Artificial intelligence in secondary schools: implications for administrators across four leadership dimensions Open Access

Artificial intelligence (AI) usage among postsecondary students has risen from 74% in 2023 to 94% in 2025 (Kumar and McGray, 2025), representing a fundamental shift in how educational work is conceived, executed, and evaluated. While current AI in education (AIEd) builds upon decades of academic research and development, with educational technologists anticipating transformative potential since the 1980s (Williamson and Eynon, 2020), the recent proliferation of accessible generative AI (GenAI) tools has created an implementation reality that educational systems were unprepared to address (Dawson, 2023). However, institutional policy responses have been characterized by inconsistency and reactive positioning, with many guidelines prioritizing workforce preparation over critical examination of AI's risks and lacking coherent implementation frameworks (Henry, 2025; Sarkar and Kumar, 2024). This policy vacuum has necessitated bottom-up leadership responses, exemplified by University College London's Faculty of Law adopting secured, in-class assessment protocols to address concerns about AI overreliance (Henry, 2025). Such localized initiatives underscore how educational leaders navigate uncharted territory while exercising what Selznick (1957) termed “institutional leadership”—the critical responsibility of infusing organizational actions with value beyond technical requirements—as authorities struggle to provide guidance. This leadership challenge becomes particularly acute at the secondary school level, where administrators must balance multiple dimensions of their leadership role while confronting technological change.

Widespread AI adoption occurs within a research and policy landscape disproportionately focused on postsecondary education, creating a critical knowledge gap regarding AI usage in secondary schools, which serve as formative environments for developing digital citizenship and ethical reasoning (Chiu, 2021). This disparity is particularly concerning given that nearly 60% of K-12 principals currently use AI tools (Kaufman et al., 2025), yet fewer than 10% of K-12 schools globally have established AI policies (UNESCO, 2024). International research confirms this trend, with principals and teachers demonstrating sufficient knowledge of ChatGPT despite concerns about ethical use and cognitive impacts (Cetin et al., 2024). The policy vacuum is especially pronounced in Canada, where, despite preliminary guidelines in Ontario, Quebec, Alberta, and British Columbia, no comprehensive federal or provincial framework exists to guide implementation (Canadian Teachers’ Federation [CTF], 2025). The absence of coherent policy architecture creates substantial risks related to equity, fairness, privacy, and professional standards, while potentially exacerbating existing educational inequalities across diverse school communities.

This research gap has profound implications for secondary school administrators whose decisions fundamentally shape AI integration within their institutions. These leaders determine conditions under which educators and students encounter AI technologies, establishing ethical frameworks, resource allocations, and professional development (PD) priorities across what scholars recognize as multiple, intersecting dimensions of educational leadership (Leithwood et al., 2004) that ultimately influence whether AI serves as a tool for educational enhancement or becomes a source of institutional fragmentation and inequity.

This paper addresses this research gap by examining how secondary school administrators can lead the ethical and responsible integration of AI within complex organizational environments that demand both technological innovation and preservation of educational values. Through systematic analysis of current practices and theoretical examination of administrative functions, the paper presents a nuanced understanding of where AI can appropriately augment leadership work and where human discretion remains essential.

To systematically analyze the intersection of AI and secondary school leadership, the study employed established dimensions of educational leadership theory that collectively encompass the breadth of administrative responsibilities in contemporary schools. Building on foundational work that conceptualizes educational administration as a multifaceted practice (Leithwood et al., 2004; Sergiovanni, 1996), this multidimensional approach recognizes that school administrators operate within complex organizational environments where different types of leadership functions require different capabilities, some enhanced by AI while others demand distinctly human judgment and conventional modus operandi. These dimensions were selected for their comprehensive coverage of administrative functions while maintaining analytical clarity and avoiding both the fragmentation of examining numerous specialized theories and the oversimplification of a unitary leadership model. This multidimensional framework aligns with Chiu's (2021) holistic approach to AI curriculum design, which emphasizes the interconnection of content, product, process, and praxis, recognizing that AI integration cannot be understood through a single lens.

Specifically, the study draws upon four complementary leadership frameworks. First, instructional leadership (Edmonds, 1979; Hallinger, 2005; Leithwood et al., 2004) emphasizes the administrator's role in improving teaching and learning outcomes through curriculum support, PD, and educational vision-setting. Second, managerial leadership (Bush, 2007; Mintzberg, 1973) focuses on operational efficiency and organizational functionality, including resource allocation, policy implementation, and administrative coordination. Third, strategic leadership (Andrews, 1971; Davies and Davies, 2006) addresses long-term planning and institutional direction-setting, encompassing environmental scanning, organizational adaptation, and sustainable improvement initiatives. Finally, relational leadership (Burns, 1978; Noddings, 1984; Shapiro, 2006) foregrounds human connections, values-based decision-making, and stakeholder trust-building as fundamental to effective educational governance.

Decades of empirical validation confirm such leadership frameworks’ utility in capturing distinct administrative functions (Hallinger and Heck, 1998; Robinson et al., 2008). The frameworks persist in scholarship and practice precisely because they provide analytical clarity while reflecting the actual work of school administrators. Besides, AI integration offers unprecedented opportunities to address these historical limitations: augmenting instructional leadership with personalized learning analytics, elevating managerial functions beyond mere efficiency, enhancing strategic planning with predictive modelling, and potentially strengthening rather than replacing relational connections through more efficient administrative processes.

The present study's theoretical framework enables analysis that moves beyond simplistic technology adoption models toward a sophisticated understanding of how AI integration intersects with the complex realities of educational leadership. By examining administrative functions through multiple theoretical lenses simultaneously, the study sheds light on not only what administrators are currently doing with AI, but also what they should be doing to ensure that technological integration serves educational equity, maintains institutional trust, and preserves the human elements that define effective school leadership. The study's analysis focuses on identifying functions that inherently require human discretion, cultural sensitivity, and relational judgment. These capacities remain fundamentally human even as AI systems become increasingly sophisticated and adept at mimicking such human traits.

These four leadership dimensions collectively address the scope of secondary school administrative practice without excessive overlap. Instructional and strategic leadership capture the core educational mission, while managerial leadership addresses operational realities. Ethical/relational leadership provides the values-based foundation underlying all educational decision-making, particularly crucial when considering technologies that may reshape human relationships within school communities. Together, these dimensions encompass both the technical and adaptive challenges that Heifetz and Linsky (2017) identify as central to educational leadership, while avoiding the fragmentation that might result from examining numerous specialized leadership theories. Alternative frameworks, such as transformational leadership (Bass and Riggio, 2006) or distributed leadership (Spillane, 2006), while valuable, either overlap significantly with this study's chosen dimensions or address organizational dynamics beyond the scope of individual administrative decision-making that forms the study's analytical focus. Recognizing that AI also impacts other administrative domains such as facilities management, budgeting, student discipline, and board governance, this analysis maintains focus on these four core leadership dimensions to provide analytical depth rather than superficial breadth. These ancillary functions often operate through the four dimensions examined here: budgeting through managerial leadership, discipline through relational leadership, and governance through strategic leadership.

This scoping review explores the research question “what is required to prepare secondary school administrators to lead the ethical and responsible integration of AI?” by analyzing administrative functions across the four leadership dimensions and examining how AI may facilitate these functions. The analysis explores their implications, both positive and negative, including equity concerns. Scoping reviews are optimal for initial assessments in emerging areas (Grant and Booth, 2009).

The following keywords were used: “AI education AND administrators OR principals AND secondary school OR high school” and “Leadership theory AND education AND/OR AI.” Sources were obtained from Google Scholar, library databases (Omni), and grey literature, including news articles, reports, and policy documents. Results were limited to post-2018, corresponding with ChatGPT's initial development (Marr, 2023), though older sources were included to establish theoretical frameworks and reference existing policy. Sources emphasizing secondary education were prioritized, with higher education sources excluded, except for two that addressed policy contexts (Dawson, 2023; Henry, 2025). From broad searches across multiple databases, 21 sources meeting relevance criteria were selected, representing a purposive sample from the extensive literature rather than being restricted to specific journals or geographic regions.

Sources were analyzed to determine their most applicable leadership dimension, then synthesized to form a cohesive narrative about how administrators may apply AI within their respective leadership functions and the implications for educational leadership. The analysis reveals a complex landscape where AI integration intersects differently with each leadership dimension, creating opportunities for administrative enhancement and risks of fundamental role transformation. Notably, findings indicate significant misalignment between where AI is deployed most frequently (relational leadership functions) and where it may be most appropriate (certain managerial and strategic functions). This pattern suggests administrators navigate AI adoption without sufficient theoretical grounding or policy guidance, potentially undermining the human connections research identifies as central to effective educational leadership.

Administrative functions were analyzed to identify how educational leadership activities align (or misalign) with AI integration across the four established dimensions. Instructional leadership functions include organizing PD for teachers and planning opportunities for students. Managerial functions encompass teacher evaluations, staffing decisions, and resource allocation. Strategic functions include completing ministry reports and school improvement plans (SIPs). Relational functions include drafting communications, crisis response, mental health alerts, and monitoring academically struggling students.

AI delegation in instructional leadership fundamentally challenges educational expertise and professional authority. It has been experimented across the globe—in Indonesia (Halomoan et al., 2024), in the U.S. (Kaufman et al., 2025), in China (Hou et al., 2024)—and instructional leadership's union with AI has also been recently explored (Bixler and Ceballos, 2025). Hallinger's (2005) principal-as-lead-learner model faces disruption when administrators use AI for PD design. The transformation to principal-as-AI-curator represents a qualitative shift in instructional authority.

AI may help administrators organize PD for teachers (selecting content, drafting activities) and plan student opportunities, including credit recovery and postsecondary exploration. Administrators readily use AI for teacher-oriented functions. As one principal in a U.S.-based RAND survey noted:

Similarly, Cetin et al. (2024) found that school principals and teachers identified lesson planning as a primary advantage of ChatGPT use; however, they expressed concerns about weakening students' cognitive skills and the potential for academic dishonesty. Chiu's (2021) study of 24 teachers across 12 K-12 schools further identified teacher–student communication and flexibility as critical components of AI curriculum implementation, suggesting that administrators must consider teachers' perspectives when designing AI-related PD.

AI also assists with student opportunity planning. Ontario secondary schools commonly organize “credit recovery” days in which students complete unsubmitted work to earn course credits. Administrators may also plan university/college fairs and allocate time for the Ontario Universities' Fair (Ontario Ministry of Education [OME], 2013).

AI can help administrators plan such opportunities while minimizing instructional time loss. Research indicates principals currently use AI for school scheduling, field trips, and events (Kaufman et al., 2025). However, AI-mediated opportunity planning for postsecondary exploration may reinforce existing educational stratification if algorithmic recommendations reflect historical patterns of student streaming that have disproportionately disadvantaged racialized and working-class students.

Whether AI facilitates teacher PD or student opportunities, the goal is to boost teacher capacity and student learning. But is this how AI is most often used? In an interview, Kaufman reveals the answer remains unknown:

Efficiency-focused AI implementation raises critical equity concerns, as apparently neutral systems may reproduce systemic inequalities by privileging dominant educational discourses while marginalizing pedagogical approaches serving Indigenous students, newcomers, and students with disabilities (Benjamin, 2019).

In Ontario, where the Truth and Reconciliation Commission (TRCC, 2015) calls to action compel educational leaders to honour Indigenous ways of knowing, AI-generated PD content may inadvertently reproduce colonial educational frameworks unless explicitly designed to challenge them. This concern is particularly acute where recent policy initiatives emphasize culturally responsive pedagogy and anti-oppressive education frameworks requiring human judgment and cultural competence rather than algorithmic efficiency.

The RAND survey finding that principals use AI to generate “sentence stems to help teachers have restorative conversations” (Kaufman et al., 2025, p. 14) exemplifies concerning techno-managerial approaches to complex relational challenges. This transforms restorative justice, which fundamentally requires cultural sensitivity and contextual understanding, into scripted interactions mediated by algorithmic recommendations. Such applications risk reducing teachers to implementers of AI-generated protocols rather than professional decision-makers capable of responsive, culturally relevant practice. This reduction of professional judgment to algorithmic compliance threatens the adaptive expertise distinguishing effective educational leadership from mere management.

These professional judgment concerns become equally complex in managerial leadership functions, where AI applications in human resources, evaluation, and resource allocation raise parallel efficiency-versus-equity questions.

Managerial leadership emphasizes administrative efficiency in operational duties essential for school functionality. Bush (2007, p. 396) argues that “achieving functional schools is an essential requirement if learning is to take place,” but cautions against reducing educational leadership to mere administrative efficiency. Within secondary schools, AI integration in teacher evaluation, human resources, resource allocation, scheduling, compliance monitoring, and data management presents both opportunities and substantial risks (Adams and Thompson, 2025).

The intersection of AI with human resource functions represents a consequential application of managerial leadership technology. According to the RAND survey, administrators report using AI for core managerial duties including writing job descriptions, developing interview questions, and supporting teacher evaluation and feedback processes. One principal noted using AI to process descriptive teacher information as an evaluation starting point, reducing completion time by approximately 50% (Kaufman et al., 2025). However, this efficiency focus raises fundamental questions about professional judgment in personnel decisions.

AI-assisted hiring processes were initially championed as solutions to documented human prejudices in recruitment, including discrimination against candidates with ethnic names and gender biases (Bertrand and Mullainathan, 2004; Moss-Racusin et al., 2012). Yet AI systems trained on historical data often perpetuate and systematize these biases (Ajunwa, 2021; Buolamwini and Gebru, 2018). The troubling irony: technologies implemented to overcome human prejudice instead encode and amplify historical discrimination patterns affecting racialized candidates and those from non-traditional educational backgrounds. In secondary schools, where teacher diversity significantly impacts student outcomes, particularly for Indigenous, newcomer, and racialized populations, this algorithmic bias could systematically exclude educators whose lived experiences align with student community needs.

AI applications in resource allocation represent another critical managerial function in which technology intersects with equity considerations. Predictive models can analyze historical data to inform budget decisions, identify schools requiring support, and optimize resource distribution. However, these systems risk perpetuating historical inequities: schools serving predominantly low-income or racialized communities may receive algorithmic recommendations reflecting past underinvestment rather than current needs under the guise of data-driven decision-making.

The RAND findings reveal a concerning equity dimension: principals in high-poverty schools are approximately half as likely as those in low-poverty schools to receive AI implementation guidance, despite similar usage rates (Kaufman et al., 2025). This guidance gap means schools serving the most vulnerable populations—those requiring thoughtful, equity-conscious AI integration—navigate implementation with the least institutional support.

While AI can streamline data collection, identify patterns in observations, and generate preliminary feedback drafts, evaluative judgment about teaching effectiveness requires human discretion, cultural competence, and relational understanding. The evaluation time reduction reported by RAND participants may represent efficiency gains but could also signal the commodification of professional judgment—transforming nuanced assessments into algorithmic outputs.

AI-generated evaluation content may standardize feedback approaches, potentially undermining culturally relevant evaluation practices. In Ontario, where recent policy initiatives emphasize anti-oppressive education and Indigenous pedagogies, standardized AI evaluation tools may conflict with efforts to recognize diverse teaching approaches serving historically marginalized populations.

Bush's (2007, p. 396) warning that principals and educators who do not “own” innovations but merely implement external changes “without enthusiasm” becomes particularly relevant. Managerial AI implementation requires distributed leadership approaches engaging teachers, support staff, and community in defining appropriate technological boundaries rather than imposing top-down efficiency mandates that undermine relational foundations of effective schools.

Strategic leadership encompasses innovation and long-term planning, requiring leaders to map institutional futures while establishing direction and “moral purpose” (Davies and Davies, 2006, p. 123). This demands environmental scanning, trend analysis, and adaptive capacity-building. AI integration presents analytical capabilities alongside challenges to wisdom-based decision-making essential for effective strategic leadership.

Ontario's School Effectiveness Framework mandates principals and vice-principals to complete SIPs while superintendents provide systemic alignment (OME, 2013). The RAND survey reveals principals increasingly employ AI for policy development, improvement planning, survey creation, and data analysis (Kaufman et al., 2025). This practice presents a double-edged sword for strategic leadership effectiveness.

On one hand, AI-assisted policy development may identify issues that administrators might unintentionally omit due to cognitive limitations, time constraints, or unconscious oversight (Kahneman, 2011; Simon, 1997; Sweller, 1988). On the other hand, when AI generates policy content, administrators may lack intimate knowledge of document details, potentially hampering subsequent decision-making. Leaders responsible for policy implementation may forget which problems have been addressed and which remain unattended, creating gaps between policy existence and practical application—what implementation scholars term the “knowing–doing gap” (Pfeffer and Sutton, 2000).

AI excels at processing datasets and identifying patterns to support long-term planning, enabling administrators to analyze demographic trends, academic performance, and resource allocation over multiple years. However, algorithmic analysis relies on historical patterns, potentially missing discontinuous changes or emerging phenomena requiring strategic adaptation. AI systems privilege quantifiable metrics while undervaluing qualitative indicators like community relationships and cultural responsiveness.

Most critically, strategic leadership requires integrating “people, contextual, and procedural wisdom” to establish institutional vision and moral purpose (Davies and Davies, 2006, p. 134). Such integration of combining human insight, cultural understanding, and reflective practice remains fundamentally human. When administrators delegate vision-setting to AI systems, they risk excluding stakeholder voices and marginalizing community knowledge. In Ontario, where the TRCC (2015) calls to action mandate incorporating Indigenous perspectives in institutional planning, AI-generated strategic documents may reproduce colonial frameworks unless explicitly designed to challenge them.

Concerns about cognitive offloading due to AI overuse, identified by Gerlich (2025), become particularly relevant in strategic contexts where sustained analytical thinking distinguishes effective educational leadership from administrative management. If administrators increasingly rely on AI for environmental analysis and planning generation, they may experience attenuation of strategic thinking capabilities essential for institutional leadership.

Effective strategic AI implementation requires careful boundary-setting: leveraging algorithmic capabilities for data processing while preserving human judgment for vision-setting, values integration, and stakeholder engagement. This aligns with Bush's (2007) warning that externally imposed innovations lacking authentic ownership may fail to generate implementation enthusiasm. Yet relational leadership, which focuses on human connections and values-based decision-making, represents the dimension in which AI integration raises the most fundamental questions about preserving educational leadership's essentially human character.

Relational leadership evokes human connections while attending to values, equity, and stakeholder trust, bridging theory and practice through reflective and critical decision-making (Shapiro, 2006). This dimension encompasses interpersonal foundations enabling all other leadership functions. Yet paradoxically, relational leadership represents where AI integration is most prevalent, raising questions about preserving authentic human connection.

Administrative duties involving relational leadership include drafting communications, crisis response, mental health alerts, and monitoring academically struggling students, including IEP development. According to RAND and EdWeek surveys, communication drafting represents principals' primary AI application, encompassing emails, speeches, and stakeholder correspondence (Banerji, 2025; Kaufman et al., 2025). One principal described: “I use [ChatGPT] for weekly newsletters and emails to staff. It helps me dump my thoughts out and put them in a more concise and coherent form” (Kaufman et al., 2025, p. 14).

While offering efficiency gains, AI-assisted communication fundamentally alters authentic leadership interaction. Relational leadership requires genuine stakeholder understanding, cultural sensitivity, and an authentic voice—qualities that emerge from sustained relationship-building rather than algorithmic text generation (Northouse, 2019). When administrators delegate communication crafting to AI, they risk creating technically proficient but relationally hollow interactions that undermine trust.

AI integration in crisis response presents complex ethical challenges. While algorithmic systems can identify concerning behavioral patterns, crisis response requires emotional intelligence, cultural competence, and relational judgment (Goleman, 1995). AI monitoring systems may create surveillance environments that undermine psychological safety essential for student development, while potentially disproportionately flagging marginalized students whose expressions are misinterpreted by systems trained on dominant cultural norms (Benjamin, 2019).

IEP development is the most relationship-intensive administrative function, requiring a deep collaborative understanding of students' strengths, family dynamics, and cultural contexts. AI-assisted IEP development may reduce complex human needs to standardized outputs, potentially overlooking unique circumstances and cultural factors essential for effective special education planning.

The prevalence of AI use in relational functions creates an “authenticity paradox”: the leadership dimension most requiring genuine human connection increasingly relies on AI mediation. Shapiro's (2006) emphasis on authentic, reflective leadership becomes particularly relevant. If administrators consistently delegate relationship-building communications to AI, they may experience attenuation of emotional intelligence and interpersonal skills that distinguish educational leaders from managers. The cognitive offloading that Gerlich (2025) identifies is especially problematic in relational contexts that require sustained empathetic communication practice.

Extensive AI use in relational leadership raises questions about institutional trust and democratic engagement. While empirical research on stakeholder reactions to AI-mediated educational communications remains limited, broader studies of algorithmic transparency suggest potential impacts on trust relationships (Riedl, 2019). As GenAI systems improve their natural language processing, cultural sensitivity, and personalization, current concerns about authenticity and detection may prove transitional as stakeholder attitudes evolve with technological and social norms. This evolutionary acceptance is already evident among students: according to Kumar and McGray (2025), 71.68% consider it acceptable to use AI for composing emails to other students, 69.3% for faculty emails, and 67.11% for administrator communications, suggesting that trust concerns about institutional authenticity may prove transitional as generational attitudes shift toward normalized AI integration.

Across all four leadership dimensions, a consistent pattern emerges: AI adoption is proceeding rapidly without adequate theoretical grounding or consideration of foundational leadership principles. The misalignment between AI's current deployment in relational functions and its potential strengths in analytical tasks suggests that efficiency concerns are overriding educational values, raising fundamental questions about the future of school leadership.

This analysis reveals critical implications spanning policy development, professional practice, and institutional equity that demand immediate attention from educational leaders and policymakers.

Our findings reveal a significant misalignment between AI's current deployment (in relational leadership functions) and where it may be most appropriate (in certain managerial and strategic functions). This pattern suggests secondary school administrators navigate AI adoption without sufficient theoretical grounding or policy guidance, potentially undermining human connections, which research identifies as central to effective educational leadership (Leithwood et al., 2004; Noddings, 1984; Sergiovanni, 1996). The prevalence of AI use in communication drafting—the most fundamentally human aspect of leadership—while underutilizing AI's analytical capabilities in data processing and resource allocation, represents a concerning inversion of appropriate technological application.

Each leadership dimension presents distinct equity challenges when AI systems are integrated without critical examination. In instructional leadership, efficiency-focused AI implementation may privilege dominant educational discourses while marginalizing pedagogical approaches serving Indigenous students, newcomers, and students with disabilities (Benjamin, 2019). Algorithmic perpetuation of historical discrimination in managerial functions, particularly hiring and evaluation, demonstrates how technologies designed to eliminate bias instead systematize and amplify existing inequities (Ajunwa, 2021; Buolamwini and Gebru, 2018). Strategic planning AI applications risk reproducing colonial frameworks unless explicitly designed to challenge them, while relational AI use may disproportionately affect marginalized communities already skeptical about institutional responsiveness.

Extensive use of AI across leadership dimensions raises questions about long-term cognitive effects on administrators who increasingly delegate analytical thinking, communication crafting, and strategic planning to AI systems. They may experience attenuation of the critical thinking, emotional intelligence, and cultural competence essential for effective educational leadership (Gerlich, 2025). This cognitive offloading phenomenon has profound implications for administrative preparation programs, which must balance teaching AI literacy with preserving the human capabilities that distinguish educational leaders from mere managers.

In contexts where AI policy remains sparse and fragmented, this study's analysis supports implementing a hybrid governance model that combines top-down regulatory oversight with participatory decision-making between administrators and multiple stakeholders. This approach, aligned with the SAFE-T Framework (Stakeholder-Aligned Fairness, Ethics, Transparency in AI-Education), ensures adaptability while maintaining ethical guardrails (Umoke et al., 2025). The agile governance model is particularly relevant to educational contexts, where complex solutions require close collaboration with end users and continuous adaptation as evidence emerges (Rigby et al., 2018).

Bush's (2007) warning that externally imposed innovations lacking authentic ownership are likely to fail becomes particularly relevant in AI implementation contexts. Our findings suggest that effective AI integration requires distributed leadership approaches that engage teachers, support staff, and community members in defining appropriate technological boundaries rather than imposing efficiency-focused mandates that may undermine relational foundations of effective schools. This participatory approach becomes essential for maintaining democratic engagement and institutional trust, particularly in diverse communities where AI-mediated interactions may be interpreted as evidence of administrative disregard for stakeholder perspectives.

These implications collectively underscore that without deliberate intervention grounded in educational leadership theory, AI adoption risks transforming school administrators from educational leaders into efficiency-focused managers, fundamentally altering the nature of secondary school governance.

This scoping review addresses what secondary school administrators require for ethical AI integration, while acknowledging several methodological and conceptual limitations that affect the scope and generalizability of findings.

The study's literature search was limited to English-language resources, introducing inherent bias toward Western-centric research and potentially excluding valuable perspectives from non-English-speaking educational contexts where AI integration may follow different patterns. In the rapidly evolving field of AI in education, the study excluded sources older than 7 years to maintain currency; however, some sources may become dated by the time of publication, given the accelerated pace of technological advancement. The relatively small corpus of 21 sources reflects the nascent state of research specifically addressing AI in secondary school administration but limits the scope of the study's analysis.

This study relies on secondary data collection rather than empirical investigation. While appropriate for exploring emerging fields (Grant and Booth, 2009), this prevents direct examination of administrator experiences and implementation outcomes. Primary research involving interviews with secondary principals actively using AI for SIPs, IEPs, and other administrative functions would provide valuable empirical validation of our theoretical framework.

Our analytical focus on four leadership dimensions constrains examination of other potentially relevant frameworks such as transformational leadership (Bass and Riggio, 2006) as examined by Hou et al. (2024) in the Chinese K-12 context or distributed leadership (Spillane, 2006). The decision to foreground structural and strategic AI applications means that critical issues, including algorithmic bias, accessibility, and digital equity—extensively examined in broader educational technology literature (Benjamin, 2019; Noble, 2018; Selwyn, 2019)—receive less detailed treatment than their importance warrants.

Findings may have limited transferability beyond the Ontario/Canadian context emphasized in the study's analysis, as AI policy landscapes, educational governance structures, and cultural considerations vary significantly across jurisdictions. The predominantly North American source base may not adequately represent AI implementation patterns in other educational systems with different leadership structures, regulatory frameworks, or technological infrastructures.

The rapid evolution of AI capabilities means that current concerns about detection, authenticity, and technical limitations may prove obsolete as technology advances. The study captures AI integration patterns at a specific moment in technological development, potentially limiting long-term relevance of specific findings while preserving broader theoretical insights about leadership dimension interactions.

The investigators' position as educational researchers working in the Ontario postsecondary system may reflect unconscious bias toward particular theoretical frameworks and policy priorities. Although attempts were made to maintain analytical objectivity, interpretation of sources and emphasis on equity concerns may reflect critical educational scholarship traditions that may not represent all perspectives on AI integration in educational leadership.

This study's analysis of AI integration across four leadership dimensions reveals several critical areas requiring empirical investigation to advance understanding of AI's role in secondary school administration.

Future research should examine long-term cognitive effects of administrators’ reliance on AI systems. Research on cognitive offloading suggests sustained use of external tools may reduce individuals’ capacity to perform thinking tasks unaided (Gerlich, 2025; Heersmink, 2015, 2016). Leadership functions ultimately depend on human judgment, yet whether similar attenuation occurs among school administrators delegating analytical or planning work to AI remains an open question. This potential attenuation raises questions for AI design and PD programs, particularly given Shapiro's (2006) hope that ethical/relational leadership would help educational administration programs become “relevant, critical and thought-provoking” learning communities that “foster moral decision making, and hopefully, in so doing, help to develop authentic and inspiring educational leaders for the future” (p. 7). Longitudinal studies tracking administrator cognitive capabilities before and after AI implementation could provide crucial insights for professional preparation programs.

Primary research involving interviews and case studies with secondary school principals actively using AI for SIPs, IEPs, and daily administrative functions would provide essential empirical validation of our theoretical framework. Such studies should examine implementation processes, stakeholder reactions, and educational outcomes across diverse school contexts to understand how theoretical concerns translate into practical realities. Similarly, research on how PD shapes educator perceptions, as demonstrated by Kurtz et al. (2022) finding shifts in teacher thinking after AI training, could inform administrator preparation programs.

Given the study's finding that principals in high-poverty schools receive less AI implementation guidance (Kaufman et al., 2025), research should investigate how socioeconomic factors affect AI integration patterns and outcomes. Studies examining differential impacts on Indigenous, newcomer, and racialized communities would address critical equity gaps in current AI education literature (Benjamin, 2019; TRCC, 2015).

The study's findings suggest potential trust implications when stakeholders discover AI-mediated communications, yet empirical evidence remains limited (Kumar and McGray, 2025; Riedl, 2019). Research should examine how parents, teachers, and community members respond to AI use in administrative functions, particularly in diverse communities where institutional trust may already be fragile.

Future research should look beyond Western-centric contexts to examine AI integration patterns across different educational systems, governance structures, and cultural frameworks. Comparative studies could illuminate whether concerns identified in Ontario contexts apply universally or reflect specific regional dynamics.

Given the policy vacuum identified in our analysis (CTF, 2025; UNESCO, 2024), research should evaluate different governance approaches to AI integration. Studies comparing top-down regulatory frameworks with participatory decision-making models could inform policy development while examining the effectiveness of hybrid approaches advocated in our analysis (Umoke et al., 2025; Williamson and Eynon, 2020).

Extended studies tracking AI integration effects on teaching and learning outcomes remain essential. As Banerji (2025, para. 23) noted regarding the RAND findings, we do not “know nearly enough” about whether AI improves educational outcomes. Multi-year studies examining student achievement, teacher satisfaction, and institutional effectiveness could address this fundamental knowledge gap.

This study analyzed administrative functions across four established leadership dimensions (Bush, 2007; Davies and Davies, 2006; Hallinger, 2005; Shapiro, 2006) to identify what is required to prepare secondary school administrators to lead the ethical and responsible integration of AI. The research revealed a complex landscape where AI integration intersects differently with instructional, managerial, strategic, and relational leadership functions, creating both unprecedented opportunities and significant risks for educational institutions.

The most significant finding is the misalignment between current AI deployment patterns and appropriate applications. While administrators predominantly use AI for communication drafting and stakeholder engagement—functions requiring authentic human connection and cultural sensitivity (Northouse, 2019; Shapiro, 2006)—they underutilize AI's analytical capabilities in data processing and resource allocation, where algorithmic support may prove more appropriate (Kaufman et al., 2025). This inversion suggests secondary school leaders navigate AI adoption without sufficient theoretical grounding or policy guidance, potentially undermining the relational foundations distinguishing effective educational leadership from administrative management (Bush, 2007).

Equity concerns permeate AI applications across all leadership dimensions, demonstrating how apparently neutral technologies systematize and amplify existing educational inequalities (Benjamin, 2019; Noble, 2018). From instructional functions that may marginalize Indigenous pedagogies (TRCC, 2015) to managerial applications perpetuating hiring discrimination (Ajunwa, 2021; Buolamwini and Gebru, 2018), AI systems trained on historical data reproduce rather than challenge systemic inequities. This pattern proves particularly concerning in diverse Ontario communities where effective leadership requires cultural responsiveness and stakeholder engagement that cannot be algorithmically mediated (Davies and Davies, 2006; Leithwood et al., 2004).

The cognitive implications of extensive AI reliance emerge as a critical area requiring immediate attention. Administrators who delegate analytical thinking, strategic planning, and communication crafting to AI systems may experience attenuation of emotional intelligence, cultural competence, and critical thinking capabilities essential for educational leadership (Gerlich, 2025; Goleman, 1995). This cognitive offloading phenomenon has profound implications for administrative preparation programs, which must balance technological literacy with preservation of distinctly human leadership capabilities (Sergiovanni, 1996; Shapiro, 2006).

Our analysis supports implementing hybrid governance frameworks combining regulatory oversight with participatory decision-making, enabling schools to adapt AI applications while maintaining ethical guardrails (Umoke et al., 2025; Williamson and Eynon, 2020). The agile governance model proves particularly relevant for educational contexts requiring close collaboration with diverse stakeholders and continuous adaptation as evidence emerges (Rigby et al., 2018). Such approaches become essential for preserving democratic participation and institutional trust (Spillane, 2006), particularly in communities where AI-mediated interactions may signal administrative disregard for stakeholder perspectives (Riedl, 2019).

Looking forward, the rapid evolution of AI capabilities suggests that current concerns about authenticity and detection may prove transitional as technological sophistication increases and social norms adapt (Kumar and McGray, 2025). However, the fundamental tension between efficiency-focused AI implementation and relationship-based educational leadership will likely persist (Greenleaf, 2002; Hallinger, 2005; Leithwood et al., 2004), requiring ongoing vigilance and adaptive governance approaches.

This research contributes to educational leadership scholarship by providing the first systematic examination of AI integration across established leadership dimensions in secondary school contexts. As AI adoption accelerates while policy development lags (CTF, 2025; UNESCO, 2024), secondary school administrators require theoretical frameworks and practical guidance for navigating technological integration that serves rather than undermines educational equity (Edmonds, 1979). The four-dimensional analysis presented here offers such a framework while identifying critical areas for future research, particularly regarding long-term cognitive effects and empirical evaluation of AI's impact on teaching and learning outcomes (Banerji, 2025).

Ultimately, the responsible integration of AI in secondary school leadership demands more than technological competence; it requires wisdom, cultural sensitivity, and unwavering commitment to educational equity (Davies and Davies, 2006; Shapiro, 2006). As administrators shape the conditions under which both educators and students encounter AI technologies, their decisions will determine whether AI serves as a tool for educational enhancement or becomes a source of institutional fragmentation and inequity (Williamson and Eynon, 2020). The stakes could not be higher for the future of public education.

GenAI tool (Claude AI, Anthropic and Grammarly) were used to improve language clarity. No content generation, research design, or analysis was conducted using AI tools. All intellectual contributions represent original work by the authors.

The authors would like to thank the anonymous reviewers and the editor for their constructive feedback and suggestions, which helped improve the clarity of the manuscript.