The quantitative study examined staff perceptions at the University of Ilorin, Kwara State University, and Al-Hikmah University in Kwara State, Nigeria. The institutions were selected purposively because of their relevance to AI applications in teaching, data-informed decision-making, administration, and institutional management. Respondents were selected through a multistage procedure that stratified academic and administrative personnel, allocated sample places according to each institution’s staff population, and randomly selected eligible individuals within the resulting strata. Of 500 questionnaires targeted, 450 were usable, producing a 90.0% response rate. The final sample included 318 respondents from the University of Ilorin, 111 from Kwara State University, and 21 from Al-Hikmah University, with 225 academic and 225 administrative staff.

Because questionnaires were completed at a single point in time, the results show statistical associations based on respondents’ perceptions rather than demonstrating that AI adoption caused institutional change. The questionnaire contained 25 items measured on a five-point Likert scale, from 1 for strongly disagree to 5 for strongly agree. Administrative effectiveness was represented by six items covering AI chatbots for routine enquiries, automated scheduling, document processing, predictive maintenance, admissions and records management, and campus security and facilities. Pedagogical enhancement included adaptive learning paths, intelligent tutoring, automated grading, teaching feedback, personalised content, and simulations. Data-driven insight covered enrolment forecasting, dashboards, student-risk indicators, strategic planning, and resource allocation.

The dependent construct, management and digitalization of higher institutions, combined perceptions of change management, resource management, effective leadership, and governance. The researchers adapted the questionnaire from established measurement scales, revised wording for the study context, and obtained content validation from two experts holding doctorates in educational measurement and educational management. A pilot study with 40 respondents produced a Cronbach’s alpha of 0.780. In the main measurement model, all four constructs had Cronbach’s alpha and composite reliability values above 0.70, while their average variance extracted values exceeded 0.50. These results indicated satisfactory internal consistency and convergent validity for the questionnaire’s four constructs.

Most indicator loadings were also above the preferred 0.70 threshold, although the item concerning whether AI prediction models improve enrolment forecasting had a lower loading of 0.534. The researchers retained that indicator because the overall data-driven insight construct retained acceptable reliability and convergent validity and the item helped cover a relevant predictive-analytics dimension. They also reported heterotrait–monotrait ratios from 0.725 to 0.832, all below the conservative 0.85 threshold. The highest value, however, was 0.832 for administrative effectiveness and management and digitalization, indicating substantial similarity between the two constructs and requiring cautious interpretation of the strongest path.

Under the Fornell–Larcker criterion, the correlation between administrative effectiveness and management and digitalization was 0.803, exceeding the square root of the average variance extracted for those constructs, 0.751 and 0.776 respectively. The authors identified this as a potential discriminant-validity concern and said the corrected values should be verified against the original SmartPLS output. The constructs nevertheless had a stated conceptual distinction: administrative effectiveness concerned the efficiency of particular AI-enabled administrative functions, whereas management and digitalization represented broader organisational outcomes involving change management, resource management, leadership, and governance. That distinction may help explain part of the overlap, but the statistical concern remains a limitation on how confidently the strongest association can be interpreted as separation between predictor and outcome.

The contrast among the three paths suggests that the visible effects of AI within a university may emerge first in routine operational work. Chatbots, scheduling systems, automated document handling, admissions support, and facilities monitoring can directly alter staff workflows and service delivery, making their contribution easier to recognise in perceptions of institutional digitalization. The model supports a positive association between these administrative applications and the broader institutional outcome, but it does not show that automation alone guarantees comprehensive transformation. It also did not measure changes in student achievement, graduation rates, cost, employment, or other direct performance indicators. Within this sample, staff perceived a stronger institutional-level relationship for AI-enhanced administration than for AI-enhanced teaching or analytical capability considered on its own.

Pedagogical applications nevertheless showed a supported positive association with management and digitalization. Their coefficient was substantially smaller than that for administrative effectiveness, but the model links tools used for personalised learning, automated assessment, feedback, and practical simulations with wider digital processes. The study also notes that such applications require supporting systems for course delivery, assessment, curriculum management, staff development, and quality assurance. The authors caution that the size of the relationship may depend on technological infrastructure, staff digital competence, institutional readiness, leadership, governance, and effective change management. The findings therefore support treating pedagogical AI as a complementary pathway to digital transformation rather than either dismissing it or assuming that it has the same magnitude of association as administrative automation.

The non-significant result for data-driven insight is equally important. The coefficient of 0.050 was close to zero, indicating that the perceived availability and usefulness of AI-generated analytical information was not directly associated with management and digitalization in this model. The authors interpret the result as potentially reflecting a gap between generating insights and acting on them. They identify data quality, fragmented information systems, weak data governance, limited analytical capacity, institutional readiness, staff expertise, and the capacity to embed outputs in strategic decisions as possible contextual constraints. These explanations were not directly tested as separate causal mechanisms by the survey. The result also does not establish that analytics are unimportant; it shows only that their direct relationship with the measured institutional outcome was not statistically supported in this sample and model.

The findings were interpreted through a Technology–Organization–Environment lens, under which technological innovation is shaped by the technology itself, an organisation’s internal capacity, and its external environment. The empirical model tested three technology-related AI domains and one institutional outcome rather than independently measuring all three components of the framework. Government policy, funding, external stakeholders, competition, and wider societal expectations were acknowledged as relevant contextual conditions but were not operationalised. Other limitations include restriction to three institutions in one Nigerian state, reliance on self-reported perceptions, the aggregation of academic and administrative staff who may encounter different tools in different roles, and the absence of longitudinal or independent measures of institutional transformation. The results therefore should not be treated as automatically generalisable to universities in other regions or institutions at different stages of AI adoption.

Within those limits, the results establish a clear ordering of the direct associations observed: administrative effectiveness had the strongest supported relationship with institutional management and digitalization, pedagogical enhancement had a smaller supported relationship, and data-driven insight had no statistically significant direct relationship. The authors recommend an integrated institutional strategy that improves administrative workflows, supports AI-enabled teaching, and builds the data governance, system interoperability, staff competence, leadership, and change-management capacity needed to convert analytical output into action. Their findings do not establish that a particular AI product will produce better learning or institutional performance. Instead, they identify where staff perceived stronger links with digitalization and where further evidence is needed, including analyses of academic and administrative personnel separately, tests of measurement invariance across roles, and longitudinal research that follows institutions over time.

Subject of Research: Associations of AI-enabled administration, teaching, and analytics with digitalization at three Nigerian universities

Article Title: Artificial intelligence-enabled practices and management and digitalization of higher institutions in a changing world

Article References: Abdullahi, N. J. K., Muhinat Bolanle, B., Isiaka, B. A., Zubair, O. M., Mubarak, M. S., Ola-Alani, E. K., & Yahaya, L. A. (2026). Artificial intelligence-enabled practices and management and digitalization of higher institutions in a changing world. Frontiers in Education, 11, Article 1950231. https://doi.org/10.3389/feduc.2026.1950231

Image Credits: AI Generated

DOI: 10.3389/feduc.2026.1950231

Keywords: artificial intelligence, higher education, university digitalization, administrative automation, pedagogical enhancement, data-driven insight, structural equation modelling, institutional management, Nigeria, survey research, Artificial, intelligence-enabled