Dehumanising leadership: AI and the construction of post-human leadership Open Access

The rapid integration of artificial intelligence (AI) into organisational life poses profound challenges to leadership theory and practice. As AI systems increasingly assume decision-making, evaluative and strategic roles once reserved for human leaders, questions arise not only about accountability and ethics (Aziz et al., 2025; Hossain et al., 2025; Peifer et al., 2022; Tabata et al., 2025) but about the very nature of leadership itself (Van Quaquebeke and Gerpott, 2023). The image of the human leader, someone who is rational, intentional and embodied, has long shaped scholarly and popular views of leadership. Yet the growing plausibility of attributing leadership beyond the human calls this image into question. This shift offers a timely provocation: if an AI system can be treated as a leader, what does this reveal about the assumptions, expectations and cognitive processes that make leadership appear real?

This paper takes seriously the idea that AI does not simply augment leadership but destabilises its conceptual foundations. Drawing on social construction (Fairhurst and Grant, 2010), attribution (Calder, 1977 Meindl, 1995; Meindl et al., 1985), implicit leadership (Lord et al., 1984, 2020) and leadership categorisation theories (Epitropaki et al., 2013), I approach leadership as an attributional phenomenon, shaped by cognitive and cultural processes rather than grounded in intrinsic personal qualities. When observers interpret authoritative outputs as visionary or directive, they engage in the same interpretive processes used to recognise leadership in human actors, a dynamic made particularly visible in the case of AI.

This argument is particularly relevant as AI’s organisational influence accelerates. From predictive algorithms in recruitment to generative models informing strategic planning, AI technologies now produce outputs that influence high-stakes decisions across sectors (Chen, 2023; Doshi et al., 2025). Such systems, although devoid of consciousness or moral intent, are increasingly treated as legitimate sources of direction and control. Their authority is interpreted less through interpersonal qualities than through perceived alignment with prevailing expectations: offering solutions, reducing uncertainty and making decisions under pressure (Jackson and Parry, 2018). In this regard, AI reveals the hidden scaffolding of leadership attribution, raising critical questions about who or what can be accepted as a leader and under what conditions.

Rather than interpreting this development as the erosion of leadership, I suggest it provides a rare opportunity to re-examine how leadership has always been constructed. By exposing leadership’s dependence on perception, context and symbolism, AI highlights the need for a theory of leadership untethered from human personhood. The first part of this paper critically interrogates the human-centric assumptions of mainstream leadership theories. The second part explores how AI holds up a mirror to these assumptions, revealing the cognitive and ideological processes behind leader attribution. The paper concludes with a tentative model of post-human leadership that reframes leadership as an attributional and socially stabilised phenomenon, prompting reconsideration of legitimacy, responsibility and power beyond human exceptionalism.

Over the past four decades, leadership research has shifted from viewing leadership as an objective property of individuals to understanding it as a socially constructed process (Billsberry and O'Callaghan, 2024). This perspective recognises that leadership does not reside inherently in persons or positions but emerges through interpretation and shared expectations (Billsberry, 2026; Fairhurst and Grant, 2010; Meindl, 1995; Meindl et al., 1985). As such, leadership exists not as an intrinsic quality but as an attribution (Alvesson and Spicer, 2012; Junker and van Dick, 2014).

One of the earliest and most influential formulations of this perspective is Calder’s (1977) attribution theory of leadership. Calder argued that leadership is inferred by observers who retrospectively explain group outcomes by crediting individuals with influence. Leadership, in this framing, is not necessarily what causes performance but what is perceived to have caused it. Calder’s theory reveals the inferential nature of leadership and underscores the importance of cognitive processes in shaping who is deemed a leader. In an environment where AI systems generate successful outputs, such cognitive processes could equally apply to machines, given that the attribution of leadership rests less on agency and more on perceived efficacy.

Building on this, implicit leadership theory (ILT) introduced the idea that individuals hold cognitive schemas or prototypes that shape how they recognise and respond to leadership stimuli (Epitropaki et al., 2013; Lord et al., 1984; Offermann et al., 1994; Schyns, 2021). These schemas include culturally embedded expectations about what leaders look and act like, and they filter how behaviour is interpreted in organisational contexts. When a person fits these prototypes, observers are more likely to ascribe leadership status, regardless of actual effectiveness. ILT highlights the role of pre-existing mental models in leadership construction and opens the possibility that these models may adapt to incorporate non-human agents if they exhibit prototype-consistent behaviours. This implies that AI systems, if they display attributes such as analytical skill, decisiveness or impartiality, could be cognitively categorised as leaders, especially in contexts where traditional human signals like warmth or charisma are less salient.

Leadership categorisation theory (LCT) extends this logic by focusing on how people classify others as leaders through cognitive comparison with stored leadership categories (Lord and Maher, 1991; Lord et al., 1984; Pitsi et al., 2025). The process is evaluative: individuals assess the degree of match between observed behaviour and their ILT. As Epitropaki and Martin (2005) have shown, these categories are both stable and malleable, shaped by experience, cultural norms and organisational narratives. Importantly, LCT suggests that the locus of leadership is not the individual per se but the match between observed behaviour and socialised expectations. In this view, the legitimacy of AI as a leader is not a question of moral status but of categorical fit.

The social construction of leadership perspective complements these cognitive approaches by examining the relational and discursive dimensions of how leadership comes to be recognised. Rather than focusing on individual cognition alone, this strand of research examines the interactive processes through which leadership is produced, negotiated and sustained (Fairhurst and Grant, 2010). Leadership is seen as a collective accomplishment, performed through talk, ritual and symbolic action, and embedded in organisational and cultural contexts. These constructions are not neutral; they are infused with power, ideology and politics (Alvesson and Sveningsson, 2003; Billsberry and O'Callaghan, 2024). From this vantage point, the emergence of AI challenges the normative assumptions about who can be a leader and with what authority. If leadership is constituted in interaction and discourse, then the inclusion of algorithmic agents in leadership processes may reflect not a misattribution but an extension of the social conditions under which leadership is performed.

By making the attribution processes visible, we can more clearly see how non-human systems become folded into what organisations treat as leadership. This is not to suggest that AI should be equated with human leaders in terms of empathy, ethics or relational complexity, but rather to highlight that leadership recognition operates through interpretive and context-dependent mechanisms that do not inherently exclude machines.

AI is often presented as a technological tool; an instrument designed to optimise processes, enhance decision-making and support human judgment. However, in its application across domains of leadership practice such as strategic planning, resource allocation and personnel evaluation, AI does more than merely assist; it produces outputs that observers may interpret as performative within leadership contexts (Chen, 2023; Doshi et al., 2025). As AI systems are increasingly embedded into organisational life, they do not simply replicate leadership behaviours; they also reflect and, in some cases, disturb the underlying logics by which leadership is attributed and legitimised. This section explores how AI functions simultaneously as a mirror, revealing the cognitive scaffolding of leadership, and as a disruptor, unsettling long-held assumptions about its human foundations.

AI systems are already making decisions once reserved for senior organisational leaders. In recruitment, algorithms screen applicants and make shortlisting recommendations (Chamorro-Premuzic et al., 2017; Chen, 2023). In financial services, predictive systems execute trading strategies and assess credit risks (Pasquale, 2015). In health and emergency services, AI guides triage protocols and diagnostic pathways (Topol, 2019). In each case, AI is not merely providing Information but is actively shaping outcomes, exercising de facto authority in contexts with material consequences. These systems are frequently accepted as legitimate when they generate consistent, optimised and seemingly impartial results (Jarrahi et al., 2021), producing outcomes that align with leadership expectations such as reducing uncertainty and driving action (Alvesson and Spicer, 2012).

Crucially, acceptance of AI in these roles reflects not only its technical performance but also its alignment with implicit leadership prototypes. For instance, the use of natural language processing to generate strategic insights or organisational communication often mimics the tone, coherence and assertiveness expected of senior leaders (Doshi et al., 2025; Van Quaquebeke and Gerpott, 2024). In this sense, AI mirrors our own biases about what leadership should look and sound like.

Moreover, AI systems have been designed and trained on data that embeds historical assumptions and normative standards of leadership performance (Kordzadeh and Ghasemaghaei, 2022). This has led scholars to suggest that AI, rather than introducing novel leadership modalities, often reinforces prevailing leadership norms, particularly those associated with masculinised, technocratic or market-oriented ideals (Fotaki et al., 2014; Wajcman, 2017). Algorithms trained on performance reviews, leadership assessments or hiring data risk codifying and perpetuating stereotypes under the guise of objectivity. In this way, AI systems become both products and producers of leadership ideology, simultaneously reflecting and amplifying culturally sanctioned images of leadership. They hold a mirror to our conceptions of authority, not as an impartial reflection but as a curated rendering of historical biases embedded in data infrastructures (Beer, 2017).

Despite their human-made origins, AI systems can attain a form of attributed agency. Research in human–computer interaction has shown that people often anthropomorphise AI tools, assign them personality traits and respond emotionally to their outputs (Nass and Moon, 2000). In leadership contexts, this tendency extends to trusting AI decisions, deferring to algorithmic recommendations and treating systems as more competent than human counterparts in certain domains (Logg et al., 2019). This does not necessarily indicate belief in AI’s personhood but reflects the flexibility of attribution mechanisms. AI can therefore become a locus of leadership not by design, but through users’ interpretation of its performance. This tendency is magnified in high-stakes, high-complexity environments where the promise of data-driven neutrality carries significant persuasive power.

Importantly, the attribution of leadership to AI is not simply an error or misunderstanding. It reveals how fragile and malleable the concept of leadership truly is. When systems without consciousness, intention or emotion are treated as leaders, it becomes apparent that leadership does not require these qualities to be effective or accepted. In this regard, AI acts as a disruptor: it destabilises human-centric theories of leadership by showing that leadership is not bound to flesh, but to observers’ perceptions. The entry of AI into leadership spaces thus provides a unique opportunity to interrogate and perhaps abandon essentialist views of leadership that have dominated both research and practice.

AI’s most significant contribution to leadership studies may lie not in its functionality but in its epistemological disruption. By stripping away the presumption of human intentionality, AI systems highlight the interpretive and ideological processes through which leadership becomes visible and legitimate. Their presence in organisational life forces researchers and practitioners to confront difficult questions: What constitutes leadership when relational, embodied presence is removed? How do we reconcile attribution with accountability in non-human systems? And what kinds of leadership become imaginable or desirable when machines are no longer just tools, but perceived authorities? These questions do not signify the end of leadership, but its deconstruction, and with it, the possibility of reimagining its contours in light of changing technological conditions.

If leadership can be credibly attributed to AI, then it cannot be dependent on essential human qualities. This simple yet unsettling proposition calls into question the person-centric foundations of leadership theory. It argues that AI does not merely mimic leadership; it dehumanises it by exposing the myth that leadership requires personhood, emotional intelligence or moral intent. This dehumanisation of leadership does not imply the removal of value from leadership, nor does it suggest a dystopian future devoid of ethical direction. Rather, it describes a process through which leadership is dislodged from the human body and revealed as a symbolic and functional role that can be enacted, or at least perceived to be enacted, by non-human agents.

This reveals a critical paradox. Leadership theory often posits that effective leadership depends on uniquely human capacities such as empathy, ethical judgement or emotional intelligence (Goleman, 1995). Yet in practice, leaders are frequently legitimised through perceptions of outcome alignment rather than through these traits. This is where AI systems excel. They provide decisions that appear rational, data-driven and aligned with performance metrics. As a result, they can be granted the mantle of leadership through the same attribution processes that elevate human leaders. From this perspective, the widespread adoption of AI in leadership roles is not an aberration, but an amplification of leadership’s existing logic.

The implications of this shift are both conceptual and political. Conceptually, it undermines the anthropocentric orientation of leadership theory. Human exceptionalism, long assumed in theories of leader development, moral agency and relational influence, becomes harder to defend when machines are constructed as legitimate leaders through the same attribution processes applied to humans. AI systems, by virtue of their perceived impartiality and consistency, are often viewed as more trustworthy than human counterparts, especially in contexts marked by bias, fatigue or error (Logg et al., 2019). This inversion challenges foundational assumptions about the moral superiority or irreplaceability of human judgement in leadership. In turn, it prompts the need for a more reflexive leadership theory that takes attribution, perception and context as its primary materials.

Politically, AI exposes long-standing tensions in how leadership legitimacy is produced and maintained. When people are willing to attribute leadership to systems that cannot feel, relate or be held morally accountable, it becomes evident that such qualities are not central to how leadership is legitimised. Scholars have shown that authority in organisations is frequently routinised and depersonalised, relying more on procedural structures and institutionalised expectations than on the ethical or relational qualities often idealised in leadership theory (Alvesson and Spicer, 2012; Barker, 1993). AI makes these dynamics harder to ignore by demonstrating that leadership can be constructed through technical and impersonal processes. In doing so, it reveals how leadership is often validated through performance, efficiency and outcome delivery rather than through inspiration, emotional connection or moral presence.

AI also challenges the centrality of the leader–follower relationship. Much of contemporary leadership theory relies on relational constructs such as leader–member exchange, authenticity, trust and emotional contagion (Graen and Uhl-Bien, 1995; Sy et al., 2005). These theories assume that the leader’s capacity to influence depends on their ability to form meaningful relationships with others. Yet AI systems exert influence in the absence of emotional reciprocity. Their legitimacy often arises not from relational engagement but from perceived neutrality and competence (Lee, 2018). This suggests that relationality, while normatively desirable, may not be structurally necessary for leadership to occur.

In addition, the dehumanisation of leadership invites renewed attention to the role of discourse. As Fairhurst and Grant (2010) and Cunliffe (2009) have argued, leadership is constituted through language, storytelling and symbolic acts. AI disrupts these discursive practices by replacing narrative authority with statistical authority. Decisions justified by data models or optimisation algorithms do not require persuasion; they require calibration. This transforms the communicative landscape of leadership, shifting it away from meaning-making and towards procedural execution. In such environments, the leader’s voice may be less important than the system’s output. Leadership thus risks becoming a silent force that is coded, calculated and impersonal. These disruptions ultimately reflect the fragility of leadership constructs and the social scaffolding that sustains them.

Rather than mourning the dehumanisation of leadership, we might use it as an analytical opportunity. As AI takes up space in leadership domains, it creates a vantage point from which to observe leadership’s construction in action. It exposes the ways we fill ambiguous spaces of influence with attributions of authority. It challenges us to separate what we want leadership to be (such as relational, ethical, inspiring) from what it often is (such as procedural, impersonal and results-driven). And it encourages us to think beyond the human subject as the necessary site of leadership, opening the door to new theoretical possibilities.

If leadership can be attributed in the absence of personhood, intention or emotional presence, it underscores that leadership arises from attribution rather than essence and that our theories must account for this. This section offers a tentative model of post-human leadership. Post-humanism, as applied here, does not reject the value of human agency but calls for a decentring of the human subject in understanding how leadership emerges, is enacted and gains legitimacy in contemporary organisational systems (Braidotti, 2013). A post-human leadership model acknowledges the decentring of the human subject and conceptualises leadership as an emergent interpretive process rather than an individual capacity.

Post-human thinking in leadership draws inspiration from relational and distributed approaches already present in the literature. Relational leadership theory, for instance, rejects the notion of leadership as residing within individuals and instead views it as an emergent property of social processes and intersubjective meaning-making (Uhl-Bien, 2006). Distributed leadership theory similarly emphasises the coordination of multiple actors and artefacts across time and space (Gronn, 2002; Van Ameijde et al., 2009). While these frameworks remain primarily human-centred, they open conceptual space for acknowledging the agency of technologies, infrastructures and non-human systems in shaping organisational direction. Post-human leadership extends this logic by generalising beyond the human subject, treating leadership as a product of attribution and legitimation rather than as an outcome of interpersonal relations alone.

In a post-human view, leadership is not located in leaders themselves, but in the interpretations through which influence and coordination are made meaningful. Authority is recognised when outcomes are rendered intelligible and acceptable within shared expectations, rather than when they are traced to particular individuals or identities. This reframing shifts leadership theory away from personal attributes and towards the social processes through which leadership is inferred, stabilised and sustained. The core process of post-human leadership involves attribution, legitimation and reification.

Attribution refers to the process through which observers retrospectively explain outcomes, patterns of influence or moments of coordination by ascribing leadership to a focal point. In this view, leadership does not originate in intention, authority or embodied capacity, but in interpretive acts through which meaning is imposed on ambiguous organisational events. Observers seek cognitive closure by locating leadership as a cause of success, failure or direction, particularly in contexts characterised by uncertainty or complexity. This process is inherently inferential, drawing on implicit leadership theories that shape what observers expect leadership to look like and where they believe it should reside. Attribution is therefore not a neutral act of recognition but a selective and socially patterned judgement that privileges certain explanations over others. Crucially, attribution operates independently of agency; leadership may be ascribed even when no actor intended to lead or when influence emerged through diffuse or impersonal processes. Leadership, at this stage, remains provisional and contestable, dependent on whether these attributions resonate beyond the individual observer. Attribution thus marks the initial moment at which leadership is brought into being as a meaningful social category, without presuming its stability or legitimacy.

Legitimation refers to the process through which provisional leadership attributions gain social acceptance and normative credibility beyond the individual observer. Whereas attribution explains how leadership is first inferred, legitimation explains how these inferences come to be shared, reinforced and treated as appropriate within a given social context. Legitimation draws not only on shared expectations about leadership but only on implicit theories about what counts as a valid explanation and authoritative knowledge, which shape how leadership claims are evaluated and normalised (Eichler and Billsberry, 2023). Leadership attributions are more likely to be legitimised when they align with culturally embedded expectations about authority, influence and responsibility and when they resonate with prevailing institutional narratives about what leadership should look like (Suchman, 1995). This process is deeply social, unfolding through discourse, repetition and tacit agreement rather than formal validation. Legitimation does not require evidence of superior competence or moral intent; it requires coherence with existing frames of sensemaking and power relations (Fairhurst and Grant, 2010). Through legitimation, leadership begins to appear less as an interpretation and more as a reasonable or even obvious reading of events. However, at this stage, leadership remains contingent; its acceptance depends on continued alignment with shared expectations and ongoing social reinforcement. Legitimation thus marks the transition from private inference to socially sanctioned meaning, setting the conditions under which leadership can persist and solidify.

Reification refers to the process through which legitimated leadership attributions become taken for granted and treated as objective features of organisational reality. At this stage, leadership is no longer experienced as an interpretation or judgement but as something that simply exists, embedded in roles, structures, routines or focal points of authority. Reification occurs when alternative explanations recede from view and leadership attributions no longer require justification or active reinforcement. Through repetition and institutionalisation, leadership becomes detached from the contingent processes that produced it and is instead perceived as natural, durable and self-evident (Alvesson and Spicer, 2012; Berger and Luckmann, 1966). This transformation gives leadership its apparent solidity, allowing it to endure even when outcomes change or agency is ambiguous. Crucially, reification does not signal accuracy or legitimacy in any normative sense; it signals interpretive closure. Once reified, leadership becomes resistant to scrutiny, critique or re-attribution, making it difficult to disentangle influence from authority or responsibility from position.

This model invites reconsideration of the criteria through which leadership is interpreted and legitimised. If leadership is no longer tied to vision, charisma or emotional presence, then metrics of legitimacy must evolve. In post-human configurations, what is recognised as credible leadership may shift away from interpersonal influence towards criteria that appear objective, impersonal or outcome-based. This shift has implications for leadership development and education, which have traditionally focused on cultivating individual capabilities and reflective self-awareness (Bolden and Gosling, 2006). A post-human orientation demands that we also develop competencies in systemic oversight, algorithmic literacy and ethico-technical governance. It reframes the idea of leadership as one interpretive node among many in a broader ecology of influence.

Critically, post-human leadership is not a utopian vision. It raises important concerns about opacity, bias and disconnection. It risks reducing leadership to procedural optimisation and diminishing the role of human judgement, empathy and dissent. For this reason, the model is presented not as a normative ideal, but as an analytical tool: a way of understanding the emergent dynamics of leadership in algorithmic organisations. It highlights the need for reflexive engagement with the systems we build and the meanings we attribute to their outputs. Rather than resisting the implications of post-human conditions for leadership theory, a post-human approach urges us to confront their ethical and political consequences directly and critically.

The practical significance of the analysis presented in this paper lies not in the automation of decisions, but in how leadership is communicated, recognised and sustained under contemporary organisational conditions. This has significant implications for how leadership is cultivated and enacted in organisations. If leadership is understood as a product of attribution rather than an inherent property of individuals, then development practices aimed at those whose work involves influencing others need to reflect this interpretive foundation. Rather than focusing primarily on cultivating personal traits, such development should help individuals understand the processes through which legitimacy is constructed around their actions. This includes surfacing implicit leadership theories, exposing the prototypes that guide followers’ sensemaking and equipping individuals to engage more reflexively with the contexts in which leadership is attributed (Billsberry and O’Callaghan, 2024). By shifting attention from personal attributes to the dynamics of attribution, development efforts can better prepare organisational members to navigate environments where authority is negotiated rather than assumed.

At a more immediate level, this approach highlights the importance of developing awareness of how everyday actions are read and evaluated by others. Strengthening this perceptual sensitivity enables people to align their intentions more closely with how their behaviour is interpreted and to reduce the unintended attributions that arise in ambiguous or fast-moving situations. These interpretive skills are especially important in contexts where authority is decoupled from individual human actors, since organisational members must remain alert to how leadership attributions form around impersonal, collective or technologically mediated sources of influence. Building this form of attributional literacy offers a grounded and practicable way to translate an attributional model of leadership into everyday organisational life.

From a leader education and development perspective, the argument advanced in this paper points to a shift towards cultivating the communicative work through which leadership is made intelligible and credible in practice. Leaders must learn how to explain their reasoning, articulate the boundaries of their judgement and situate decisions within broader narratives of purpose, responsibility and constraint. Rather than relying on outcomes or positional authority to speak for themselves, leadership requires the ongoing articulation of why particular courses of action were taken and how alternatives were weighed. This form of interpretive work helps others locate leadership in the person who engages with uncertainty, rather than in processes, routines or artefacts that merely produce results. Developing these capabilities foregrounds leadership as an enacted, discursive practice, sustained through explanation and engagement rather than assumed authority. In this sense, leadership development becomes less about decisiveness and more about the capacity to render judgement visible and accountable.

A further implication concerns the risk of misattributing leadership under post-human conditions, particularly when authoritative outputs are produced without clear human agency. Because authoritative outputs can appear confident, consistent or analytically sophisticated, organisational members may sometimes read these outputs as carrying a degree of authority that no machine possesses. The practical challenge is to structure decision processes so that human judgment remains visible and accountable, rather than receding behind algorithmic outputs. This may involve clarifying when an AI system is providing input rather than direction, communicating the limits of what such systems can meaningfully “know,” and ensuring that the rationale for consequential decisions is articulated by those responsible for making them. In this sense, misattribution is less a technological failure than an interpretive one, and sustaining clarity about the boundary between support and authority is essential in hybrid human–AI environments.

From a governance perspective, this analysis suggests that organisations need explicit arrangements to prevent leadership from being quietly read into AI-generated outputs. When analytic systems routinely produce recommendations that shape consequential decisions, governance becomes the means by which responsibility is kept human and visible. This includes clearly specifying who has final decision authority, requiring that AI outputs are treated as inputs rather than directives and ensuring that named individuals remain accountable for outcomes. It also involves establishing clear routes for challenge and override, so that organisational members can question or escalate decisions that rely heavily on algorithmic recommendations. These arrangements are not about granting authority to AI, but about preventing its outputs from becoming authoritative by default. In this sense, governance functions as a practical counterweight to reification, ensuring that leadership remains a socially accountable judgement rather than an implicit property of technical systems.

In this paper, I have argued that the growing attribution of leadership to AI systems reveals more about the nature of leadership than about the capabilities of machines. AI generates decisions and outputs that people may frame as leadership, not because they constitute leadership in essence, but because of how they are interpreted. Drawing on attribution theory, ILT and constructionist perspectives, I have shown that leadership is not an intrinsic human trait, but a contextual and perceptual phenomenon. The perceived legitimacy of AI-led decisions demonstrates the extent to which leadership is conferred through framing, expectation and outcome alignment, rather than through human distinctiveness.

To engage with this shift constructively, I have proposed a tentative model of post-human leadership grounded in the processes of attribution, legitimation and reification. This model reframes leadership not as a capacity possessed by individuals or systems, but as a socially constructed phenomenon that comes into being through interpretation, gains durability through shared acceptance and acquires apparent reality through taken-for-grantedness. While AI served as a critical provocation for this analysis, the resulting model is not specific to algorithmic contexts. Rather, it offers a general account of how leadership is constructed, stabilised and rendered durable. The post-human leadership model invites leadership studies to move beyond human exceptionalism and to treat leadership as an emergent social accomplishment rather than an inherent human property.