High-tech Solutions in Low-tech Contexts: Supporting Skill-focused Lessons in Remote Classrooms

As the Country Director of EducAid Sierra Leone, I was invited to present our innovative work with AI for education, at the 4th edition of the GPE KIX Continental Symposium on Research and Innovation in Education in Africa.

EducAid is a Sierra Leonean non-profit organisation and UK registered charity, running schools and school improvement programmes working to transform education in Sierra Leone. We have been working to bring high-quality education to Sierra Leone for more than 30 years, growing from one school serving 20 learners in September 2000, to a strong team of 150 serving over 250,000 learners in 2025/26. EducAid partners with the Ministry of Education to ensure the District Ministry teams have the necessary support to tackle learning poverty. 

In Sierra Leone, education faces significant challenges including high student-teacher ratios, limited infrastructure, and widespread learning poverty. For example, in 2022, 95.8 percent of 10-year-old children were unable to read and understand an age-appropriate text [1]. With teachers often responsible for large classes and few schools having access to electricity or internet (6% and 1%, respectively), providing quality education in this resource-constrained environment is a formidable task. Compounding these issues is the low qualification level among many teachers, which further impacts the quality of education delivered.

These challenges are shared by other countries across sub-Saharan Africa. However, as this project illustrates, through a people-centred approach, we can collaborate and use AI to tackle learning poverty even in the most remote contexts.

EducAid used AI for education in collaboration with Fab AI in 2022, who created a ChatGPT-based AI chatbot – TheTeacher.AI. This chatbot leverages GPT-3.5 Turbo technology and is carefully adapted for the local context where mobile network coverage includes 93% 2G (83% 3G and 50% 4G) coverage. It operates on WhatsApp, which is known for its robustness under poor network conditions and data efficiency, as well as its familiarity among teachers, enabling accessibility even in the most underserved schools.

EducAid implemented preliminary research to see what levels of support were needed for teachers to feel comfortable asking the chatbot questions and to enable them to ask questions in such a way as to get useful answers. The tool was piloted by assessing teacher use in four groups (40 teachers per group) with various exposure to the tool: 

1. ⁠Group 1: Early access and introduction to the tool.
2. ⁠Group 2: Access and a short introductory level training.
3. ⁠Group 3: Access, full training on crafting prompts, and coaching.
4. ⁠Group 4: Served as the control group with access only in the final month but with no other support.

We found that teachers with coaching used the tool most and for the widest range of uses. Despite the very real infrastructure challenges, TheTeacher.AI illustrated the promising role AI can play in supporting educators in low-resource settings. We learned that successful implementation hinges on careful adaptation to local conditions, continuous human oversight to ensure quality and accuracy, and leveraging existing technologies that are already embedded in community use.

In addition, this work revealed that TheTeacher.AI can serve diverse educational needs, including concept clarification, lesson planning, and writing support. Thus far, teachers have used the tool to generate vocabulary lists, explanatory texts, practice questions, and more across various subjects like mathematics, science, and business economics. 

“For students in our schools, AI tools help by giving lessons that match how each student learns. Whether you need extra help or can move faster, AI helps you learn at your own pace.” EducAid School Leader, Port Loko District, March 2025.

Drawing on this success, EducAid has been further supporting teachers to use a range of AI tools that go beyond the WhatsApp-based chatbot, and to use them for designing engaging lessons. Implementing the use of AI tools (Pi.ai, Perplexity.ai, ChatGPT, CoPilot and even Suno.ai for creating learning songs) teachers are learning to create content, glossaries, activities and questions (including practice exam questions). The result is more engaging lessons, a movement away from just presenting content and understanding how to create opportunities for learning skills and practising them.

But how can teachers give a class of children access to these materials when there is no printer and limited paper? This is where the low-tech piece comes in! Teachers have adapted the tool to their low-tech environments by transcribing the AI output onto flattened cardboard boxes, creating a range of activities that can be used in a carousel. 

A 5-station carousel might include activities such as the above.

In a 60-minute lesson, learners might spend 10 minutes on a carousel having started with a short introductory explanation and then ending with an exit ticket. With a focus on creating opportunities for children to practice speaking, listening, reading and writing about the correct vocabulary, the lessons are more interactive and result in genuine learning and skills acquisition.

This approach has been found, in early observations, to enable even under-resourced teachers and schools to focus on creating opportunities for practising core skills and using newly acquired vocabulary to explain the concepts being studied. This creates tools for learner-centred and skills-focused teaching.

As we expand this work into more and more schools (in 2026, we will be working with 1200+ schools at both primary and secondary level), we will be assessing the impact on teacher workload and student learning. 

Although the majority of the symposium presentations focused on the high-tech applications for which AI can be used to manage and improve education services, EducAid’s presentation was unique in showing how students without devices can benefit from an AI-enhanced solution. Strong engagement at the Symposium confirms both the relevance of low-tech, inclusive approaches and the growing global need for practical AI solutions that leave no learner behind. We thank GPE-KIX, a joint endeavor with IDRC, for the opportunity to share this work that is helping bridge the digital gap and reduce learning poverty in one of the world’s most underserved countries. 

[1] Learning Poverty is defined as the % of 10-year-old children (in and out of school) who cannot read a 150 word age appropriate text, https://www.educationpolicydashboard.org/indicator_timetrend/sle/44/44