What is new is that the technology being diffused is an actor with delegated discretion, which is why its diffusion is also a redistribution of authority.

3. Theoretical Framework

Three frameworks structure the analysis, each addressing a different level at which power is organized: the intra-organizational, the global-systemic, and the inter-organizational field.

3.1 Capital, field, and habitus

Bourdieu’s sociology treats social life as a struggle for position within relatively autonomous fields, waged with different species of capital and guided by habitus, the durable dispositions that make certain actions feel natural to actors in a given position (Bourdieu, 1977, 1990). Capital takes economic form (budget, compute, data-acquisition capacity), cultural form (technical and governance expertise, domain knowledge), social form (the relationships that secure data, partnerships, and preferential access), and symbolic form (the prestige and legitimacy of being seen to operate responsibly or at the frontier) (Bourdieu, 1986). Read through this lens, agentic AI is a stake over which these capitals are contested and converted, not merely a tool that firms possess or lack.

3.2 World-systems analysis

World-systems analysis positions the global economy as a hierarchy of core, semi-periphery, and periphery, in which surplus flows toward actors that control the most valued capacities (Wallerstein, 2004). In the political economy of computation, those capacities are foundational models, large-scale compute, and the standards that govern interoperability and audit. Critical accounts of data colonialism and platform capitalism show how value is appropriated through the control of connective infrastructure and the conversion of social life into data (Couldry & Mejias, 2019; Srnicek, 2016; Crawford, 2021). Agentic AI extends this dynamic: when planning, action, and standards are anchored in core infrastructures, peripheral organizations may act through agents they do not control and on terms they did not set.

3.3 Institutional isomorphism

DiMaggio and Powell (1983) explain why organizations in a field come to resemble one another through three mechanisms: coercive pressure from regulation and dependency, mimetic imitation of perceived leaders under uncertainty, and normative pressure from professions that define legitimate practice. The framework is well suited to a fast-moving technology whose governance templates are still forming. Evidence from the adjacent domain of AI ethics is indicative: a global mapping found convergence on a small set of principles alongside persistent divergence in how they are implemented (Jobin et al., 2019), the signature pattern of isomorphism operating before stable templates exist.

3.4 An integrated framework

The three lenses are complementary rather than redundant. Bourdieu specifies the intra-organizational stakes and the actors who win or lose position; world-systems analysis situates those struggles within a global hierarchy of computational capacity; institutional isomorphism explains why responses converge across organizations and how legitimacy is conferred. Together they move the study of agentic AI beyond questions of technical feasibility toward questions of power, value capture, and legitimacy. The research design below operationalizes this integration.

4. Research Design

This is a conceptual, theory-building study. Its aim is to construct and integrate a framework that explains an emerging phenomenon and to derive propositions for subsequent empirical testing, rather than to test hypotheses against primary data. The design follows three commitments: explicit selection logic for the theories and settings analyzed, a transparent analytical procedure, and clearly stated scope conditions.

The three frameworks were selected for complementary levels of analysis and for their established standing in organizational sociology and the political economy of technology. Bourdieu’s framework supplies a micro-to-meso account of authority and legitimacy within organizations; world-systems analysis supplies a macro account of global asymmetry; institutional isomorphism supplies a meso account of inter-organizational convergence. Each has an extensive record of application to technological and institutional change, which allows the present synthesis to build on settled conceptual foundations rather than improvised constructs.

The three settings—corporate management, tourism, and the technology supply base—were chosen by theoretical rather than statistical sampling, that is, for the analytical contrast they provide. Corporate management is the general locus in which authority is delegated to agents and is therefore where intra-organizational redistribution is most visible. Tourism is an intermediation-intensive, experience-and-service sector in which global distribution platforms occupy core positions and destinations occupy peripheral ones, which makes core–periphery dynamics and value-capture asymmetries observable in a single value chain (Tussyadiah, 2020; Majid et al., 2023). The technology supply base is the set of actors that build the models, orchestration layers, and reference architectures and thereby shape the field itself. The three were not selected to represent the economy but to expose, respectively, the firm-level, global-systemic, and field-level mechanisms the framework predicts.

The analytical procedure had four steps. First, the phenomenon and its diffusion were characterized from current technical and analyst literature (Wang et al., 2024; Gartner, 2024). Second, the phenomenon was read through each framework to derive analytical claims about capital redistribution, global asymmetry, and institutional convergence. Third, these claims were consolidated into propositions stated at the firm, sectoral, and global levels. Fourth, the design implications of the propositions were synthesized into a governance model. Throughout, secondary peer-reviewed and analyst sources were used to ground each claim and to discipline interpretation against the existing record on AI, labor, governance, and tourism (Acemoglu & Restrepo, 2020; Acemoglu et al., 2022; Jobin et al., 2019; Floridi, 2013).

Two scope conditions bound the argument. First, the empirical referent is the enterprise context of 2024 and 2025; claims about a phenomenon at this stage of diffusion are provisional and time-indexed. Second, the propositions are interpretive and conceptual. They are offered as testable conjectures, not as established findings, and the analysis makes no claim to quantify the magnitude of the effects it describes.

5. The Reorganization of Power Within the Firm

When a firm delegates action to agents, the practical question of who configures, supervises, and answers for those agents becomes a question of organizational power. Authority migrates toward the actors who control the data pipelines, the orchestration layer, and the governance apparatus, because these are the points at which an agent’s behavior is determined and defended. Budget and decision rights accrue to leaders who can demonstrate both measurable returns and credible control, while functions whose status rested on routine execution—manual reconciliation, sequential sign-off—lose ground. In Bourdieu’s terms, technical and governance competence is converted into cultural and symbolic capital, and the conversion rate favors hybrid profiles that combine domain knowledge with fluency in agent safety and oversight (Bourdieu, 1986).

Proposition 1. The adoption of autonomous agents relocates authority toward actors who control the data, orchestration, and governance of those agents, converting technical and governance competence into cultural and symbolic capital and displacing functions built on routine execution.

This redistribution is institutionalized in a recognizable set of practices: a policy layer that specifies permitted actions and escalation thresholds; an assurance layer of red-teaming and scenario testing; a telemetry layer of logs, rationales, and uncertainty measures; and oversight bodies with the authority to halt deployment. Within this emerging space, firms compete not only on what their agents can do but on their capacity to prove that the agents are under control. Demonstrable control—governability—becomes a stake in its own right and a resource that can be accumulated and displayed.

Proposition 2. Agentic AI gives rise to a distinct organizational field of agent governance, with its own stakes and forms of capital, in which the demonstrable controllability of agents (governability) becomes a competitive resource alongside capability.

The habitus of managers and staff shifts accordingly. Supervising software actors—reading dashboards, interpreting confidence scores, setting escalation rules—comes to feel as ordinary as reading a financial statement, while frontline work moves from execution toward the handling of exceptions and the refinement of policy. Human labor becomes more deliberative and synthetic and less repetitive, a recomposition consistent with evidence that AI reshapes the task content of work rather than simply eliminating jobs (Acemoglu et al., 2022; March, 1991; March & Simon, 1958).

Proposition 3. The supervision of autonomous agents reconfigures the managerial habitus from the supervision of people toward the supervision of software, shifting human labor from execution toward oversight, exception handling, and policy specification.

6. Sectoral Dynamics: Management and Tourism

6.1 Management and operations

In general management, agents compress the interval between a signal—a shift in demand, a supply disruption—and a response, and they reduce coordination latency across procurement, logistics, and service. Yet the language of efficiency can obscure a struggle over symbolic capital: which unit owns a decision, and whose metrics define success. When agent dashboards privilege a single class of indicator, typically cost, they can function as instruments of symbolic domination, naturalizing one coalition’s priorities as the firm’s objective. The analytical implication is that governance should distinguish agent-authorizable from human-reserved decisions, maintain a plurality of indicators across financial, ethical, and service dimensions, and submit the metric set itself to periodic review so that no coalition captures the definition of success (Floridi, 2013).

6.2 Tourism and hospitality

Tourism makes the stakes unusually legible because its value chain spans a global core of distribution platforms and a periphery of destinations, operators, and residents. Agents orchestrate dynamic packaging, personalized itineraries, real-time service recovery, and revenue management on the demand side, and scheduling and resource allocation on the supply side (Tussyadiah, 2020; Majid et al., 2023). Read through world-systems analysis, the platforms that mediate distribution occupy core positions and capture intermediary rents, while peripheral destinations accept default rules that shape their visibility and pricing power and absorb the social and environmental costs of the visits these systems route to them (Srnicek, 2016; Couldry & Mejias, 2019).

Proposition 4. In intermediation-intensive sectors such as tourism, agent-mediated distribution concentrates value capture in platform cores while externalizing social and environmental costs to peripheral destinations, unless agents are configured to encode local constraints.

The same lens identifies the conditions under which the dynamic can be altered. Destination management organizations and operator associations can pool anonymized demand signals into regional data cooperatives and host local agent stacks that negotiate terms with core platforms and embed local norms—sustainability caps on fragile sites, local-vendor quotas, and resident quality-of-life constraints—directly into the objective functions of the agents that allocate demand. Where cultural capital in the form of local knowledge, language, and storytelling is translated into agent-readable constraints, destinations can convert it into symbolic value—experiences perceived as authentic—rather than surrendering identity to generic global templates (Bourdieu, 1986). The principal risk to manage is homogenization: the algorithmic flattening of distinct places into interchangeable inventory.

6.3 The technology supply base

Technology suppliers concentrate economic capital in compute, data, and orchestration, and they accumulate symbolic capital by authoring the reference architectures that define what a safe agent looks like. Institutional pressures amplify this influence: as regulators, consultants, and professional bodies circulate the same templates, vendor designs become de facto standards (DiMaggio & Powell, 1983; Jobin et al., 2019). For buyers, the analytical implication is to treat governance capability as something to retain rather than fully externalize—favoring interoperable and open interfaces, maintaining second-source options to limit lock-in, and developing internal competence in agent-policy design and evaluation so that the capacity to govern does not migrate entirely to the supplier.

7. The Global Political Economy of Agentic AI

Three levers determine advantage in the agentic economy: access to high-quality and lawful data, including operational telemetry; the compute required to train, adapt, and run agents at scale; and control over the standards that govern formats, safety taxonomies, and audit expectations. Core actors tend to hold all three, which places peripheral organizations in a position of data and standards dependency, where the cost of deviating from core templates is high (Wallerstein, 2004; Crawford, 2021). This is the mechanism by which agentic AI can reproduce, rather than merely reflect, global asymmetry: the more that planning, action, and assurance are anchored in core infrastructures, the more peripheral actors operate through systems they do not control.

Proposition 5. Because foundational models, compute, and standards are concentrated among core actors, agentic AI tends to reproduce core–periphery asymmetries, positioning peripheral organizations as rent-paying users of core infrastructures unless countervailing arrangements alter their bargaining power.

The periphery is not, however, without recourse. Public investment in compute and data infrastructure, procurement policies that require interoperability and portability, and semi-peripheral coalitions of universities, public laboratories, and industry consortia can co-develop contextual agents anchored in local languages and regulatory traditions (Mazzucato, 2013). Tourism again supplies a concrete illustration: itinerary and ranking agents that systematically favor high-rent segments would deepen value-capture asymmetry, whereas policy-aligned agents that encode sustainability constraints and redirect demand toward under-represented communities could make the same technology an instrument of more equitable distribution. The broader point, in Polanyi’s sense, is that agentic markets remain embedded in social and political arrangements that can be designed to constrain extraction rather than accelerate it (Polanyi, 1944).

8. Institutional Convergence and Governance Design

Across these levels, organizations facing similar pressures arrive at similar arrangements. Regulation and dependency exert coercive pressure toward auditability, risk classification, human override, and traceability, which pushes firms toward common logging schemas and lifecycle controls. Uncertainty about returns and safety encourages mimetic imitation of the architectures of perceived leaders—typically a policy–reasoner–executor stack with safety filters and tiered autonomy. Professional communities in risk, law, and machine-learning safety exert normative pressure through standards, handbooks, and certifications, forming a shared professional vocabulary. The convergence on AI ethics principles documented by Jobin et al. (2019), accompanied by divergence in implementation, is the expected signature of these mechanisms operating in a field whose templates are not yet settled.

Proposition 6. Coercive, mimetic, and normative pressures drive organizations toward convergent agent architectures and governance templates, reducing variance in how autonomy is governed and making conformity to emerging templates a condition of legitimacy.

Convergence has a cost. Mimetic imitation that ignores context can flatten local distinctiveness, which is especially damaging in tourism and public services, and dependence on core templates can erode the internal capacity to evaluate and adapt them. A governance model is therefore useful not as a compliance checklist but as a means of retaining deliberate control while meeting legitimate external expectations. The seven layers in Table 1 organize the decisions that delegation to agents requires, from the definition of purpose to the management of learning.

Table 1. A seven-layer model for governing autonomous AI agents.

Layer	Function	Representative elements
Purpose and scope	Establish why agents are used and where they are barred	Success metrics; out-of-scope decisions such as high-stakes employment actions
Policy and constraints	Define the boundaries of autonomous action	Permitted actions; data access; rate, cost, and escalation thresholds
Safety and assurance	Test behavior before and during deployment	Red-teaming; adversarial simulation; robustness checks under distribution shift
Observability	Make agent behavior legible and auditable	Standardized logs of prompts, tool calls, and outputs; rationales; uncertainty estimates
Control and intervention	Keep humans able to direct and stop agents	Graded autonomy (observe, propose, execute); checkpoints; emergency stop
Accountability	Assign responsibility for agent decisions	Role mapping for approval, sign-off, and incident reporting; periodic board reporting
Learning and adaptation	Improve agents under controlled change	Feedback from human overrides; post-incident review; change-controlled retraining

Note. The layers are cumulative rather than sequential; each remains active throughout the agent lifecycle. The model is offered as an analytical organization of governance decisions, not as a maturity standard or compliance instrument.

Read against the propositions, the model also clarifies the stakes of the field analyzed in Sections 5 and 8: the observability, control, and accountability layers are precisely where governability is produced and displayed, which is why competence in them functions as capital, and why their externalization to suppliers transfers power as well as work.

9. Discussion

Taken together, the seven propositions support a final claim that integrates the three lenses. As agents act on behalf of organizations, external audiences—regulators, partners, boards, and the public—assess deployments less by raw capability than by visible evidence of responsible control. Audits, certifications, and governance disclosures become the symbolic markers through which legitimacy is granted, with the consequence that symbolic capital can become as decisive as technical performance in securing contracts and trust (Bourdieu, 1986; Jobin et al., 2019; Zuboff, 2019).

Proposition 7. The legitimacy of agentic AI deployment increasingly depends on symbolic markers of responsible control, such that symbolic capital becomes a competitive resource alongside, and at times ahead of, technical performance.

The contribution to each body of theory can now be stated precisely. To Bourdieusian organizational sociology, the article extends field and capital theory to the supervision of autonomous software and identifies governability as an emergent species of capital, showing that AI functions as a stake in field struggles rather than as an inert capability that firms simply hold. To the political economy of AI, it specifies the mechanisms—model control, compute concentration, and standard-setting—through which agentic systems can reproduce core–periphery asymmetry, connecting the organizational analysis to accounts of data colonialism and platform capitalism that have so far operated at the level of infrastructure rather than the firm (Couldry & Mejias, 2019; Srnicek, 2016; Crawford, 2021). To institutional theory, it applies the isomorphism framework to a technology whose governance templates are still forming, characterizing agent governance as a nascent field in which convergence pressures operate ahead of settled standards and in which the documented convergence-with-divergence pattern in AI ethics is the visible trace (DiMaggio & Powell, 1983; Jobin et al., 2019).

The argument also speaks to a live debate about automation and work. Where the displacement literature measures the substitution of labor by machines (Acemoglu & Restrepo, 2020; Acemoglu et al., 2022), the present account reframes the question as one of authority: the central change is not only how many tasks are automated but how decision rights are redistributed among human and non-human actors and which actors accumulate the capital to govern them. For management and tourism scholarship, the practical reframing is that the strategic question is not whether to adopt agents but how to do so in ways that preserve plural objectives, local distinctiveness, and a fair position within the global division of computational labor.

10. Limitations and Future Research

The study is conceptual, and its claims carry the limitations of conceptual work. The seven propositions are interpretive conjectures grounded in established theory and secondary evidence rather than findings from primary data, and the analysis does not estimate the magnitude or scope of the effects it describes. Its empirical referent is the enterprise context of 2024 and 2025; because the phenomenon is at an early stage of diffusion, the patterns identified are time-indexed and may shift as the technology and its regulation mature. The three settings were chosen for analytical contrast rather than representativeness, so the framework should be transferred to other sectors with care.

These limitations define a research agenda. Comparative field studies could test Propositions 1 through 3 and 6 by examining how agent-governance fields and managerial habitus vary across sectors and regulatory regimes. Longitudinal labor research could trace the recomposition of work that Proposition 3 anticipates and identify the new forms of skill, precarity, or empowerment that follow. Cross-regional political-economy research could assess Propositions 4 and 5 by examining the conditions under which semi-peripheral coalitions, public compute, and regional data cooperatives shift bargaining power. Across all of these, the construct of governability proposed here—and its measurement as a form of capital—offers a tractable target for operationalization and testing.

11. Conclusion

Autonomous AI agents are not only a technical advance but an organizational and geopolitical force that reorders who holds authority, how value is captured, and which actors can credibly claim legitimacy. Reading their rise through Bourdieu, world-systems analysis, and institutional isomorphism shows a redistribution of capital and the formation of a new field within firms, the reproduction of core–periphery asymmetries across them, and a convergence on governance templates that makes demonstrable control a condition of legitimacy. The article’s contribution is to integrate these levels into a single framework, to identify governability as an emergent form of capital, and to translate the analysis into a governance model and a set of testable propositions. The strategic and scholarly task that follows is to govern for difference: to build interoperable, audited, and human-aligned agent systems that honor local context while meeting the legitimate demands of a global field.

References

• Acemoglu, D., Autor, D., Hazell, J., & Restrepo, P. (2022). Artificial intelligence and jobs: Evidence from online vacancies. Journal of Labor Economics, 40(S1), S293–S340. https://doi.org/10.1086/718327

• Acemoglu, D., & Restrepo, P. (2020). Robots and jobs: Evidence from US labor markets. Journal of Political Economy, 128(6), 2188–2244. https://doi.org/10.1086/705716

• Bourdieu, P. (1977). Outline of a theory of practice. Cambridge University Press.

• Bourdieu, P. (1986). The forms of capital. In J. G. Richardson (Ed.), Handbook of theory and research for the sociology of education (pp. 241–258). Greenwood Press.

• Bourdieu, P. (1990). The logic of practice. Stanford University Press.

• Brynjolfsson, E., & McAfee, A. (2014). The second machine age: Work, progress, and prosperity in a time of brilliant technologies. W. W. Norton.

• Castells, M. (1996). The rise of the network society. Blackwell.

• Couldry, N., & Mejias, U. A. (2019). Data colonialism: Rethinking big data's relation to the contemporary subject. Television & New Media, 20(4), 336–349. https://doi.org/10.1177/1527476418796632

• Crawford, K. (2021). Atlas of AI: Power, politics, and the planetary costs of artificial intelligence. Yale University Press.

• Davenport, T. H., & Ronanki, R. (2018). Artificial intelligence for the real world. Harvard Business Review, 96(1), 108–116.

• DiMaggio, P. J., & Powell, W. W. (1983). The iron cage revisited: Institutional isomorphism and collective rationality in organizational fields. American Sociological Review, 48(2), 147–160. https://doi.org/10.2307/2095101

• Floridi, L. (2013). The ethics of information. Oxford University Press.

• Gartner. (2024). Top strategic technology trends for 2025: Agentic AI. Gartner, Inc.

• Jobin, A., Ienca, M., & Vayena, E. (2019). The global landscape of AI ethics guidelines. Nature Machine Intelligence, 1(9), 389–399. https://doi.org/10.1038/s42256-019-0088-2

• Katsamakas, E., Pavlov, O. V., & Saklad, R. (2024). Artificial intelligence and the transformation of higher education institutions: A systems approach. Sustainability, 16(14), 6118. https://doi.org/10.3390/su16146118

• Latour, B. (2005). Reassembling the social: An introduction to actor-network theory. Oxford University Press.

• Majid, G. M., Tussyadiah, I., Kim, Y. R., & Pal, A. (2023). Intelligent automation for sustainable tourism: A systematic review. Journal of Sustainable Tourism, 31(11), 2421–2440. https://doi.org/10.1080/09669582.2023.2246681

• March, J. G. (1991). Exploration and exploitation in organizational learning. Organization Science, 2(1), 71–87. https://doi.org/10.1287/orsc.2.1.71

• March, J. G., & Simon, H. A. (1958). Organizations. Wiley.

• Mazzucato, M. (2013). The entrepreneurial state: Debunking public vs. private sector myths. Anthem Press.

• Polanyi, K. (1944). The great transformation: The political and economic origins of our time. Beacon Press.

• Srnicek, N. (2016). Platform capitalism. Polity Press.

• Tussyadiah, I. (2020). A review of research into automation in tourism: Launching the Annals of Tourism Research curated collection on artificial intelligence and robotics in tourism. Annals of Tourism Research, 81, 102883. https://doi.org/10.1016/j.annals.2020.102883

• Wallerstein, I. (2004). World-systems analysis: An introduction. Duke University Press.

• Wang, L., Ma, C., Feng, X., Zhang, Z., Yang, H., Zhang, J., Chen, Z., Tang, J., Chen, X., Lin, Y., Zhao, W. X., Wei, Z., & Wen, J.-R. (2024). A survey on large language model based autonomous agents. Frontiers of Computer Science, 18(6), 186345. https://doi.org/10.1007/s11704-024-40231-1

• Zuboff, S. (2019). The age of surveillance capitalism: The fight for a human future at the new frontier of power. PublicAffairs.

#AgenticAI #AutonomousAgents #ArtificialIntelligence #OrganizationalPower #CriticalSociology #Bourdieu #WorldSystemsTheory #InstitutionalIsomorphism #AIGovernance #PoliticalEconomyOfAI #FutureOfWork #DigitalTransformation #TourismTechnology #ManagementStudies #DataColonialism #PlatformCapitalism #ResponsibleAI