Can a platform help a pupil without replacing the insight of their teacher? In South Africa, school group Advtech is betting on artificial intelligence to personalise exercises, while seeking to expand its tools across its regional network.
Technology already in use across the group’s schools
On 21 October 2026, Advtech announced the acquisition of a 25% stake in MathU, a South African educational technology company. The deal gives it a seat on MathU’s board and exclusive rights to its current and future AI-based educational solutions in the markets where the group operates.
The amount was not disclosed in the public announcement. However, Advtech’s interim results state that the stake was acquired on 16 July 2026 for 15.8 million rand in cash. The group had already integrated MathU’s platform into its schools division in 2023, under the Advlearn brand.
The tools are used in subjects including mathematics and physical sciences. They offer activities adapted to pupils’ levels and guided explanations. Advtech also plans to extend the offering to English. AI features can direct exercises based on a learner’s answers; on their own, they do not guarantee improved results.
Investment in an expanding network
The deal is part of the development of Advtech, which is headquartered in Johannesburg, South Africa. The group operates schools and higher education institutions in several African countries. Its results for the first half of 2026 report 119,197 enrolments, including 47,730 in schools and 71,467 in higher education.
Its school activities outside South Africa are also growing: the group recorded 13,161 pupils there in February 2026, up from 11,550 a year earlier. These figures help explain the commercial appeal of digital tools that can be deployed across several campuses. They do not, however, show how access to these tools will be distributed among institutions, school levels or families.
For Advtech, an equity stake goes beyond a supply contract. It gives the group a role in MathU’s governance and exclusivity over the solutions concerned. In return, MathU gains a partner with a network of schools in which to test, distribute and develop its products. The precise terms of this exclusivity were not detailed in the announcements reviewed.
The challenge: integrating digital tools with teachers’ work
Adaptive software can identify recurring errors and suggest targeted exercises. It can also give staff insights into the difficulties encountered by a class. But the quality of learning will depend on how this information is used: teachers must be able to interpret it, correct the tool’s limitations and adapt their support.
This question is particularly important in a private network whose schools serve pupils with different means and needs. Access to digital support can supplement lessons, but it also depends on the availability of devices, connectivity and time for guidance. The announcement does not specify the arrangements for deployment beyond Advtech campuses or any possible user fees.
MathU says it wants to extend its solutions to more pupils. Its MathU Society initiative plans to provide mathematics and science support to learners from underserved communities. Both companies have announced this social dimension; however, the public information available does not yet provide a timetable, the number of beneficiaries or measured results.
A signal for digital education in southern Africa
The partnership illustrates a shift in the market: established school groups are seeking to integrate tools developed by local technology companies. The regional challenge is not simply to deploy AI, but to ensure it works with the curricula, languages and teaching conditions specific to each country. Advtech’s presence in several markets could facilitate this adaptation, but does not guarantee it.
The next concrete steps will therefore be worth watching: the planned launch of the English module, the expansion of MathU Society and how teachers use the data generated by the platform. For pupils and their families, the value of the investment will be measured less by the sophistication of the algorithms than by whether support is genuinely accessible and useful for learning.



