Kai Coach asks the questions. The student does the thinking.

Every turn ends with a question.

Kai Coach never writes for the student, and it never makes the inference for them. It reads where the student stands, and it asks the one question that moves their thinking forward. When a student stalls, Kai Coach offers a definition or points to a passage, and then it asks again. The student keeps the pen from the first sentence to the last. This discipline separates a coach from a chatbot, because a chatbot answers the question while a coach teaches the student to answer it.

Kai Coach may supply the term. It never supplies the reasoning.

In the History pilot, across 389 instrumented discussion and peer-response sessions, every coach turn contained a question.

Grounded in the course itself.

Kai Coach works only from the materials the instructor assigns: the readings, the primary sources, the problem sets, and the exam questions. It sends students back to the specific passage that holds the evidence, and it asks them to cite that passage in their own work. General chatbots draw on the entire internet, and in our tests they rarely touched the assigned readings at all. Kai Coach keeps the student's attention on the texts the course actually teaches. That focus explains why citation of assigned readings rose so sharply in History 2C.

Built around the places where students struggle.

Kai Coach draws on eight years of teaching and of close reading of how students write and reason. To build the first version, Elizabeth Landers analyzed sixty-four History 2C essays from 2025 and identified the patterns that held students back. These patterns included quotations dropped in without analysis, theses that described rather than argued, and hedges that never arrived at a claim. Kai Coach recognizes each of these patterns and asks the question that moves the student past it. Every new course begins with the same step: a conversation with the instructor about where their own students have historically fallen short.

Different tasks call for different coaching.

A single style of questioning cannot serve every assignment, so Kai Coach adapts its approach to the task. On the History 2C final exam, it worked in three modes: one that helps students build an argument, one that asks them to critique an answer an AI produced, and one that asks them to reason from inside a historical perspective. In Physics, it reflects with students after an exam, calibrates their confidence before the next one, and probes their reasoning on weekly problems. In Business Ethics, it moves through three phases, which we call Mechanics, The Twist, and Judgment. Throughout that module, Kai Coach stays evenhanded and never offers its own ethical verdict.

It knows when to step back.

A good coach recognizes when a student is ready, and Kai Coach does too. When a draft meets the bar, Kai Coach confirms it and releases the student to submit, rather than inventing new objections. When a student begins to circle without progress, Kai Coach notices the plateau and changes its approach. Students may return as many times as they wish, because the History 2C study showed that the number of passes did not by itself produce improvement. The quality of the thinking in each pass is what matters.

Integrity by design, not by policing.

Kai Coach responds to generative AI by strengthening student work instead of policing it. It gives students a reason and a method to engage with the course material and to cite it in their own words. Because the coaching is part of the course, instructors can read every session and download it, rather than wondering what happened in another browser tab. The responsibility for the thinking stays with the student, where it belongs. As our motto puts it, the thinking has to be theirs.