This chapter illustrates how Artificial Intelligence (AI) is integrated into educational impact evaluation as an enabler—not a replacement—that speeds up the validation of findings, broadens analytical coverage, and enables working with both quantitative and qualitative, structured and unstructured data, while ensuring traceability and methodological rigor. The evolution is described in “two waves”: first, AI reshaped performance evaluation through formative and adaptive feedback; then, it started transforming impact evaluation by enhancing targeting, multi-source triangulation, and causal attribution. The chapter introduces an operational typology (ML, NLP, LLMs, and data integration/governance). It demonstrates its role as a methodological co-designer during the design, implementation, and reporting phases, highlighting that the purpose of evaluation and causal frameworks remains central. Risks such as solutionism, biases, and confusion between prediction and causality are discussed, along with their mitigation strategies—including human oversight, outcome and equity metrics, lightweight MLOps, governance, and privacy by design.