Integrating artificial intelligence in primary education: empirical insights of school leadership perspectives Open Access

The integration of artificial intelligence (AI) into school leadership and management represents one of the most significant developments in contemporary education. This study focuses on the role of school leadership in the adoption and integration of AI in primary schools. Ιn this context, integration refers to the process by which school leaders adopt, use, and manage AI technologies within their professional practices.

Predictive analytics and real-time data help school leaders identify challenges, make informed decisions, and plan strategically (Fusarelli and Fusarelli, 2024; Wang, 2021). Adopting and integrating AI is primarily the responsibility of school leaders (Bixler and Ceballos, 2025), who set strategic direction, allocate resources, and guide professional development (Tyson and Sauers, 2021). Τheir beliefs and attitudes toward AI, influence how effectively it is integrated (Marrone et al., 2025).

While literature on leadership and AI is growing worldwide, little research explores how leadership factors, mainly attitudes, self-efficacy, and decision-making, influence the integration of AI in schools. Most studies focus on teachers΄ competencies, leaving a gap concerning principals΄ beliefs and efficacy. This study addresses that gap by investigating four dimensions (attitudes, self-efficacy, perceived usefulness, and challenges) in Greek primary schools.

The Greek context provides a valuable background. After the COVID-19 pandemic Greece emphasized digital literacy by aligning its educational plan with the European Union (EU)’s Digital Education Action Plan 2021–2027 (European Commission, 2020). AI readiness and digital literacy are promoted by initiatives like the AI4EDU project, the ahedd Digital Innovation Hub, and Innovation Centers (Kanellopoulou et al., 2025; Vacalopoulou et al., 2024; Vagelatos et al., 2025). Despite these efforts, there is still little data on how Greek school leaders engage with AI. Reports reveal limited systematic training, unequal digital infrastructure, and fragmented policy guidance (Αnysiadou and Gkliati, 2025; Kalogeratos and Pierrakeas, 2023). Such restrictions underline the need for a closer look at how school leaders perceive and navigate AI integration.

This study focuses on three dimensions (attitudes, self-efficacy, and perceived usefulness) and the challenges school leaders face, to examine how leadership factors influence AI integration in Greek primary schools.

Accordingly, the study addresses the following research questions:

These research questions aim to reveal the main factors shaping leaders΄ engagement with AI in Greece. The limited scholarly work in this context justifies this case study, offering empirical evidence on how leadership beliefs drive technological innovation in education and providing practical insights for policy and professional development.

The structure of the paper is as follows: The next section presents the literature review, followed by the theoretical framework of the study. The next section details the research methodology and data analysis procedures. The findings are then presented and discussed in relation to previous research. Finally, the paper concludes with the key implications, limitations, and directions for future research.

This paper presents literature on ΑΙ and school leadership, under the following four thematic areas: (1) the evolving role of school leaders in AI integration, (2) AI-assisted decision-making and ethical considerations, (3) AI self-efficacy as important leadership competency, and (4) challenges to AI integration. These themes reflect the conceptual base developed to understand how school leaders perceive and manage AI in education.

School leaders are expected to include AI into administration and pedagogy, while promoting innovation and positive attitudes toward technology (Karakose, 2024; Van Quaquebeke and Gerpott, 2023).

But, Tyson and Sauers (2021) point out, that the main challenge is not the lack of adequate technological infrastructure, but school leaders΄ poor comprehension of technology΄s value. This was evident during the COVID-19 pandemic, which accelerated digital transformation, revealed readiness gaps, and changed leaders΄ perceptions toward professional development and digital competence (Constantia et al., 2023; Kafa, 2025; Yang et al., 2025).

AI may enhance the quality of school leadership by facilitating data-driven decision-making. Machine learning and real-time data help leaders make decisions (Wang, 2021). However, depending too much on algorithms could impair professional judgment (Igbokwe, 2024).

Therefore, human oversight remains indispensable. AI is not an expert leader or manager but rather a participant in collaborative decision-making (Dai et al., 2025). As Igbokwe (2024) further argues a culture of openness, reflection, and shared accountability is necessary for sustainable AI use.

The expectations for educational leadership are evolving as a result of the increasing use of AI. School leaders are expected to adapt to digital tools that influence management, communication, and decision-making. AI self-efficacy leaders΄ confidence about comprehending and using AI is central to appropriate integration. Professional learning, resource availability, and institutional support enhance self-efficacy, driving adoption (Berkovich and Eyal, 2025; Erdoğan et al., 2025).

Additional research points out that the mediating factor between support and AI usage is self-efficacy (Dai et al., 2025; Wang and Cheng, 2021). Even in the midst of ongoing difficulties, leaders who feel confident to apply AI are more likely to use it in a meaningful and responsible way. Transformational leadership can further strengthen this process through a culture of creativity, trust, and reflective practice (Arar et al., 2025; Fullan et al., 2023). Therefore, the key to moral and effective AI leadership is to enhance self-efficacy and supporting environments.

Despite growing awareness, leaders still have numerous challenges to overcome. Insufficient training, funding, and policy guidance remain major barriers (Anysiadou and Gkliati, 2025; Hales et al., 2025; Marrone et al., 2025). These constraints directly affect AI adoption and digital transformation readiness. Table 1 summarizes principals΄ perceptions of AI integration across different education systems.

The study took place in Greece, an EU member state with a highly centralized educational system. According to Raptis et al. (2020), the Ministry of Education, Religious Affairs, and Sports is responsible for establishing regulatory frameworks, creating and implementing educational policies. Due to limited strategic initiatives before the pandemic crisis, Greece has historically experienced minimal incorporation of technology in educational environments. Furthermore, school leaders were unprepared to handle the digital issues that arose during the COVID-19 epidemic (Constantia et al., 2023). However, in the post-COVID era, there is a growing movement to recognize the role that school leadership plays in the digital modernization of educational institutions (Raptis et al., 2024). Additionally, principals have embraced more collaborative and participatory leadership practices, using new technologies to transform schools (Raptis et al., 2025).

The recent strategic orientations reflected in the Ministry of Education, Religious Affairs, and Sports' annual report (2024) demonstrate a clear and strengthened institutional will to accelerate digital transformation in all Greek schools. This commitment has recently taken a more organized shape through initiatives as the Strategic Plan for Digital Transformation in Education and the National Digital Academy, both of which prioritize the incorporation of ΑΙ, robotics, and digital literacy (Ministry of Education, Religious Affairs and Sports, 2024). These initiatives are closely related with the broader Digital Transformation Strategy 2020–2025, that launched Greece's national agenda for digital modernization in public sectors, including education (Ministry of Digital Governance, 2020).

Greece's digital strategy fosters a highly technological environment by supporting innovative educational programmes that include technology like robots and ΑΙ. Lifelong learning and AI-focused training enhance employment skills, while initiatives such as Digital School II and Skills for the Future support infrastructure, teacher training, and leadership development (European Commission, 2020). In this policy context, the integration of AI in Greek schools remains at an early stage, making it a suitable setting to investigate how school leaders΄ self-efficacy and organizational support shape AI adoption.

The theoretical foundation of this study draws on three complementary perspectives: the Technology Acceptance Model (TAM), Bandura's Self-Efficacy Theory, and Ajzen's Theory of Planned Behavior (TPB). These frameworks provide a logical means for understanding the cognitive and behavioral processes that shape school leaders΄ adoption and integration of AI technologies.

The TAM (Davis, 1989) proposed two major constructs, perceived usefulness and perceived ease of use. Both determine individuals΄ attitudes and intentions toward the adoption of new technologies. Most research applies TAM to examine the adoption of technology and attitudes toward AI in education (Al-Shorman et al., 2025; Mastour et al., 2025).

A later extension of the TAM (Venkatesh and Davis, 2000) acknowledges that adoption decisions are also influenced by other external variables like organizational support, subjective norms, and contextual constraints. These variables affect behavioral intention by moderating perceived usefulness and ease of use. In this paper, TAM is expanded to include challenges as contextual moderators reflecting how environmental and institutional conditions on leaders΄ willingness to adopt AI.

In addition to TAM, the Self-Efficacy Theory explains that performance and motivation are determined by the strength of an individual΄s belief in his capability (Bandura, 1977). Regarding school leadership, AI self-efficacy could be considered as leaders΄ confidence in implementing and promoting AI tools in their schools. Leaders with high self-efficacy may collaborate with others to transform technological potential into meaningful innovation (Berkovich and Eyal, 2025).

The relationship between organizational support and adoption intent is mediated by self-efficacy, which also predicts readiness for AI integration (Wang and Chuang, 2024). Confidence and resilience are further reinforced throughout digital transitions by transformational leadership (Arar et al., 2025; Fullan et al., 2023).

Ajzen's (1991) TPB builds upon TAM by presenting the concept of Perceived Behavioral Control, which illustrates how individuals΄ perceptions of resources and external limitations impact their ability to act. Challenges like limited infrastructure, lack of expertise, or institutional rigidity serve as external restrictions that can either help or hinder behavior. This theoretical integration goes on to relate to the challenges school leaders face in integrating AI.

In this paper, the TAM, Bandura's Self-Efficacy Theory, and Ajzen's TPB jointly provide the conceptual baseline for understanding how school leaders perceive and integrate AI tools in educational settings. Therefore, the study seeks to address four main research questions, from different theoretical angles. RQ1 relates to the attitudinal component of TAM, reflecting school leaders΄ openness to AI integration. RQ2 is based on Bandura΄s Self-Efficacy Theory and complementary to perceived ease of use in TAM and explores confidence among leaders on usage of AI leading as a driver of engagement. RQ3 investigates the ways in which AI supports creativity by utilizing TAM΄s perceived usefulness dimension. In line with the extended TAM and Ajzen΄s concept, RQ4 addresses ethical, technological, and organizational challenges that might hinder adoption. These frameworks capture the interaction of internal drivers (attitudes, self-efficacy, perceived usefulness) and external moderators (challenges).

This research used a quantitative method to gather data through an online questionnaire (Creswell and Poth, 2016). A structured online questionnaire (61 items) was developed covering four sub-themes reflecting the theoretical framework: (1) attitudes toward AI, (2) self-efficacy in using AI, (3) perceived usefulness in decision-making, and (4) challenges to AI integration.

The instrument was piloted with a small group of school principals to test reliability and clarity. Participants΄ perceptions were recorded through a 5-point Likert-type scale, in which 1 corresponded to complete disagreement and 5 to complete agreement, allowing quantitative assessment of their level of agreement with each research dimension. Table 2 contains sample questions per four sub-themes, along with their theoretical background.

Convenience sampling was used, which made it possible to recruit a diverse group of school leaders from four districts in Greece (Attica, Thessalia, South Aegean, Crete) during the 2024–2025 academic year. Convenience sampling was employed to find accessible participants who were willing to share their opinions about ΑΙ. An email invitation was sent to all primary school units across the four selected districts of Greece, and the final sample comprised 262 school leaders.

The distribution of participants across the four districts was as follows: Attica (n = 72, 27.5%), Thessalia (n = 68, 26.0%), South Aegean (n = 66, 25.2%), and Crete (n = 56, 21.3%). Other demographic data collected appear in Table 3.

Despite its methodological limitations, convenience sampling provided a practical means to capture context-specific data from school leaders in Greece, including those in island regions, thereby facilitating future comparative analyses with urban research. Data collection took place from early December 2024 to mid-March 2025, and the participation was voluntary. Informed consent was obtained electronically before participation, data anonymity and confidentiality were guaranteed, and approval for the research was granted by the University of the Aegean Deontology Committee (Approval No2. 19/10//2024).

The quantitative data collected through the structured questionnaire were analyzed using descriptive and inferential statistical methods. Descriptive statistics (e.g. means, standard deviations, and frequencies) were calculated to provide an overview of the demographic characteristics of the participants and to explore the main challenges faced by school leaders when introducing AI into the educational reality. To investigate the connection between the dimensions of school leadership and the integration of ΑΙ in schools, Spearman rank correlation was used.

The predictive power of leadership dimensions, including perceived AI usefulness, leadership attitudes and opinions on AI, and self-efficacy in AI use, on the reported level of AI integration in educational settings was also investigated using multiple linear regression analysis. Cronbach's alpha was used to evaluate the scales' internal consistency. All subscales demonstrated strong reliability, with values ranging from 0.710 to 0.942 and an overall reliability coefficient of a = 0.893, indicating high construct reliability. All statistical analyses were conducted using SPSS (Version 25.0). In addition, Mann–Whitney U and Kruskal–Wallis tests were conducted to explore differences across sociodemographic groups.

The descriptive analysis of school leaders' responses showed that they have a moderate understanding and ability to use ΑΙ systems (see Table 4). Specifically, the statement about school leaders' capacity to identify possible uses of AI systems received the highest mean score (M = 2.92, T.A. = 0.88), indicating that school leaders have started to conceptually understand the value of AI in educational administration. However, they lacked specialized knowledge, as indicated by their lowest mean score (M = 2.63, SD = 0.98) in explaining the technical parts of AI, such as machine learning procedures. A mean score of 2.73 (SD = 0.91) was also given to the assertion that school administrators successfully use AI systems, underscoring the limited practical use of these technologies. The range of values (1–5) shows variation in experiences and opinions, while the median value (3.00) in all three statements indicates neutrality or caution among participants.

According to frequency analysis, 42% of school leaders agreed (scores 4–5) that they could identify potential use cases of AI systems, while only 28% reported using AI effectively in their daily work. In contrast, 35% stated that they could explain how machine learning works, indicating a serious lack of technical expertise.

Overall, these findings suggest that school leaders perceive three primary challenges: (a) limited experience with AI tools, (b) lack of technical knowledge about AI and machine learning mechanisms, and (c) insufficient targeted professional training and institutional support for integrating AI.

In addition, the integration of AI within primary education institutions in the Greek educational system shows positive and statistically significant correlations with the three key dimensions of school leadership (see Table 5).

There is a weak but significant correlation (ρ = 0.159, p < 0.05) between the integration of AI in educational organizations and perceived usefulness of AI for decision-making. This suggests that school leaders are more likely to incorporate AI into their routine decision-making processes when they possess greater awareness of its potential benefits. However, the modest strength of this relationship implies that awareness is not enough on its own. Contextual variables such as inadequate training, resource limitations, and ethical issues may also contribute to this weak association.

The relationship between self-efficacy in using AI and its integration in educational institutions is moderate and statistically significant (ρ = 0.340, p < 0.01). This suggests that school leaders' confidence in their ability to manage and control AI systems is directly related to the technology's successful adoption. Leaders who feel competent in applying AI are more likely to engage actively in integration processes and promote innovation within their schools.

The integration of AI in educational institutions is moderately and statistically significantly correlated with attitudes and opinions about AI (ρ = 0.370, p < 0.01), underscoring the importance of favorable emotional reactions and positive attitudes toward AI in encouraging its practical adoption. Positive attitudes appear to increase willingness to experiment with AI tools, support teacher development, and foster a school climate that values ethical and responsible AI use.

To further investigate the predictive ability of leadership attributes toward the degree of AI integration in educational settings, a multiple linear regression analysis was carried out. According to the results, 20% of the variance in the integration of AI in schools can be explained by the model (R² = 0.20, ΔR² = 0.19), suggesting that the chosen predictor variables have a moderate explanatory power (see Table 6).

In the regression analysis, the usefulness of AI in decision-making was not a statistically significant predictor (β = 0.00, p > 0.05) when considering the individual predictors. This suggests that the mere perception of AI's usefulness in decision-making is insufficient to predict its actual incorporation in educational contexts. In contrast, self-efficacy (β = 0.21, p < 0.001) and attitudes toward AI (β = 0.31, p < 0.001) were significant positive predictors of AI integration. These findings confirm that internal psychological factors, rather than only perceived functional value, determine AI adoption in Greek schools.

Consequently, initiatives to promote AI integration should focus on enhancing school leaders' self-efficacy and cultivating favorable views towards the technology.

Additional non-parametric analyses examined whether sociodemographic factors relate to school leaders' engagement with AI. Gender did not affect attitudes toward AI (U = 6231.00, p = 0.110), but female principals reported lower AI self-efficacy (U = 5482.50, p = 0.003) and perceived usefulness (U = 5345.50, p = 0.001) than males. Age differences also appeared: younger leaders, especially those under 29 and those aged 40–49, showed higher self-efficacy than those over 60 (H (4) = 17.67, p = 0.001). Teaching experience influenced perceived challenges, with principals in their first 10 years reporting fewer barriers than those with 11–20 years (H (3) = 8.19, p = 0.042). No effects emerged for years in leadership roles. Overall, younger, less experienced, and male leaders appear more prepared to integrate AI.

This study provides an understanding of Greek school principals΄ perceptions of and interactions with ΑΙ in their workplace. There is a gap between the theoretical discussion about AI and its practical application in schools. Although many school leaders show interest in AI, their understanding remains moderate. Limited funding, insufficient infrastructure, and lack of targeted training also contribute to this situation (Kafa, 2025; Ng et al., 2025).

AI adoption in education is more than just implementing new technology. For successful use, leader΄s beliefs and attitudes are vital (Tyson and Sauers, 2021). When school leaders feel supported, they are more likely to guide teachers confidently toward innovation. Leadership that adapts AI use to each school΄s context, as noted by Marrone et al. (2025), can bridge the gap between theory and practice. This finding is consistent with earlier research emphasizing that understanding must precede effective implementation (Aliane et al., 2023; Koukaras et al., 2025).

Principals with positive attitudes toward AI usually support staff training, allocate resources, and promote experimentation (Sposato, 2024; Tursunbayeva and Gal, 2024). Τhese efforts are strengthened by a culture that supports moral and learner-centered AI (Chan et al., 2024). Therefore, positive attitudes drive acceptance toward technological change (Pietsch and Mah, 2025). According to Wang et al. (2023), this process is often hindered by limited institutional support.

School leaders' positive attitudes and self-perceived AI abilities are more strongly associated with successful AI implementation than perceived usefulness. The sociodemographic findings show that male, younger and less experienced leaders tend to feel more confident using AI. These differences highlight the central role of self-efficacy in TAM and TPB, suggesting that confidence and perceived control strongly shape leaders΄ intentions to adopt AI. This supports digital leadership theories, such as the TAM. As demonstrated by this model, user confidence and ease of use are critical factors in the adoption of new technology (Arar et al., 2025; Davis, 1989). These results imply that school leaders' interactions with AI are significantly influenced by their personal views rather than merely an objective assessment of AI's advantages (Ghamrawi et al., 2024; Marrone et al., 2025). The integration of new technologies is often hindered by external factors such as inadequate infrastructure, institutional rigidity, and concerns over data ethics, which weaken the connection between intention and actual behavior. This explains why, despite good intentions, adoption levels tend to be modest. The findings΄reference to “significant mistrust” is not a rejection but rather a reflection of a lack of confidence in technological issues, data protection, and reliability. Reluctance can be reduced and trust enhanced with consistent technical support and practical training (Anysiadou and Gkliati, 2025).

These findings make clear how vital it is for school leaders to become more digitally literate. According to Kafa (2025), similar trends were observed in Cyprus, highlighting the necessity of continuous professional development targeted at effectively integrating digital devices into leadership roles and school operations. According to Ng et al. (2025), collaboration with public institutions, universities, and experts can also lead to sustainable professional learning. The educational policy should involve collaboration between the Ministry of Education, the private sector, and higher education institutions to sustain digital transformation.

Concerns remain about readiness within Greece's centralized educational system. Staff training and cultivating a positive outlook on AI are among these concerns (Anysiadou and Gkliati, 2025; Kalogeratos and Pierrakeas, 2023). Specifically, data analytics, algorithmic literacy, and ethics must be covered in professional development courses (Rodafinos, 2024). Although some studies show that school leaders' positive opinions on AI's role in education can have a beneficial impact, there remains significant mistrust due to a lack of understanding of AI fundamentals. Although many leaders are optimistic about AI, but a lack of understanding makes them hesitant. Future research should be conducted on how trust is built between school principals and AI tools over time and how institutional support helps them use AI in daily decision-making processes.

This study offers valuable insights into how Greek school leaders perceive and apply AI in education. In order to assist leaders, make well-informed decisions, it underlines the necessity of professional development that connects theory to practical educational objectives (Blankesteijn et al., 2024). In addition to providing technical skills, these programs need to foster data literacy, ethical reasoning, and strategic thinking competencies (Rodafinos, 2024). The subgroup findings further suggest that training may need to be differentiated, offering targeted support for older, female, or more experienced leaders who reported lower AI self-efficacy.

The development of an open and supportive school culture that encourages experimentation, teamwork, and ethical awareness is an example of a second important implication (Akyazi, 2023). This cultural change takes time and requires ongoing dialogue and gradual integration of AI into teaching and management (Kim and Lee, 2023).

From a policy perspective, integrating AI necessitates going beyond technometric approaches. It should deal with the emotional, cognitive, and organizational dimensions that shape school leadership (Arar et al., 2025). This implies that support structures such as funding, infrastructure, and ethics must reflect the daily realities of principals. Well-designed policies can enable rather than hinder AI integration.

Meanwhile, this study acknowledges certain limitations. The research uses self-reported quantitative data, which might contain personal bias or misjudgment of competence. Results cannot be generalized to different contexts, since the sample is restricted to a single national educational system. Future research should employ mixed-method or longitudinal approaches, to capture the development of AI use and its impact on school practices. Comparative studies between different educational systems may reveal how cultural, policy, and structural variations influence AI integration. Furthermore, it is crucial to understand the broader socio-cultural and political contexts driving the use of AI in Greek schools. This dynamic should be analyzed with respect to the four educational dimensions, highlighting how local contexts shape AI integration.

Future research should involve multiple stakeholders (teachers, parents, students, and policymakers) to build a holistic understanding of AI integration. This type of research will assist in the formation of collaborative structures and fair policies that encourage ethical and sustainable AI use in education. Future studies, should explicitly test the role of challenges within TAM and TPB models to clarify how external constraints affect the relationship between attitudes, self-efficacy, and behavior.

The study highlighted significant aspects of the attitudes, opinions, and preparedness of Greek school administrators toward the use of ΑΙ in learning environments. Although there is increasing theoretical recognition of AI's potential in educational practices, its actual use in the Greek educational system is still somewhat restricted. The findings showed how school leaders' technical capabilities and lack of confidence in the use and functioning of AI systems impede the effective deployment of ΑΙ.

AI usage in decision-making is not a reliable indicator of AI integration in schools, according to an analysis of leadership traits and AI integration's relationship. Instead, the intention of school leaders to integrate AI in classrooms is highly contingent on their personal beliefs, such as their positive expectations about the technology and their self-efficacy beliefs about their abilities. As the most reliable predictor, attitudes indicate the need to foster a positive climate and culture that embraces technological innovation in the classroom.

To sum up, the successful incorporation of ΑΙ into Greek education necessitates the improvement of school administrators' abilities via focused professional development, and institutional government support. These measures must be accompanied by suitable infrastructure and the reinforcement of collaboration among academic institutions, technical players, and ministries. The entire potential of AI to improve administration, instruction, and learning will only be realized by educational institutions through a concerted and inclusive approach.