One is good for politics.

The other is bad for society.

The good story is access.

Enrollment at basic and secondary levels has climbed over the past decade, and Ghana outperforms the sub-Saharan African average—nine in ten basic school-age children are enrolled.

However, only 54 percent of Primary 4 pupils are proficient in English and 46 percent in Mathematics, against a national target of 90 percent by age 10. According to the World Bank's Human Capital Index, a Ghanaian child can expect 11.6 years of schooling by age 18.

Adjusted for what they actually learn, that drops to 5.7 years. Nearly six years of schooling produce no measurable learning. This is the quandary. Budgets are constrained.

Teachers are stretched across increasingly large classrooms. Resources do not match our ambition.

We need to do more with less, and we need to do it to improve access and quality simultaneously. AI could be part of the answer, but only if we design for our context, measure what matters, and insist on partnership terms that protect our interests.

The good news is that Ghana has started well. What follows next is where most countries fail including the US

This month, Sal Khan, founder of Khan Academy, which serves over 100 million learners worldwide, publicly admitted that his flagship AI tutor, Khanmigo, hasn't gained adoption in practice.

Khanmigo had every advantage: first access to GPT models, Microsoft backing, Gates Foundation grants, and state subsidies. The failure was not due to resources.

It was their method. The team built the tool, then pushed it onto teachers and students. When students did not use it, the chatbot was made more intrusive. 

When that did not work either, teachers were blamed, and then students were blamed. The humans had to adapt to the tool, not the other way around.

Ghana took the opposite approach.

In October 2025, the Ministry of Education rolled out subject-specific AI apps, built with PlayLab, to all 712 Senior High Schools, supporting 68,000 teachers delivering Ghana's new SHS curriculum to 1.4 million students. 

The apps do not sit beside teachers' work but are embedded inside the coaching, lesson preparation and curriculum delivery practices teachers already use through our national Professional Learning Communities, which meet weekly in every school.

A recent third-party study shows 77 percent adoption in the national roll-out, with more than half of teachers engaging multiple times a week and reporting they feel more motivated, supported and willing to stay in the profession.

The design choices behind this result are worth naming. The apps are built on Ghana's own curriculum materials - teacher manuals and standards, learner resources, and frameworks on national values.

The partnership matters just as much: T-TEL, a Ghanaian education non-profit, leads implementation; PlayLab provides the AI tool with subsidised compute from Amazon Web Services and Anthropic; the Mastercard Foundation funds the pilot; and the Ministry of Education acts as anchor client, defining what Ghana needs.

This is not a top-down deployment but a collaborative intervention with Ghana's educators and education system at the centre.

There is also a policy argument for why this partnership approach matters.

In July 2024, the African Union Executive Council endorsed the Continental AI Strategy in Accra on how AI should be developed and deployed across the continent by respecting context, data sovereignty, infrastructure limitations, and local capacity building.

Ghana is one of the first AU member states implementing AI in education in a way that is consistent with this leading framework.

Ghana has done the first hard thing: it has built a contextually grounded, government-anchored deployment. Three things deserve attention now. Impact measurement is where most edtech investments fail. Today, the dominant metric across AI edtech is engagement ‒ how many users, how many chats.

It is the same quantity-over-quality trade-off that Ghana is trying to escape in education more broadly. What we should be measuring is whether literacy and learning-adjusted years are improving, and at what cost. We should insist on independent outcome measurement as a condition for funding scaled pilots.

Then, by standardising the measurements and publicising outcomes, African governments can also learn from each other's experiments and better benchmark partners.

Procurement leverage is a continental opportunity. Procurement is where partnership terms either protect our interests or quietly erode them. African governments currently negotiate with large AI vendors individually, with limited technical capacity and no shared standards. 

A continental procurement framework ‒ covering impact assessments, vendor transparency on training data, data localisation, and open API requirements - would shift the power dynamic. NEPAD and the African Union are actively developing one.

Ghana is well positioned to shape it and to lead implementation by convening and collaborating with leading African Ministries of Education.
Digital and AI fluency is becoming table stakes for the economy.

As we strengthen early digital and AI fluency for teachers, we need to develop a roadmap for enabling student fluency as well. Nations that manage this transition will not just develop great local talent. They will build globally competitive workforces for remote work, and with them, a growing and resilient tax base.

The risk with early pilot successes is that we mistake the announcement for the outcome. Partnerships get press. Pilots get photographs.

Ensuring real impact is not glamorous, but it is arguably the only metric that matters. Soon, the conversation should shift from how many teachers, schools and students have AI access to whether students’ learning outcomes are meaningfully improving.

Ghana is early, well-positioned, and laying the foundation for a scalable way to support our teachers and students.
The job now is to insist ‒ as citizens, as policymakers, as funders ‒ on the measurement discipline for innovation pilots like these that tells us what's working for the outcomes that really matter. Increasing access wins headlines and votes. Improving learning should, too.