Artificial Intelligence-Supported Feedback for Teacher Professional Development: Opportunities and Challenges

Abstract

Few levers in teacher professional development are as well supported by evidence, or as hard to supply at scale, as timely individual feedback on classroom practice. Artificial intelligence is now being offered as a way out of that bind. This article reviews the recent international literature on AI-supported feedback for teachers, published between roughly 2017 and 2026, and reads the resulting opportunities and risks as one connected argument before turning to the Republic of Kazakhstan and the work of the JSC National Center for Professional Development “Örleu.” Tools built on natural language processing and large language models, among them M-Powering Teachers and TeachFX, can return cheap, consistent, descriptive feedback on how teachers talk; in at least one large randomised trial they shifted both teaching practice and student outcomes. The same literature explains why enthusiasm should be tempered. The reliability of automated judgements is still unsettled. Models carry bias, recordings raise real questions about privacy and surveillance, teachers may be deskilled or their relationships with students worn thin, and the evidence remains young, short-term, and heavy on self-report. For Kazakhstan one further constraint is decisive: Kazakh is a low-resource language that current speech and language models handle poorly. Reading the UNESCO AI Competency Framework for Teachers (2024) against Kazakhstan’s 2024–2029 policy settlement, the article makes the case for a staged, human-centred model in which AI-supported feedback widens the reach of mentoring instead of standing in for it.