By Alfred Ouko, Founder & CEO · July 2025 · 8 min read
We ran an internal study of 240 Smartious students over the 2024 academic year. Half used our Mshauri AI tutor for 30 minutes daily. Half did not. Here is what we found.
We took 240 students preparing for IGCSE, A-Level, or KCSE. We matched pairs by starting grade, subject, and teacher. One student in each pair got unlimited Mshauri access; the other got only the standard teacher-led tuition. Both groups sat the same mock exams throughout the year.
Students using Mshauri for at least 20 hours a month improved their mock exam grades by an average of 34% — roughly one full letter grade, e.g. from a C to a B. Students who used it less than 10 hours a month showed no significant difference. The threshold matters.
Below 10 hours a month, usage is sporadic and does not build the habit of self-directed revision. Above 20 hours, the student is using Mshauri daily and internalising its Socratic questioning style. The middle range had mixed results — consistency matters more than total hours.
Mathematics (+41%) and Chemistry (+38%) showed the largest gains, followed by Physics (+32%) and Biology (+28%). English Literature showed +19% — smaller but still meaningful. History and Geography showed only +12% — AI tutoring is less differentiating for humanities, where discussion and essay coaching from a human tutor matters more.
Mshauri did not help students who were disengaged or who had personal issues outside academics. Students under significant family stress, or those with undiagnosed learning differences, needed human intervention before any AI tool could help. We do not claim Mshauri solves motivation — it only helps students who are already trying.
We are publishing this data alongside a white paper in Q1 2026. We also made Mshauri freely available to all enrolled Smartious students — not sold as a separate product — precisely because it works best when combined with human tutoring, not as a replacement.
AI tutoring is not a gimmick, but it is also not a silver bullet. Our data shows it amplifies what a motivated student is already doing. For families trying to decide whether to add an AI component, the question is not whether it works — it is whether your child will actually use it for 20 hours a month.