Agentic AI as a Strategic Capability in Service Economies: Evidence From Banking and Tourism

Hassan, Issa

doi:10.65326/u7y566743

Agentic AI as a Strategic Capability in Service Economies: Evidence From Banking and Tourism "Agentic Artificial Intelligence"

Oct 30, 2025
17 min read

Updated: 5 days ago

Author: Issa Hassan

ORCID iD: 0009-0001-4071-058X

Affiliation: Swiss International University (SIU)

Received 3 July 2025; Revised 27 August 2025; Accepted 5 September 2025; Available online 30 October 2025; Version of Record 30 October 2025.

DOI: https://doi.org/10.65326/u7y566743

Volume 2, December 2025, (10020)

Abstract

Agentic artificial intelligence—systems that can perceive context, reason with memory, call external tools, and act toward goals with varying degrees of autonomy—has rapidly moved from experimental demos to production roadmaps in service economies. This article reframes agentic AI not merely as a technological capability but as a field of power that redistributes capital (economic, social, cultural, and symbolic), reorganizes organizational isomorphism, and re-articulates core–periphery relations in global markets. Building on Bourdieu’s theory of capital and fields, world-systems analysis, and institutional isomorphism, I analyze how agentic AI reconfigures decision rights, risk, and value capture in two emblematic service sectors: banking and tourism. I advance (1) a sociotechnical capability stack for agentic AI, (2) a governance and assurance framework oriented to procedural justice and fairness over time, and (3) a mixed-methods research agenda capable of isolating productivity, quality, and equity effects. The contribution is a critical yet constructive account that treats agentic AI as both organizational technology and social institution, offering executives, regulators, and scholars a vocabulary and blueprint for responsible adoption.

Keywords: agentic AI, service economy, banking technology, travel and tourism, Bourdieu, world-systems, institutional isomorphism, governance, fairness, organizational learning

1. Introduction

Service economies coordinate knowledge under uncertainty, from retail banking to destination management. For two decades, the automation frontier in these sectors was defined by rule engines and predictive analytics; generative models then widened it by turning unstructured language and images into operational signals (Brynjolfsson & McAfee, 2014; Huang & Rust, 2018). Agentic artificial intelligence (AI) marks a further shift: systems now link perception, reasoning, and action, planning tasks, orchestrating external tools such as pricing engines and booking systems, critiquing their own output, and escalating to humans under explicit uncertainty thresholds (Vanneste & Puranam, 2024). The defining feature is not better prediction but delegated initiative—the capacity to take consequential action with bounded autonomy.

The managerial promise is well rehearsed: shorter queues, faster approvals, personalized itineraries, and fewer operational backlogs. I argue that the stakes are more structural than this framing suggests. Agentic AI reconfigures who holds which forms of capital, how organizations come to resemble one another under institutional pressure, and how value and risk travel across the core–periphery geography of the world economy. The salient question is not only what can be automated but who becomes legitimate to decide, supervise, audit, and profit (Kellogg, Valentine, & Christin, 2020; Zuboff, 2019).

1.1 Research gap

Three literatures speak to this moment but do not yet meet. First, organizational research on AI has clarified the automation–augmentation tension and the conditions under which human–AI combinations add value, but it largely treats AI as a decision aid rather than an acting agent embedded in a field of competing interests (Raisch & Krakowski, 2021; Vaccaro, Almaatouq, & Malone, 2024; Bankins, Ocampo, Marrone, Restubog, & Woo, 2024). Second, sector studies of AI in banking and in tourism document adoption patterns, efficiency gains, and customer responses, yet remain descriptive about the social and institutional restructuring that delegated autonomy sets in motion (Singh, Mishra, Kumar, & Bag, 2025; Knani, Echchakoui, & Ladhari, 2022; Doborjeh, Hemmington, Doborjeh, & Kasabov, 2022). Third, work on algorithmic fairness and AI governance offers principles and audit techniques, but rarely connects them to the strategic logic by which firms convert legitimacy into economic advantage (Selbst, Boyd, Friedler, Venkatasubramanian, & Vertesi, 2019; Mittelstadt, 2019; Jobin, Ienca, & Vayena, 2019). The gap is therefore conceptual: there is no integrated account that explains how agentic AI simultaneously redistributes organizational capital, propagates institutional isomorphism, and re-articulates core–periphery dependence in service markets. Absent such an account, executives optimize for efficiency while underestimating the legitimacy, labor, and value-capture dynamics that determine whether adoption is durable.

1.2 Research questions and approach

This article addresses four questions. (a) What sociotechnical capability stack is necessary for responsible agentic AI in services? (b) How does agentic AI redistribute forms of capital across workers, firms, and customers? (c) How do institutional and world-system pressures shape adoption trajectories in banking and tourism? (d) What measurement strategies can separate genuine productivity gains from quality, fairness, and legitimacy effects? I pursue these through a theory-building synthesis that integrates Bourdieu’s theory of capital and fields, world-systems analysis, and institutional isomorphism, and applies the integrated lens to two analytically contrasting service sectors. The contribution is threefold: a capability stack for agentic AI, a governance and assurance framework centered on procedural justice and temporal fairness, and a set of theoretical propositions with an accompanying empirical agenda.

2. Theoretical Background

2.1 Capital, field, and habitus

Bourdieu (1986) holds that actors compete within fields for position using convertible forms of capital: economic (resources), cultural (credentials and know-how), social (networks), and symbolic (recognized legitimacy). Agentic AI enters organizations as objectified cultural capital—best practices codified in prompts, policies, and playbooks—and as symbolic capital, a signal of competence and modernity. Its deployment can elevate technical and risk teams who curate tools, thresholds, and logs while devaluing routine clerical roles whose tacit practice becomes embedded in agentic workflows (Kellogg et al., 2020; Chen & Chan, 2024). Because capital is convertible, early adopters can transmute symbolic capital into economic capital—market share, revenue per customer—and back into recruitment prestige and partnerships. The frontier of advantage is thus less model accuracy than the conversion rate among capitals: how know-how and legitimacy crystallize into revenue and regulatory latitude, a dynamic that resonates with the microfoundations of dynamic capabilities (Teece, 2007).

2.2 World-systems and the AI value chain

World-systems analysis foregrounds structural inequality in global production networks (Wallerstein, 1974). Agentic AI ecosystems instantiate a new core in model and infrastructure provision, while many service firms—especially in the periphery and semi-periphery—consume models and tools with limited bargaining power. Customer conversations, documents, and itineraries flow toward core infrastructure, where value capture concentrates (Zuboff, 2019). Tourism, frequently situated in peripheral and seasonal economies, risks becoming a raw-data exporter paying rents to core platform providers; banking in emerging markets may depend on imported risk models and guardrails, reshaping exposure to regulatory sovereignty. The strategic implication for peripheral contexts is to pursue data localization, shared sectoral utilities, and negotiated standards that retain a fair share of value.

2.3 Institutional isomorphism

DiMaggio and Powell (1983) describe how organizations converge in structure under coercive, mimetic, and normative pressures. Agentic AI accelerates this convergence: policy engines, audit logs, and human-in-the-loop checkpoints harden into standardized expectations, while vendor blueprints and regulatory consultation papers codify what good practice looks like (Jobin et al., 2019). Mimetic pressure is acute in banking, where firms fear lagging on cost-to-income ratios, and in tourism, where firms fear missing personalization. Normative pressure grows as risk, audit, and data professionals articulate codes of practice and certifications. Isomorphism can raise a safety baseline, but it may also dull experimentation and elevate core-economy practices to the status of the universal, crowding out local knowledge.

2.4 An integrative lens: agentic AI as a field of power

Read together, the three traditions describe one phenomenon at three levels. Bourdieu specifies the intra-organizational struggle over who controls and benefits from agentic capability; institutional theory specifies the inter-organizational convergence that standardizes how that capability is governed; world-systems analysis specifies the global stratification that determines where value is captured. Treating agentic AI as a field of power, rather than a neutral toolkit, makes these levels analytically commensurable and motivates the case analysis that follows.

3. Research Design

This is a conceptual, theory-building article rather than an empirical study, and its claims are interpretive. The design follows three steps. First, I conducted an integrative reading of the literatures identified in Section 1.1—organizational AI, sector studies of banking and tourism, and algorithmic fairness and governance—selecting sources for their theoretical leverage on capital, institutions, and global stratification rather than for exhaustive coverage. Second, I applied the integrated lens to two analytical cases. Third, I derived a set of falsifiable propositions and an associated empirical agenda (Sections 7 and 10).

The case selection is purposive and contrastive. Banking and tourism are both knowledge-coordinating service sectors in which agentic AI is moving into production, which makes them comparable; they differ on dimensions that sharpen theory. Banking is a high-regulation, high-stakes domain where consequential decisions (credit, anti-money-laundering, collections) are bound by formal accountability and where legitimacy turns on visible due process. Tourism is a lighter-regulation, demand-volatile domain where value hinges on perceived fairness, experiential quality, and seasonal capacity, and where peripheral operators are especially exposed to dependence on imported infrastructure. The contrast lets the same theoretical mechanisms be observed under different institutional and world-system conditions, supporting analytical rather than statistical generalization.

The analytical procedure maps each case onto the three theoretical levels: I identify the decision archetypes where agents act, trace how each archetype redistributes forms of capital, and locate the isomorphic and core–periphery pressures that shape adoption. Because the work is sociomaterial—capability is constituted in the entanglement of people, policies, and tools rather than in the model alone—the unit of analysis is the governed decision workflow, not the model (Orlikowski, 2007). The scope is deliberately bounded: I treat service decisions in which an agent can take or recommend consequential action under human oversight, and I exclude fully autonomous, safety-critical control as well as consumer-facing entertainment uses. The propositions are offered as theoretically grounded conjectures to be tested, not as validated findings.

4. Agentic AI as a Sociotechnical Capability

4.1 A capability stack

Responsible agentic AI rests on six interdependent layers. Data foundations provide governed, lineage-aware access with privacy by design. The reasoning layer combines frontier language models with task-specific components for retrieval, planning, and self-critique. Tooling and orchestration supply secure tool catalogs—payments, customer relationship management, revenue management, booking—together with workflow control over cost and latency. Safety and governance add policy filters, thresholding, redaction, immutable logs, and appeal pathways. Role design coordinates specialized agents (planner, analyst, critic, compliance, executor) through shared memory and arbitration. Finally, the experience layer equips supervisors with interfaces to explain, approve, and amend agent actions. The stack reframes agentic AI as a dynamic capability whose value lies in orchestration and recombination rather than in any single component (Teece, 2007).

4.2 Capability envelopes and progressive autonomy

Agents should operate within explicit capability envelopes that enumerate actions permitted without approval, permitted under conditional approval, and prohibited. Progressive autonomy then moves from advisory output, to constrained actions with automatic rollback, to conditional autonomy under continuous performance and drift monitoring. The envelope is itself a site of symbolic struggle: which function—compliance, operations, or marketing—wins the right to set thresholds is a question of power in the field, and it determines whether augmentation or automation dominates in practice (Raisch & Krakowski, 2021).

4.3 Instrumentation for learning

To avoid mistaking macro- or selection-driven gains for agent effects, organizations need counterfactuals captured during shadow operation: a record of what a trained human would have done in parallel. Such instrumentation converts tacit know-how into objectified cultural capital—playbooks, prompts, and red-team cases—preserving institutional memory as roles shift, and it operationalizes the exploration–exploitation balance that sustains learning over time (March, 1991; Vaccaro et al., 2024).

5. Banking: Compliant Personalization as Field Reconfiguration

5.1 Decision archetypes and agent roles

Four archetypes are illustrative. In onboarding and know-your-customer triage, planner agents extract and validate documents while compliance agents enforce policy and escalate anomalies. In small-business credit renewal, analyst agents reconcile financial statements with transaction graphs, risk tools compute exposure, compliance agents draft disclosures, and humans approve. In collections and care, negotiation agents propose hardship plans while fairness monitors enforce offer parity across comparable borrowers. In fraud and anti-money-laundering investigations, multi-agent teams cross-reference alerts, narrative summaries, and network graphs under auditable traces. Surveys of AI in banking confirm that these are precisely the functions where adoption is concentrating, alongside robo-advisory and personalized service (Singh et al., 2025; Barile, Secundo, & Bussoli, 2024). Each archetype embeds cultural capital (risk knowledge) in policy and prompts, augments social capital (relationship networks) through customer-facing agents, and stages symbolic capital (soundness) through transparent rationales.

5.2 Redistribution of capital and labor

Agentic workflows decompose once-holistic banker tasks into supervision plus exception handling. Mid-career analysts who curate prompts, critique rationales, and sign off on drift become pivotal, which elevates the cultural capital of those adept at scrutinizing machine reasoning while devaluing routine documentation (Kellogg et al., 2020; Chen & Chan, 2024). The redistribution is not automatic or benign: if design excludes frontline staff, tacit knowledge of local customers—a form of social capital—can be erased, yielding brittle decisions and reputational loss. Whether the net effect is augmentation or hollowing-out depends on how envelopes and oversight roles are designed (Raisch & Krakowski, 2021).

5.3 Fairness and procedural justice

Fairness in banking must be a temporal commitment rather than a one-time report. Equalized error rates, false-positive differentials, and denial explanations require monitoring over months, because both data distributions and agent behavior drift (Selbst et al., 2019). Equally, procedural justice—clear reasons, accessible appeals, and timely remediation—matters as much as statistical parity, since legitimacy hinges on visible due process and on customer trust in the system’s reasoning (Glikson & Woolley, 2020; Pasquale, 2015). Explanations should therefore reference policy and data lineage, not merely model internals. Randomized branch-level rollouts and difference-in-differences against matched cohorts can separate agent effects from macroeconomic shifts, converting credibility into durable bargaining power with regulators.

6. Tourism and Hospitality: Adaptive Sensing and Experience

6.1 Event-driven sensing and pricing

Tourism thrives on volatile demand. Sensing agents read event calendars and transport capacity; pricing agents propose rate and inventory changes; experience agents assemble packages of transfers, tours, and food and beverage; and operations agents adjust staffing—while supervisors enforce caps and fairness norms. Reviews of AI in hospitality and tourism map a comparable migration from descriptive analytics to service robots and conversational agents that act on the guest journey (Knani et al., 2022; Doborjeh et al., 2022; Kim, So, & Wirtz, 2022).

6.2 Perceived fairness and symbolic capital

Pricing power is bounded by social meaning. Even when a revenue model is technically correct, customers may read sharp increases as opportunistic, and symbolic capital—brand warmth—erodes when communications lack reasons (Glikson & Woolley, 2020). Agents should therefore generate explanations (a city-wide conference, limited inventory, an honored loyalty guarantee) and offer goodwill gestures when thresholds are crossed. Because hospitality is co-produced, agentic messaging must preserve an authentic human rescue path so that automation does not become a hollow performance of service.

6.3 Core–periphery dependence

Destination operators in peripheral economies often depend on imported agent stacks and remote data centers. Without local capacity they export behavioral data while importing pricing logic, concentrating value capture in core providers (Wallerstein, 1974; Zuboff, 2019). A counter-strategy is cooperative infrastructure: regional alliances that pool data under shared governance, train sector-specific retrieval corpora, and negotiate platform terms, thereby reclaiming a share of economic and symbolic capital.

7. Cross-Case Synthesis and Propositions

Across both sectors, the same mechanisms recur under different institutional and world-system conditions. I state them as falsifiable propositions; each is interpretive at present and is paired with a testable empirical claim in Section 10.

Proposition 1 (Capital conversion). The strategic advantage organizations derive from agentic AI depends less on model accuracy than on their capacity to convert cultural and symbolic capital—codified know-how and recognized legitimacy—into economic capital; firms with stronger conversion capability capture disproportionate value.

Proposition 2 (Intra-firm redistribution). Agentic workflows decompose holistic service roles into supervision and exception handling, elevating staff who curate and audit agent reasoning and devaluing routine documentation, thereby redistributing cultural capital within the firm.

Proposition 3 (Procedural and temporal legitimacy). In consequential service decisions, perceived legitimacy depends more on procedural justice and on fairness sustained over time than on aggregate efficiency or single-point statistical parity; because data and behavior drift, point-in-time fairness overstates equity.

Proposition 4 (Isomorphic convergence). Coercive, mimetic, and normative pressures drive convergence on standardized agentic-governance templates, raising a safety baseline while narrowing local experimentation and privileging core-economy practices.

Proposition 5 (Core–periphery value capture). Where peripheral service firms adopt imported agent stacks without local data and inference capacity, they export behavioral data and import decision logic, concentrating value capture in core providers; cooperative regional infrastructure attenuates this effect.

Proposition 6 (Counterfactual instrumentation). Without parallel counterfactual capture, organizations will misattribute macro- or selection-driven gains to agentic AI; credible causal estimation is a precondition for durable strategic claims.

Proposition	Core claim	Primary theoretical anchor	Illustrative case
P1	Advantage flows from capital conversion, not accuracy alone	Bourdieu; dynamic capabilities	Both
P2	Roles split into supervision and exception handling; capital redistributes	Bourdieu; algorithmic control	Banking
P3	Legitimacy rests on procedural and temporal fairness	Procedural justice; fair-ML critique	Both
P4	Governance templates converge, narrowing experimentation	Institutional isomorphism	Both
P5	Imported stacks concentrate value capture in the core	World-systems analysis	Tourism
P6	Counterfactuals are required to attribute gains credibly	Organizational learning	Both

Note. Each proposition is a theoretically grounded conjecture intended for empirical testing; “Both” indicates a mechanism observed in banking and tourism, and the listed case denotes where it is most pronounced.

8. Governance and Assurance

Governance for agentic AI combines ex ante and ex post controls. Ex ante, capability envelopes enumerate permitted actions and thresholds per agent, two-person rules guard consequential moves such as limit changes or high-impact pricing, policy engines codify suitability and data minimization, and scenario libraries stress-test agents against adversarial prompts, tool abuse, and fairness-critical situations such as surge pricing during emergencies. Ex post, immutable and replayable traces support audit, incident response, and appeals; counterfactual explanations clarify what would have happened under alternative policies or data; and drift, cost, and human-override metrics feed back into the envelopes. Layered auditing of the underlying models—governance, process, and output—gives these controls technical grounding (Mökander, Schuett, Kirk, & Floridi, 2023).

Principles such as beneficence and justice gain force only when attached to practices: redaction by default, opt-in personalization, tiered explanations for customers and auditors, and a refusal of unbounded autonomy in financially or emotionally consequential contexts (Mittelstadt, 2019). Because converging governance templates can ossify into mere compliance, assurance should be designed to preserve, not foreclose, local experimentation (DiMaggio & Powell, 1983; Jobin et al., 2019).

9. Discussion

The central argument is that agentic AI succeeds when organizations earn legitimacy in the eyes of customers, workers, and regulators; efficiency is necessary but not sufficient. Deployments that maximize short-term metrics while minimizing due process invite backlash, regulatory friction, and the erosion of brand meaning, whereas those that make reasons visible accumulate the symbolic capital that stabilizes adoption (Glikson & Woolley, 2020; Pasquale, 2015).

The study makes three theoretical contributions. To Bourdieusian field theory, it specifies a concrete contemporary site—the capability envelope and the supervision role—where the struggle over capital and its conversion is contested, extending the theory of capital to a sociomaterial setting in which know-how is objectified in prompts and logs (Bourdieu, 1986; Orlikowski, 2007). To institutional theory, it shows how vendor blueprints and regulatory templates function as carriers of isomorphism specific to autonomous systems, and it identifies the trade-off between a rising safety baseline and narrowing local experimentation (DiMaggio & Powell, 1983; Jobin et al., 2019). To world-systems analysis, it reframes data and inference dependence as a mechanism of core–periphery value capture in services, and it names cooperative infrastructure as a counter-strategy (Wallerstein, 1974; Zuboff, 2019).

The account also speaks to live debates in organizational AI. The automation–augmentation paradox is reinterpreted here as a contest over who sets capability envelopes, locating the paradox in the politics of threshold-setting rather than in technology alone (Raisch & Krakowski, 2021). The literature on human–AI combinations is extended from the question of whether such combinations help, on average, to the institutional conditions under which they remain legitimate and fair over time (Vaccaro et al., 2024; Choudhary, Marchetti, Shrestha, & Puranam, 2025). Finally, by distinguishing measured efficiency from credibly attributed effects, the analysis connects to macroeconomic caution about overstated AI productivity, underscoring that strategic claims require counterfactual evidence (Acemoglu, 2025).

Strategic implications follow the world-system position of the firm. Core economies should invest in procedural legitimacy, interoperable logs, and audit interfaces, while guarding against the complacency that isomorphic comfort breeds. Semi-peripheral firms should dual-source models, localize retrieval corpora, and build regional assurance services that monetize local language and policy nuance. Peripheral firms should prioritize data sovereignty and cooperative infrastructure, negotiating region-bound inference and sharing audit artifacts across destination and banking networks to retain symbolic and social capital.

10. Limitations and Future Research

This article is conceptual, and its propositions are interpretive rather than tested. The two-case design supports analytical, not statistical, generalization, and the choice of banking and tourism, though theoretically motivated, leaves open how the mechanisms operate in sectors such as healthcare or public administration. The framework also assumes a level of governance maturity that many firms have not reached, so its prescriptions may apply unevenly. These limits define a clear agenda.

The propositions are designed to be testable. Productivity and attribution claims (P1, P6) can be examined with randomized branch- and property-level rollouts, stepped-wedge designs, and difference-in-differences against matched units, with causal mediation used to decompose gains into retrieval, planning, and tool-use components (Vaccaro et al., 2024; Acemoglu, 2025). Redistribution claims (P2) call for longitudinal study of supervision roles and for ethnography of supervisor–agent interaction, capturing how organizational dispositions meet agent affordances—who trusts, who resists, and why (Kellogg et al., 2020; Orlikowski, 2007). Legitimacy and fairness claims (P3) require pre-registered equity audits with thresholds and remediation plans, tracking error and recovery dispersion across segments and seasons in tourism and adverse-action reasons in banking, since fairness fluctuates with data mix (Selbst et al., 2019; Glikson & Woolley, 2020). Isomorphism and value-capture claims (P4, P5) invite comparative and cross-national designs that trace how governance templates diffuse and how data and inference dependence shape the distribution of returns (DiMaggio & Powell, 1983; Singh et al., 2025). Across these designs, the priority is to separate genuine quality and equity effects from efficiency gains rather than to assume they coincide.

11. Conclusion

Agentic AI in service economies is best understood as a sociotechnical institution that redistributes capital, standardizes governance, and reshapes global value chains. Banking shows how compliant personalization can compress decision cycles while demanding rigorous fairness over time; tourism shows how adaptive sensing can lift revenue while depending on trust-building explanations and community inclusion. Treated as a field of power rather than a toolkit, agentic AI invites strategies that convert cultural and symbolic capital into durable economic value without sacrificing justice or autonomy. The contribution of this article is an integrated lens and a set of propositions that turn that claim into a testable research program: define capability envelopes, specialize roles, instrument counterfactuals, and commit to longitudinal fairness, so that novelty becomes legitimacy and legitimacy becomes sustainable advantage.

References

Acemoglu, D. (2025). The simple macroeconomics of AI. Economic Policy, 40(121), 13–58. https://doi.org/10.1093/epolic/eiae042
Bankins, S., Ocampo, A. C., Marrone, M., Restubog, S. L. D., & Woo, S. E. (2024). A multilevel review of artificial intelligence in organizations: Implications for organizational behavior research and practice. Journal of Organizational Behavior, 45(2), 159–182. https://doi.org/10.1002/job.2735
Barile, D., Secundo, G., & Bussoli, C. (2024). Exploring artificial intelligence robo-advisor in the banking industry: A platform model. Management Decision. Advance online publication. https://doi.org/10.1108/MD-08-2023-1324
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.
Brynjolfsson, E., & McAfee, A. (2014). The second machine age: Work, progress, and prosperity in a time of brilliant technologies. W. W. Norton.
Chen, Z., & Chan, J. (2024). Large language model in creative work: The role of collaboration modality and user expertise. Management Science, 70(12), 9101–9117. https://doi.org/10.1287/mnsc.2023.03014
Choudhary, V., Marchetti, A., Shrestha, Y. R., & Puranam, P. (2025). Human–AI ensembles: When can they work? Journal of Management, 51(2), 536–569. https://doi.org/10.1177/01492063231194968
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
Doborjeh, Z., Hemmington, N., Doborjeh, M., & Kasabov, N. (2022). Artificial intelligence: A systematic review of methods and applications in hospitality and tourism. International Journal of Contemporary Hospitality Management, 34(3), 1154–1175. https://doi.org/10.1108/IJCHM-06-2021-0767
Glikson, E., & Woolley, A. W. (2020). Human trust in artificial intelligence: Review of empirical research. Academy of Management Annals, 14(2), 627–660. https://doi.org/10.5465/annals.2018.0057
Huang, M.-H., & Rust, R. T. (2018). Artificial intelligence in service. Journal of Service Research, 21(2), 155–172. https://doi.org/10.1177/1094670517752459
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
Kellogg, K. C., Valentine, M. A., & Christin, A. (2020). Algorithms at work: The new contested terrain of control. Academy of Management Annals, 14(1), 366–410. https://doi.org/10.5465/annals.2018.0174
Kim, H., So, K. K. F., & Wirtz, J. (2022). Service robots: Applying social exchange theory to better understand human–robot interactions. Tourism Management, 92, 104537. https://doi.org/10.1016/j.tourman.2022.104537
Knani, M., Echchakoui, S., & Ladhari, R. (2022). Artificial intelligence in tourism and hospitality: Bibliometric analysis and research agenda. International Journal of Hospitality Management, 107, 103317. https://doi.org/10.1016/j.ijhm.2022.103317
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
Mittelstadt, B. (2019). Principles alone cannot guarantee ethical AI. Nature Machine Intelligence, 1(11), 501–507. https://doi.org/10.1038/s42256-019-0114-4
Mökander, J., Schuett, J., Kirk, H. R., & Floridi, L. (2023). Auditing large language models: A three-layered approach. AI and Ethics. Advance online publication. https://doi.org/10.1007/s43681-023-00289-2
Orlikowski, W. J. (2007). Sociomaterial practices: Exploring technology at work. Organization Studies, 28(9), 1435–1448. https://doi.org/10.1177/0170840607081138
Pasquale, F. (2015). The black box society: The secret algorithms that control money and information. Harvard University Press.
Raisch, S., & Krakowski, S. (2021). Artificial intelligence and management: The automation–augmentation paradox. Academy of Management Review, 46(1), 192–210. https://doi.org/10.5465/amr.2018.0072
Selbst, A. D., Boyd, D., Friedler, S. A., Venkatasubramanian, S., & Vertesi, J. (2019). Fairness and abstraction in sociotechnical systems. In Proceedings of the Conference on Fairness, Accountability, and Transparency (pp. 59–68). Association for Computing Machinery. https://doi.org/10.1145/3287560.3287598
Singh, R. K., Mishra, R., Kumar, S., & Bag, S. (2025). Applications of artificial intelligence for optimizing banking and financial operations: A systematic literature review and future research agenda. Journal of Economic Surveys, 39(5), 2284–2302. https://doi.org/10.1111/joes.12693
Teece, D. J. (2007). Explicating dynamic capabilities: The nature and microfoundations of (sustainable) enterprise performance. Strategic Management Journal, 28(13), 1319–1350. https://doi.org/10.1002/smj.640
Vaccaro, M., Almaatouq, A., & Malone, T. (2024). When combinations of humans and AI are useful: A systematic review and meta-analysis. Nature Human Behaviour, 8, 2293–2303. https://doi.org/10.1038/s41562-024-02024-1
Vanneste, B. S., & Puranam, P. (2024). Artificial intelligence, trust, and perceptions of agency. Academy of Management Review. Advance online publication. https://doi.org/10.5465/amr.2022.0041
Wallerstein, I. (1974). The modern world-system I: Capitalist agriculture and the origins of the European world-economy in the sixteenth century. Academic Press.
Zuboff, S. (2019). The age of surveillance capitalism: The fight for a human future at the new frontier of power. PublicAffairs.