Darsel's AI Math Tutor Fits Inside a WhatsApp Message

Hundreds of millions of children sit in classrooms every day and still leave without basic math skills. Darsel, a nonprofit education technology organization, attacks this problem with a deliberately modest looking tool: an AI powered math chatbot that lives inside WhatsApp, Facebook Messenger and plain SMS. It matters because it reaches students that almost every other education technology product skips, those without laptops, tablets or reliable internet, and early evidence suggests it works.

The scale of the learning crisis is staggering. The World Bank estimates that about 125 million children lack basic literacy and numeracy despite attending school. In Jordan, where Darsel operates nationally for grade 7 mathematics, the country ranked 73rd out of 81 in math in the 2022 edition of the OECD's PISA assessment. Darsel's answer is to deliver personalized math practice through the messaging apps students and their families already use every day.

A Tutor That Runs on the Phone You Already Own

Most adaptive learning apps assume a smartphone, a fast connection and a quiet place to study. Darsel assumes the opposite. Students send it a message the way they would message a friend, and the chatbot serves curriculum aligned math questions tuned to their level. An adaptive algorithm decides what to ask next based on each answer, so a struggling student gets scaffolded practice while a stronger one moves ahead. The platform holds more than 500,000 questions reviewed by education specialists, and gamified features keep students coming back.

Teachers are part of the loop rather than bystanders. Darsel sends teachers real time usage reports through the same messaging channels, showing who is practicing, who is stuck and where the class needs help. Optional in class recommendations help teachers reinforce exactly what the data says students are missing. In resource constrained school systems where diagnostic testing is rare, this alone changes how classrooms work.

From a Cancelled Spring Break to National Scale

Darsel was founded in 2021 by Abdulhamid Haidar, who studied math, computer science and economics at MIT and worked at McKinsey before turning to education. During the COVID-19 lockdowns of spring 2020, with a spring break suddenly cancelled, he built a chatbot prototype that could deliver math practice over low bandwidth channels and tested it in Nigeria. The idea earned Stanford's Social Innovation Fellowship in 2021, and Darsel joined Y Combinator's Winter 2022 batch, an unusual path for a nonprofit focused on public school systems in low and middle income countries.

The growth since has been rapid. Darsel reports more than 264,000 students reached across over 2,600 schools in Jordan, Nigeria and India, who have collectively answered more than 269 million questions. Early users solved tens of millions of questions even before the platform's current scale, and usage has compounded as ministries of education integrate it into regular math instruction.

What the Jordan Pilot Found

The most interesting evidence comes from Jordan. In 2023, Darsel was adopted by more than 2,000 public schools for grade 7 mathematics, reaching over 70,000 students. According to the Fund for Innovation in Development, which is financing further evaluation, students who used Darsel improved their math performance by an amount comparable to three months of additional high quality instruction. For a chatbot that costs almost nothing per student to run, that is a remarkable return.

Now researchers want to know if the result holds under the strictest test. J-PAL Middle East and North Africa, the regional office of the lab founded by Nobel laureate Esther Duflo and Abhijit Banerjee, is running a randomized controlled trial of Darsel with more than 20,000 grade 6 students across six districts in Jordan, in partnership with the Jordanian Ministry of Education, which has endorsed the program through a multi year agreement. The study will measure math skills, motivation, confidence and even career aspirations, and will examine whether refugee students, about 7 percent of Jordan's public school enrollment, benefit disproportionately.

Why Messaging Beats Apps in Poor Communities

The deeper lesson from Darsel is about distribution. Education technology has produced dazzling products that quietly fail in the places with the worst learning outcomes because they assume infrastructure that does not exist. Messaging apps have the opposite property: they are already installed, already understood and already cheap to use. By building inside that rail, Darsel sidesteps the device gap, the data cost gap and the digital skills gap in one design decision.

The model also scales at software speed without requiring armies of tutors. One adaptive system can serve a national grade cohort, and every question answered makes the targeting sharper. Teachers gain insight without extra paperwork, and ministries gain a real time view of learning that standardized tests only offer once a year.

The Questions That Remain

Darsel's story is promising but unfinished. Engagement is voluntary, and students who never message the bot get nothing from it, so motivation design matters as much as math design. Effects on deeper skills, not just practice question performance, await the randomized trial results. And a chatbot, however good, does not fix crowded classrooms, teacher shortages or poverty itself.

Still, the direction is clear. When the personal tutor that wealthy families pay for can be compressed into a text message, the marginal cost of extra practice approaches zero. If the J-PAL trial confirms the pilot findings, Darsel may become the template for how AI reaches the students who need it most, not through a shiny app, but through the humble chat thread.