Artificial Intelligence Integration in STEM Education: An Evidence-Constrained Systems Readiness Assessment

Artificial intelligence (AI) is entering university STEM education at a pace that may exceed the capacity of curricula, professional development, infrastructure, and governance to adapt. This study develops an evidence-constrained systems-readiness framework based on two parallel surveys conducted in Albania in January 2026, comprising 664 student responses (approximately 282 substantially complete entries) and 92 academic staff responses from public and private higher education institutions. As respondent-level data were unavailable for this secondary analysis, the investigation relied on statistics that could be defensibly reconstructed from aggregate reports, including chi-square tests, Holm-adjusted one-sample tests, Welch comparisons, effect sizes, Wilson confidence intervals, denominator and covariance sensitivity analyses, and report-level qualitative triangulation. The framework integrates reported curriculum maturity, responsible AI practices, academic capacity, infrastructure, governance, student preparedness, and perceived employability alignment, while distinguishing perceptions from demonstrated competence and independently verified institutional performance. Student and staff distributions across the four levels of AI integration showed no statistically detectable difference (χ²(3) = 1.28, p = 0.734, Cramér's V = 0.069), with reported curriculum maturity scores of 34.1/100 and 30.0/100, respectively. Students reported a substantial gap between the availability of project-based learning opportunities and the perceived effectiveness of practical learning (Hedges' g ≈ 0.80). Academic staff rated existing infrastructure below the scale midpoint but projected institutional readiness over the following three years above the midpoint. The perceived demand–preparedness gap remained positive under assumed within-respondent correlations ranging from 0 to 0.6. The study contributes an implementation-agnostic and reproducible systems-readiness framework that identifies asynchronous institutional constraints, preserves uncertainty arising from aggregate data, and translates perception-based findings into testable pathways for AI integration in university STEM education.