Kenya replaced its entire school system — 8-4-4 to the Competency Based Curriculum — faster than teaching materials could follow. Learners and teachers were asked to work to a new curriculum design without textbooks matched to it, and the first KJSEA cohort sat the exam in late 2025 with very little to revise from.
On the outgoing system the cost of that gap is measurable: only 27.18% of 2025 KCSE candidates scored C+ or above, the grade that opens direct university entry. The learners below that line are not less able. They are less resourced.
The way Kenyan families close the gap is private tuition, and it costs more than most households can carry. What a good tutor actually gives a student is narrow and specific: a complete worked solution, in their own curriculum, at the moment they are stuck. I built Grade AI because that specific thing can now be delivered at almost no marginal cost — and if it can, then whether a child gets it should stop depending on what their parents earn.
Grade AI is a digital textbook first and a chatbot second — a deliberate choice, because a learner who is stuck does not know what to ask.
Content is organised by the KICD curriculum hierarchy: a student opens their grade, subject and topic and finds the concept explained, worked examples carried through step by step, the misconceptions that usually trip people up, and practice with answers. Chat sits underneath as a fallback when the page does not answer the question, and the tutor is grounded in the curriculum content with a refusal system rather than left to improvise.
There are four roles: students, teachers who verify and assign, parents who can see progress, and school admins. Teachers matter structurally — content enters the system as draft and stays unverified until a teacher signs it off.
Explanations for maths and symbolic content are programmatically animated rather than AI-generated video, so an equation on screen is correct by construction, not by luck.
It runs as an installable web app and reaches learners over WhatsApp, which is where Kenyan families already are. Schools pay through M-Pesa. Learners' data is handled under the Kenya Data Protection Act with parental consent captured at signup.
School by school rather than download by download.
We are live with paying schools at KES 50,000 per package, billed through M-Pesa, and more than 1,000 learners have used the platform. The early route to market was deliberately unglamorous: before the backend existed we ran click-to-WhatsApp ads as schools closed ahead of the national exams and delivered tutoring manually, which told us within weeks what parents would actually pay for.
A guest mode lets anyone try five questions with no signup, which is how most teachers meet it first.
Content build has been the real constraint on spread, so it has been sequenced rather than rushed: Grade 6 rebuilt from a zero-omission KICD audit into 115 units, 121 units across six subjects for Grades 1–3, and a deep rebuild of Grade 9 across maths, integrated science, English, social studies, Kiswahili and agriculture.
Next is Grade AI Prep, an IGCSE and A-Level tier for international schools, which cross-subsidises the CBC work rather than competing with it.
Go to mygradeai.com. You can open the guest mode and ask five questions without signing up or giving us anything — that is the fastest way to see whether the explanations are any good.
If you are a teacher or a school and want it for your learners, email info@mygradeai.com and we will set up your school and walk you through assigning content.
If you are a parent, sign up at mygradeai.com, pick your child's grade, and you will see the same curriculum your child is sitting, with the worked solutions attached.
The biggest change was abandoning what we started as. Grade AI began as a KCSE tutoring assistant on WhatsApp — chat first. Watching real learners use it showed the flaw: a student who is stuck often cannot phrase the question, and a chat window gives them nothing to read. We rebuilt around a structured, browsable textbook with complete worked solutions, and demoted chat to a fallback. That was the right call and it cost us months.
We also changed how explanation videos are made. AI video generation produced beautiful nonsense for maths — wrong equations, confidently rendered. We moved symbolic and mathematical content to programmatic animation, which is accurate by construction, costs almost nothing per video, and keeps files small enough for Kenyan mobile data.
We narrowed scope too. Rather than covering every grade shallowly, we sequenced: KCSE Form 3 and 4 as a revenue bridge across the two remaining exam seasons, and Grade 9 CBC as the permanent product.
And we corrected ourselves on competition — we had claimed the market was wide open, then found Soma by CBC Edu Kenya already there with a free tier.
