Universities race to adopt AI, but teaching capacity lags behind

New evidence suggests that the speed of AI adoption in higher education is masking deep structural weaknesses that threaten the sustainability and educational value of AI-driven teaching. A comprehensive regional study shows that while AI is widely recognized as essential to the future of higher education, its integration into everyday teaching remains uneven, fragmented, and misaligned with the needs of educators and learners.

The study, titled Empowering Teaching in Higher Education Through Artificial Intelligence: A Multidimensional Exploration, published in the journal Sustainability, is based on large-scale survey data from Zhejiang Province in China. It analyzes institutional strategies, faculty readiness, and student demand across 81 universities, more than 4,000 instructors, and over 24,000 students, revealing a system racing ahead on policy and infrastructure while struggling to translate ambition into effective, equitable, and pedagogically sound teaching practice.

Universities push AI strategies, but disciplinary imbalances persist

At the institutional level, the study finds that AI has become firmly embedded in higher education planning. Nearly all surveyed universities have already launched or are preparing AI-related education initiatives, reflecting strong policy momentum and alignment with national digital transformation goals. This widespread institutional commitment signals that AI is no longer treated as an optional enhancement but as a core component of future talent development.

However, beneath this apparent progress lies a pronounced imbalance in how AI is integrated across academic disciplines. AI-related courses are heavily concentrated in science, engineering, and technology fields, while areas such as agriculture and medicine receive minimal attention. This skewed distribution suggests that universities are prioritizing disciplines with existing technical capacity and clearer industry demand, leaving other fields behind despite their growing need for AI-enabled innovation.

The study also highlights a mismatch between infrastructure investment and pedagogical development. Many universities have prioritized the construction of AI laboratories, intelligent teaching environments, and experimental facilities. While these investments demonstrate commitment to modernization, they are not matched by comparable progress in curriculum design or teaching materials. AI textbooks, structured course content, and coherent teaching frameworks lag significantly behind hardware deployment, creating a situation where facilities exist without sufficient pedagogical scaffolding to support meaningful learning.

Practical training platforms and industry collaboration present another area of partial progress. Some universities have established AI learning centers, collaborative innovation hubs, and practice bases in partnership with enterprises. Yet coverage remains limited, and many initiatives are still in planning rather than operation. The result is a system where policy frameworks and infrastructure expansion move faster than the development of sustainable teaching ecosystems that connect universities, industry, and learners.

Taken together, the institutional findings point to a pattern of infrastructure-first development. Universities are building the visible components of AI integration while struggling to align disciplines, curricula, and practice-oriented learning at the same pace. The study warns that without corrective action, these imbalances risk entrenching inequality between disciplines and limiting the educational impact of AI investments.

Teachers Support AI in Principle but Lack Training and Practical Pathways

Faculty attitudes toward AI emerge as broadly positive, yet constrained by limited opportunity and support. Most instructors report basic familiarity with AI concepts and express strong agreement that students should master AI-related knowledge. Support for introducing AI into teaching is similarly high, indicating that resistance to technology is not the primary barrier to adoption.

Despite this willingness, the study identifies a pronounced gap between teachers’ understanding of AI and their ability to apply it in practice. Formal training opportunities are scarce. Only a small share of faculty have participated in on-campus AI training programs, while the vast majority report no structured professional development in this area. At the same time, demand for external training is high, suggesting that instructors recognize the need to upskill but lack accessible institutional pathways to do so.

Industry experience represents an even greater bottleneck. Very few instructors have engaged in extended AI-related practice within enterprises or applied research settings. This lack of real-world exposure leaves many teachers with theoretical awareness but limited insight into how AI tools function in professional contexts. As a result, their ability to integrate AI meaningfully into teaching remains constrained, reinforcing a cycle where AI education stays abstract rather than applied.

Participation in AI curriculum development and teaching reform is similarly limited. Only a minority of faculty are directly involved in designing AI-related courses or adjusting teaching methods to reflect technological change. Many instructors remain observers rather than contributors to curriculum innovation, often due to insufficient resources, unclear institutional incentives, or lack of confidence in applying AI pedagogically.

Ethics and academic integrity present another critical challenge. While most teachers express concern about AI’s potential impact on academic honesty, assessment fairness, and research integrity, their understanding of formal AI ethics guidelines is weak. This disconnect creates risk. Without clear training and governance frameworks, instructors may struggle to balance innovation with responsible use, particularly as generative AI tools become more accessible to students.

The study concludes that faculty challenges are structural rather than attitudinal. Teachers are not resisting AI but are constrained by limited training, weak industry links, and insufficient institutional support. Addressing these gaps requires systemic reform, including integrated training programs, recognition of industry experience in career progression, and stronger alignment between teaching workloads and professional development opportunities.

Students Want Practical AI Skills, but Education Supply Falls Short

From the student perspective, the study reveals strong enthusiasm for AI paired with frustration over limited access. More than half of surveyed students express high interest in AI, viewing it as essential for career development, technical relevance, and long-term employability. This interest cuts across disciplines, signaling that AI literacy is no longer seen as the domain of computer science alone.

Despite this demand, only a minority of students have enrolled in AI-related courses. The gap is not driven by lack of motivation but by insufficient educational supply. Many universities offer limited AI course options, particularly outside science and engineering fields, leaving students unable to pursue structured learning even when interest is high.

Students’ preferences further underscore the mismatch between supply and demand. Learners prioritize foundational computing skills, commonly used AI tools, applied case studies, and hands-on experimentation. They are less interested in purely theoretical instruction and more focused on practical competencies that can be transferred to real-world contexts. This emphasis reflects broader shifts in labor markets, where applied digital skills are increasingly valued.

Learning pathways also reveal structural constraints. Most students rely primarily on university-provided courses for AI knowledge, with online self-study serving as a supplement rather than a replacement. This reliance places significant responsibility on institutions to expand and diversify AI offerings. When universities fail to meet this demand, students’ learning trajectories stall, and interest risks dissipating over time.

The study also highlights students’ preference for experiential learning. Practical experiments, project-based work, and real-world applications are viewed as the most effective ways to master AI concepts. Yet such formats remain underdeveloped in many institutions due to limited infrastructure integration, insufficient teacher training, and weak industry collaboration.

Overall, the student findings reinforce the conclusion that AI education is constrained by supply-side limitations. Interest is high, expectations are clear, but institutional capacity has not yet caught up with demand. Without expanded course availability, interdisciplinary integration, and practice-oriented teaching models, universities risk failing to convert student enthusiasm into meaningful skill development.