Abstract
The accelerated diffusion of generative artificial intelligence in higher education has outpaced the institutional rules that define where machine assistance should end and academic judgement should begin, leaving universities without clear boundaries for substituting for teaching labour and placing the quality education and decent work agendas at risk. This study examines how Kazakhstan's educational potential relates to labour productivity and develops a strategic model that allocates functions between the instructor and artificial intelligence. The study employs annual time-series data from the Bureau of National Statistics of the Republic of Kazakhstan for 2018-2025, covering eleven labour market and human capital indicators. It applies Pearson correlation analysis to indicator levels, first differences, and growth rates, and is complemented by functional-systemic classification and expert-analytical generalisation. The results reveal that gross value added per employee correlates strongly with wage employees (r = 0.995, p < 0.001), with employees holding higher and postgraduate qualifications (r = 0.913, p = 0.002), and with technical and vocational graduates (r = 0.931, p = 0.001), and inversely with total (r = -0.857, p = 0.007) and youth unemployment (r = -0.886, p = 0.003). In first differences, the education-productivity association falls to r = 0.297 (p = 0.518) and to r = 0.074 (p = 0.876) in growth rates, while wage employment retains r = 0.797 (p = 0.032). The findings suggest that level associations reflect a shared trend rather than an annual return in graduate numbers, and that productivity responds to the composition of competencies rather than to the volume of educational output.