Khan—s AI tutor: board hears demo of Khanmigo as a way to scale 1:1 tutoring

Sal Khan, founder of Khan Academy, demonstrated an AI tutoring and teaching-assistant product (Khanmigo) during the chancellor—s AI educational series at the Jan. 14 Board of Governors meeting.

Khan framed the tool as an attempt to reach Bloom—s long-standing observation that one-to-one tutoring produces dramatically higher learning outcomes than lecture: "Bloom—s 2-sigma problem" suggests tutoring can move students far beyond average classroom performance. Khan argued the combination of on-demand video resources, adaptive practice and a chat-based, Socratic AI tutor can scale many elements of that effect.

Demonstrations shown to the board and in the room included:

- Socratic problem-solving coaching in algebra and distributed-step prompts rather than simply giving answers; the AI asked students to explain their steps before nudging them to correct distribution errors. - Live coding help that helped a student reason about why an animation only moved one object, prompting the student to inspect a variable that controlled the other object. - A writing coach workflow that records student-AI exchanges, gives formative feedback and provides teachers with a log and analytics about what the student worked on and how long it took. Khan said the transparency can reduce misuse: "If a student copies and pastes from another source, the teacher can see that the AI wasn—t used to co-construct the piece and that discrepancy can be followed up."

Khan emphasized guardrails: in K-12 deployments conversations could be recorded and viewable by teachers, and teachers can set whether an assignment allows AI support. He said teachers can also toggle into a "teacher mode" where the AI acts as a lesson-planner or a guide to craft active-learning hooks. Khan and board members discussed concerns: accuracy, hallucination, privacy, cheating and computational energy costs. Khan acknowledged trade-offs and framed the technology as neutral: "an amplifier of human intent," stressing that the responsible use by educators is critical to outcomes.

Chancellor Christian said the Khan presentation fit the board—s educational series on AI and follows pilot work already underway in the system. Board members asked how to preserve human connection, how to use AI to support remedial gateway learning (math), and the potential to use AI plus mastery-based assessments for credit-for-prior-learning. Khan suggested competency-based validation plus teacher proctoring as a route to award credit where mastery is demonstrated.

The demonstration generated many immediate questions from board members; chancellor—s office staff said they will continue piloting AI tutors with attention to evaluation, safety and faculty training.

Provenance: transcript of Sal Khan—s presentation and Q&A (SEG 6156—6700 and surrounding blocks).