Artificial intelligence is becoming increasingly embedded in business management, influencing decisions in strategy, operations, marketing, human resources, finance, and organizational control. Its managerial significance no longer lies only in its capacity to process information faster than humans, but in its growing ability to recommend, rank, predict, allocate, and sometimes decide. This shift raises important questions about how organizations should govern AI when it begins to affect managerial judgment itself. The central problem addressed in this review is that management research has often treated AI as a performance-enhancing tool while giving less sustained attention to its governance consequences. Three tensions remain particularly fragmented: the delegation of decision authority to algorithmic systems, the maintenance of organizational accountability in distributed human-machine arrangements, and the conditions under which managers trust or distrust AI-assisted decisions. These issues are analytically distinct but practically interdependent. The objective of this critical review is to synthesize literature on AI in business management through the integrated lenses of authority, accountability, and trust. Rather than presenting AI adoption as an inevitable route to efficiency, the review interrogates the organizational assumptions behind AI-enabled decision-making. It asks how AI changes managerial discretion, responsibility, oversight, and confidence in organizational decisions. The review concludes that AI governance in management must move beyond technical performance and address the institutional conditions under which AI-assisted decisions are authorized, explained, contested, and trusted. Authority, accountability, and trust should not be treated as separate implementation concerns but as a connected governance triad. Future management research should therefore conceptualize AI not merely as a tool, but as a socio-technical actor that reshapes managerial responsibility and organizational control.
Artificial intelligence has moved from a peripheral analytics capability to a central component of managerial decision-making across contemporary organizations. Management scholarship increasingly recognizes that AI is not simply another information technology, because it can learn from data, generate predictions, recommend actions, and reshape how organizational decisions are made [1]. This development has produced the automation–augmentation paradox: AI can extend managerial capability while simultaneously displacing or constraining human judgment [1].
The business management literature has often emphasized AI’s potential to improve performance, creativity, productivity, and decision quality. Studies of AI capability and organizational performance suggest that firms can benefit when AI is embedded in managerial routines, data infrastructures, and innovation processes [2]. Yet this performance-oriented framing risks underplaying the governance dilemmas created when AI systems begin to influence decisions that were previously understood as managerial responsibilities.
A critical problem is that AI changes the locus of decision authority without always making that shift explicit. Organizational decision-making structures increasingly involve hybrid arrangements in which managers, experts, data scientists, and algorithmic systems jointly shape outcomes [3]. Such arrangements complicate the question of who decides, because decision authority may be formally retained by managers while being materially shaped by algorithmic rankings, predictions, and recommendations [4].
This review addresses the need for a more integrated management perspective on AI governance. It argues that decision authority, organizational accountability, and managerial trust must be examined together, because authority without accountability becomes irresponsible delegation, accountability without authority becomes symbolic compliance, and trust without governance becomes managerial vulnerability [5]. The article therefore proceeds by mapping AI in business management, critically reviewing decision authority and oversight, and preparing the basis for later analysis of accountability, trust, research gaps, and implementation implications.
AI in business management now spans multiple domains, including strategy, operations, marketing, human resources, and organizational decision support. In marketing, AI is used for customer analytics, personalization, targeting, segmentation, and automated interaction, shifting marketing management toward data-intensive prediction and intervention [6]. In human resource management, AI supports recruitment, screening, performance evaluation, workforce analytics, and employee monitoring, but these applications also raise concerns about fairness, privacy, and procedural integrity [7].
The managerial role of AI differs across decision contexts because some systems merely support human judgment while others actively structure or automate decisions. Decision-support systems provide analysis for managers to interpret, decision-augmentation systems recommend or rank alternatives, and decision-automation systems execute or trigger actions with limited human intervention [8]. This distinction is essential because governance requirements become more demanding as AI moves from informational support to operational decision agency [3].
In strategic and organizational contexts, AI can influence not only isolated decisions but also the routines through which organizations define problems and evaluate knowledge. Research on learning algorithms shows that algorithmic systems can reshape organizational knowing by changing what counts as evidence, expertise, or a valid decision basis [9]. This means that AI adoption is not simply a technical implementation issue, because it affects organizational interpretation, managerial attention, and the distribution of epistemic authority [4].
Despite these implications, much AI management research still assumes that adoption is desirable when it improves efficiency, prediction, or scalability. This assumption is visible in studies that emphasize AI capability, performance, and creativity, but it becomes incomplete if governance risks are treated as secondary implementation barriers rather than central managerial concerns [2]. Table 1 maps the landscape of AI applications in business management and their decision-making roles.
Table 1. AI Applications in Business Management: Domains, Decision Types, and Levels of Autonomy
Business management domain | Typical AI applications | Dominant decision type | Level of AI autonomy | Main governance concern |
Strategy and planning | Market sensing, scenario analysis, strategic forecasting, competitive intelligence | Long-range judgment under uncertainty | Low to medium, usually advisory or augmentative | Risk of over-reliance on predictive outputs in uncertain strategic contexts |
Operations and supply chains | Demand forecasting, inventory optimization, scheduling, routing, process automation | Repetitive optimization and resource allocation | Medium to high, often semi-automated | Loss of human discretion and limited contestability of automated operational choices |
Human resource management | Recruitment screening, performance analytics, workforce planning, employee monitoring | Evaluation of people and allocation of opportunities | Medium, frequently recommendation-based | Bias, opacity, privacy intrusion, and erosion of procedural fairness |
Marketing and customer management | Personalization, segmentation, pricing support, recommendation systems, customer analytics | Prediction of customer behaviour and targeting decisions | Medium to high, depending on automation | Manipulation, unfair targeting, and weak accountability for algorithmic customer treatment |
Finance and risk management | Credit assessment, fraud detection, risk scoring, investment analytics | Risk classification and exception detection | Medium to high in rule-based or predictive settings | Responsibility gaps when decisions are justified by model outputs |
General management control | Performance dashboards, algorithmic monitoring, workflow allocation, compliance analytics | Monitoring, evaluation, and coordination | Medium, often embedded in managerial routines | Algorithmic control, surveillance, and normalization of automated evaluation |
Decision authority is one of the most significant but under-theorized consequences of AI adoption in business management. AI systems can appear to leave authority formally with managers while practically narrowing the options managers perceive as legitimate or defensible [3]. This creates a hidden transfer of authority, where algorithmic outputs become default judgments even when human decision-makers remain nominally responsible [10].
The literature distinguishes between augmentation and automation, but this distinction is often insufficient for understanding managerial authority. Augmentation assumes that AI assists human judgment, whereas automation implies that AI replaces certain human decisions, yet many organizational systems operate in between these categories [1]. In hybrid settings, managers may retain veto power but lose interpretive control because the algorithm defines the decision frame, ranks alternatives, or sets thresholds for intervention [11].
Human oversight is commonly presented as the solution to AI-related decision risks, but oversight varies substantially in form and effectiveness. Human-in-the-loop models require human approval before action, human-on-the-loop models allow monitoring and intervention, and human-out-of-the-loop models permit automated execution with limited human involvement [12]. Table 2 contrasts models of decision authority and human oversight in AI-assisted management.
Table 2. Models of Decision Authority and Human Oversight in AI-Assisted Management: Configurations, Risks, and Governance Requirements
Authority model | Decision configuration | Role of the manager | Main risks | Governance requirements |
Human-led decision support | AI provides information, forecasts, or analysis, but the manager interprets and decides | Primary decision-maker and interpreter | Selective use of evidence, confirmation bias, underuse of valuable AI insights | Clear documentation of how AI outputs inform managerial judgment |
AI-augmented managerial decision | AI recommends, ranks, or prioritizes options, while the manager approves or modifies the decision | Final approver and contextual evaluator | Automation bias, weak challenge of algorithmic recommendations, symbolic oversight | Mandatory review criteria, explainability, and decision logs |
Shared human-AI authority | Human and AI inputs jointly determine outcomes through embedded workflows | Co-producer of the decision with partial discretion | Ambiguity over who has authority and who is accountable | Explicit allocation of decision rights and escalation pathways |
Human-on-the-loop supervision | AI executes routine decisions while managers monitor system performance and intervene when needed | Supervisor and exception handler | Delayed intervention, deskilling, and excessive confidence in system stability | Continuous monitoring, alert thresholds, audit trails, and override mechanisms |
AI-led automation | AI executes decisions with minimal human involvement | Residual controller or post hoc reviewer | Loss of managerial discretion, opacity, and responsibility gaps | Strict domain limits, periodic audits, accountability assignment, and human appeal mechanisms |
The main danger is that oversight can become symbolic when managers lack the expertise, time, or authority to challenge AI outputs. Research on professionals using opaque AI systems shows that users may either disengage from AI recommendations or defer to them despite limited understanding, especially in high-stakes judgment contexts [11]. Decision authority therefore cannot be governed only by keeping a human formally involved; it requires designing conditions under which human judgment remains informed, empowered, and accountable [13].
Figure 1 presents the integrated authority–accountability–trust framework through which AI-assisted management decisions should be governed.
Figure 1. Integrated Governance Framework for AI-Assisted Management: Linking Decision Authority, Organizational Accountability, and Managerial Trust
Organizational accountability becomes more difficult when AI systems influence decisions through complex data pipelines, model architectures, vendor infrastructures, and managerial workflows. The central challenge is not simply that algorithms may be opaque, but that responsibility becomes distributed across designers, data providers, managers, executives, users, and external technology suppliers [14]. This diffusion creates an accountability gap in which harmful outcomes may be recognized, yet no single actor appears fully responsible for authorizing, explaining, or correcting them.
The “many hands” problem is especially acute in AI-enabled organizations because decision outcomes are often produced by interacting human and technical components. Studies of algorithmic work show that algorithms can become instruments of control while managers and workers contest how decisions are made, interpreted, and enforced [15]. When organizations treat algorithmic outputs as neutral or technical, they may obscure the managerial choices embedded in data selection, model objectives, performance metrics, and implementation rules [16].
Business ethics research emphasizes that algorithmic accountability requires more than compliance with abstract ethical principles. Ethical AI principles can help identify values such as fairness, transparency, and non-maleficence, but principles alone do not guarantee responsible organizational practice [17]. This limitation is visible when firms adopt ethics statements or guidelines without creating enforceable accountability structures, review mechanisms, or channels for affected stakeholders to contest AI-assisted decisions [18].
Recent work on AI governance has begun to define organizational mechanisms for assigning responsibility, managing risk, and aligning AI systems with institutional objectives. Organizational AI governance requires decision rights, monitoring structures, escalation procedures, documentation practices, and cross-functional oversight that connect technical model management to managerial accountability [19]. However, governance frameworks remain limited if they focus on formal structures while neglecting power asymmetries, organizational incentives, and the possibility that AI can rationalize unaccountability by making managerial choices appear technically inevitable [16].
Managerial trust in AI is not a simple attitude of acceptance, but a situated judgment about whether an AI system is competent, reliable, understandable, fair, and aligned with organizational goals. Research on human trust in AI shows that trust depends on both system characteristics and user characteristics, including perceived accuracy, transparency, task context, and prior experience [5]. For managers, trust is especially consequential because AI recommendations can affect resource allocation, personnel evaluation, customer treatment, and strategic priorities.
Trust can develop when AI systems consistently provide useful outputs, but it can also become excessive when managers defer to algorithmic recommendations without adequate scrutiny. The literature on algorithm appreciation shows that people may sometimes prefer algorithmic judgment to human judgment, especially when algorithms are perceived as objective or statistically superior [20]. Yet algorithm aversion research shows that trust can collapse after visible errors, even when comparable human errors would be tolerated more easily [10].
Explainability and transparency are often presented as solutions to mistrust, but their managerial value depends on whether explanations are actionable. If explanations are too technical, too generic, or disconnected from managerial decision rights, they may produce only superficial confidence rather than informed trust [21]. Trustworthy AI therefore requires not only interpretable outputs, but also organizational routines that allow managers to question, override, document, and learn from AI-assisted decisions [22].
Trust is also shaped by ethical and relational concerns, particularly when AI affects people rather than only operational variables. Studies of resistance to medical AI show that individuals may distrust algorithmic judgment when they believe it cannot recognize human uniqueness or contextual nuance [23]. Table 3 summarises the antecedents, dynamics, and consequences of managerial trust in AI.
Table 3. Managerial Trust in AI-Assisted Decisions: Antecedents, Erosion Factors, and Rebuilding Mechanisms
Trust dimension | Antecedents of trust | Factors that erode trust | Managerial consequences | Rebuilding mechanisms |
Technical reliability | Accurate predictions, stable performance, validated outputs | Visible errors, inconsistent recommendations, poor model monitoring | Reduced adoption, selective use, or defensive decision-making | Performance audits, validation cycles, and transparent error reporting |
Explainability | Understandable rationale, clear input-output logic, usable explanations | Black-box outputs, technical opacity, unexplained ranking or scoring | Symbolic approval or blind deferral | Manager-oriented explanations and decision documentation |
Fairness and integrity | Evidence of non-discrimination, procedural fairness, ethical safeguards | Biased outcomes, privacy concerns, unfair treatment of employees or customers | Resistance, reputational risk, and legitimacy loss | Bias testing, stakeholder review, and appeal mechanisms |
Managerial control | Ability to question, override, and contextualize AI outputs | No meaningful override, rigid automation, unclear escalation pathways | Loss of discretion and accountability anxiety | Defined decision rights, escalation rules, and human review protocols |
Organizational learning | Feedback loops, post-decision review, continuous improvement | Repeated errors without correction, lack of feedback channels | Declining confidence and disengagement from AI systems | Learning routines, incident review, and model improvement governance |
A major gap in AI management research is the absence of integrated theoretical frameworks that connect decision authority, accountability, and trust. Existing studies often examine AI capability, algorithmic work, ethics, or trust as separate domains, but managerial governance problems emerge precisely because these dimensions interact [24]. Authority determines who can act, accountability determines who must answer, and trust determines whether AI-assisted decisions are accepted, challenged, or ignored.
A second gap concerns methodological imbalance. Much research on trust in algorithms relies on experiments, vignettes, or controlled decision tasks, which are useful for isolating mechanisms but limited for understanding how trust develops over time in organizations [20]. Field-based studies of AI implementation show that organizational meaning, professional identity, and existing routines strongly shape how AI is interpreted and used in practice [25-27].
A third gap is the limited attention to power and control in AI-assisted management. Algorithmic systems do not merely support decisions; they may redistribute authority, intensify monitoring, standardize judgment, and alter the conditions under which employees and managers can contest decisions [15]. Research on algorithmic management and organizational knowing suggests that AI can reshape workplace control by redefining what counts as legitimate knowledge and whose judgment matters [9].
A fourth gap concerns the practical translation of AI ethics into organizational governance. Reviews of AI ethics tools and guidelines show that organizations need concrete methods for moving from principles to operational practices, yet implementation remains uneven and fragmented [22]. Table 4 consolidates the critical gaps and suggests future research priorities.
Table 4. Critical Gaps in AI Management Research: Deficits in Theory, Method, and Evidence across Authority, Accountability, and Trust
Critical gap | Current limitation in the literature | Why it matters for management research | Future research priority |
Fragmented theorization | Authority, accountability, and trust are often studied separately | AI governance problems arise from their interaction, not from isolated variables | Develop integrated socio-technical governance frameworks |
Weak field evidence | Many studies rely on experiments, vignettes, or conceptual argument | Organizational AI use unfolds over time through routines, politics, and learning | Conduct longitudinal field studies of AI-assisted management |
Limited attention to power | AI is often framed as efficiency-enhancing rather than control-shaping | Algorithmic systems can redistribute discretion and intensify surveillance | Examine power, resistance, and contestability in AI-enabled organizations |
Underdeveloped accountability models | Responsibility is discussed ethically but less often operationalized organizationally | Firms need clear mechanisms for assigning, auditing, and enforcing responsibility | Study accountability structures, audit systems, and escalation mechanisms |
Narrow view of trust | Trust is frequently treated as adoption willingness | Managerial trust includes scrutiny, informed reliance, and justified skepticism | Investigate calibrated trust, distrust, and trust repair after AI failure |
Insufficient sectoral comparison | Evidence is concentrated in selected domains such as HR, marketing, and expert work | Governance risks differ across strategic, operational, people-related, and customer-facing decisions | Compare AI governance across sectors, functions, and institutional environments |
The first implication is that organizations must design explicit decision authority protocols before deploying AI in management processes. These protocols should specify whether AI is advisory, augmentative, supervisory, or autonomous, and they should clarify when managers may accept, reject, override, or escalate AI outputs [3]. Without such protocols, organizations risk creating informal authority shifts in which algorithmic recommendations become de facto decisions while managers remain formally accountable [1].
The second implication is that AI accountability must be institutionalized through audit structures, documentation practices, and review mechanisms. Business ethics research shows that accountability cannot be delegated to technical teams alone, because algorithmic decisions reflect organizational values, incentives, and governance choices [14]. Organizations therefore need cross-functional AI governance bodies that include management, legal, ethics, information systems, and domain experts rather than treating AI oversight as a narrow compliance function [28].
The third implication is that explainability should be treated as a managerial capability rather than only a technical model property. Managers do not merely need model transparency; they need explanations that support judgment, contestation, and responsibility in specific decision contexts [21]. This requires training managers to interpret AI outputs critically, recognize uncertainty, identify possible bias, and document the reasons for accepting or rejecting algorithmic recommendations [8].
The fourth implication is cultural and leadership-oriented. Leaders must avoid presenting AI as an objective substitute for managerial responsibility, because this framing encourages over-reliance, symbolic oversight, and rationalized unaccountability [16]. Instead, leadership should build an AI-accountable culture in which AI is used to improve decisions while preserving human responsibility, stakeholder contestability, and calibrated trust [29].
Figure 2 translates the review’s critical synthesis into a staged governance pathway for implementing AI responsibly in business management.
Figure 2. Governance Pathway for Responsible AI Implementation in Business Management: From AI Adoption to Accountable and Trustworthy Decision Systems
This critical review has argued that AI in business management cannot be understood only through the language of performance, efficiency, and prediction. As AI systems increasingly recommend, rank, evaluate, and decide, they alter the governance foundations of managerial work. The most important question is therefore not whether AI improves decisions in general, but under what conditions AI-assisted decisions remain authorized, accountable, and trustworthy.
The review has positioned decision authority, organizational accountability, and managerial trust as an interdependent triad. Authority determines how decision rights are distributed between managers and AI systems, accountability determines how responsibility is assigned and enforced, and trust determines whether managers rely on AI in a calibrated and responsible way. Treating these issues separately risks producing organizations in which AI is powerful but poorly governed.
Future management research should treat AI as a socio-technical governance challenge rather than merely as a technological resource. This requires longitudinal field studies, stronger theory on human-machine authority, deeper analysis of accountability structures, and more nuanced models of managerial trust and distrust. The practical task for organizations is not to choose between human judgment and AI, but to govern their interaction responsibly.
None
None
None
None
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Institute for Management, Business, and Accounting Studies
