Khan Academy spent six months rigorously A/B testing improvements to Khanmigo, its generative AI-powered tutor for students and assistant for teachers. The work focused on reducing response latency and personalizing tutoring using students’ Khan Academy learning history.
Across roughly 20 substantive product tests, Khan Academy tested changes related to making the math agent faster, giving Khanmigo structured Khan Academy learning history, and improving next-item correctness. In total, Khan Academy reported a 6.1 percent improvement in next-item correctness, a measure of whether students can solve the next problem independently after receiving help.
