From digital transformation to intelligent classrooms: artificial intelligence, adaptive leadership, and service delivery in global higher education—a systematic literature review

Synthesis: Nyamboga asks how artificial intelligence is reshaping service delivery in global higher education, and what decides whether that reshaping succeeds. The review synthesizes 155 peer-reviewed studies published between 2019 and 2024 and organizes them across five domains: AI integration, adaptive leadership, infrastructure readiness, service delivery transformation, and barriers to sustainability. The evidence points in one direction: AI in higher education is associated with improvements in teaching, learning, and administrative efficiency, and two of the five thematic domains, AI integration and digital infrastructure readiness, earn high certainty ratings under a GRADE-adapted framework. But the synthesis is candid about its own shape. Adaptive and personalized systems dominate the application list, adaptive leadership is the most studied leadership model, and the papers cluster in high-income regions. The organizing claim is that technological capability, leadership, and institutional readiness must move together, and that installing tools while neglecting AI Governance, infrastructure, and leadership capacity yields fragmented adoption rather than transformation.

Key Findings

1. Of 155 included studies, AI integration in higher education and digital infrastructure readiness were rated high certainty, adaptive leadership moderate, AI-enabled service delivery outcomes moderate to high, and ethical governance and sustainability low.
2. Learning analytics was the most common AI application, appearing in 46 studies (29.7%), ahead of chatbots and virtual assistants at 31 (20.0%) and predictive analytics at 29 (18.7%).
3. Adaptive leadership was the most studied leadership model at 61 studies (39.4%), against distributed leadership at 44 (28.4%), transformational leadership at 32 (20.6%) and centralised leadership at 18 (11.6%), noted for fast implementation but limited flexibility.
4. Coverage is geographically narrow: North America contributed 48 studies (31.0%), Europe 42 (27.1%) and Asia 38 (24.5%), while Africa contributed 15 (9.7%), South America 7 (4.5%) and Oceania 5 (3.2%).
5. Infrastructure readiness split across the corpus: high readiness in 59 studies (38.1%), moderate in 53 (34.2%) and low in 43 (27.7%), where low readiness was tied to limited AI adoption and fragmented systems.
6. Reporting skew is visible in the review's own bias assessment: 117 studies (75.5%) reported successful implementation outcomes, 124 (80.0%) covered short-term impact only, and just 23 (14.8%) reported on ethics, equity or governance failures.

How the review was done

The study is a systematic literature review conducted following PRISMA guidelines. Peer-reviewed studies published in English between 2019 and 2024 were retrieved from Scopus, Web of Science, IEEE Xplore, SpringerLink, ScienceDirect, and Google Scholar, selected for relevance to AI, adaptive leadership, infrastructure readiness, and service delivery in higher education. Following eligibility assessment, 155 studies met all inclusion criteria. Data were extracted using a structured framework and analyzed through thematic synthesis. Quality was graded with adapted CASP and JBI-informed criteria: 76 studies (49.0%) carried low risk of bias, 56 (36.1%) moderate and 23 (14.9%) high. The synthesis is grounded in an integrated conceptual framework drawing on the Technology-Organization-Environment (TOE) framework, Socio-Technical Systems theory, and Adaptive Leadership theory.

What the evidence says about AI and service delivery

The reviewed literature associates AI with improvements in teaching, learning, and administrative efficiency through learning analytics, intelligent tutoring systems, chatbots, and predictive modeling. Intelligent tutoring systems appeared in 27 studies (17.4%). Named examples include learning analytics dashboards at Arizona State University, intelligent tutoring in engineering and computer science at Tsinghua University, controlled generative AI for academic writing and curriculum development at the University of Oxford and University College London, and chatbots handling student enquiries at the University of Melbourne. Service delivery outcomes skewed positive, with high improvement in 64 studies (41.3%), moderate improvement in 57 (36.8%) and low improvement in 34 (21.9%). The same literature consistently flags academic integrity, authorship, and assessment validity as unresolved.

Leadership and infrastructure readiness decide the outcome

Adaptive leadership emerged as a key enabler of AI adoption, fostering innovation, managing resistance to change, and aligning institutional strategy with digital transformation goals. Institutional type mattered: public universities hosted 78 studies (50.3%), private universities 41 (26.5%) with faster adoption but uneven scaling, research-intensive universities 24 (15.5%) and emerging digital campuses 12 (7.7%). Infrastructure readiness tracked outcomes closely, with high readiness in 59 studies (38.1%) showing full AI ecosystem integration and low readiness in 43 (27.7%) showing limited adoption and fragmented systems, with disparities between high-income and low-income regions.

What this means for practice

• Instructors. Treat AI as a complement to teaching rather than a replacement: the reviewed studies associate learning analytics and intelligent tutoring with earlier identification of students needing support, but also flag academic integrity and assessment validity as open problems.
• Instructors. Check whether your institution's infrastructure can sustain a tool before adopting it; low readiness was tied to limited, fragmented systems.
• Administrators. Fund adaptive leadership capacity alongside procurement, since adaptive leadership was the most studied enabler of AI adoption success (61 studies, 39.4%).
• Administrators. Build ethical governance and sustainability review into AI programs now, because ethical governance and sustainability drew the review's lowest certainty rating and rest on limited longitudinal evidence.
• Institutions and policymakers. Commission longitudinal and equity-focused evaluation: only 31 studies (20.0%) were longitudinal, and only 23 (14.8%) reported on ethics, equity or governance failures.

Limitations

• The evidence base is geographically skewed: 128 studies (82.6%) came from North America, Europe and Asia, and only 27 (17.4%) from Africa and South America, limiting generalizability to low- and middle-income contexts.
• Positive-outcome reporting bias is documented within the review: 117 studies (75.5%) reported successful implementation, while only 38 (24.5%) reported partial or failed implementation.
• Short-term evidence dominates: 124 studies (80.0%) examined immediate outcomes and only 31 (20.0%) were longitudinal, leaving long-term sustainability under-evidenced.
• The overall certainty of evidence was assessed as moderate, with leadership findings rated moderate because of contextual variability across regions.

Citation

Nyamboga, T. O. (2026). From digital transformation to intelligent classrooms: artificial intelligence, adaptive leadership, and service delivery in global higher education—a systematic literature review. Frontiers in Education, 11:1899369.