AI Use Cases for Education

Our thinking behind the development of the AI for education use cases.

News • December 2023


We wanted to share some of the thinking that came out of developing the AI for education use cases. In this article, we outline the main opportunities for AI interventions in Foundational Literacy and Numeracy (FLN), barriers to use, and other reflections on implementation. Finally, we consider some scenarios where use cases might be applied.  


The main opportunities for the use of AI in supporting FLN were in the following broad areas:

  1. Supporting inclusion by modifying teaching to different learning levels or making learning materials and assessments accessible in different languages or for those with learning difficulties.
  2. Providing 1-2-1 support to develop student and teachers’ knowledge and skills through improved ongoing personalised mentoring.
  3. Data analytics and management and use of education information to inform teaching, learning and support needed by students and teachers.
  4. Digitisation and content creation and adaptation including of assessments, curricula and teaching and learning materials to ensure they are high quality, context specific and curriculum-aligned.
  5. Facilitating personalisation with timely personalised feedback.
  6. Enhancing educational experience to ensure teaching is engaging and at the right level. 

AI tools offer a diverse range of opportunities to improve teaching and learning in FLN. Benefits mostly provide potential to improve the quality of practice, for example through improved analysis, better teaching materials and access to personalised learning and training. There is also scope to increase alignment, for example where analysis is used to assess current materials and curriculum. Cost reductions might occur where tasks are automated or travelling for training is not required. 

Barriers to use 

When implementing AI interventions in low- and middle-Income countries (LMICs), the following are the most significant barriers to use: 

  1. Access to internet  

Sufficient levels of broadband internet access can determine whether AI tools are feasible and in some cases the scope of their functionality.  Occasional access such as once a week may be adequate to carry out a one-off task, such as checking alignment of a document or to upload or download assessment scores and generate analysis. In other cases, it may be possible to operate a tool without ongoing internet access, but this may affect its quality and responsiveness. For example, text-to-speech tools can read out an assessment to younger students without internet access, but providing ongoing feedback as the test progresses would be challenging.  

2. Access to devices  

Access to devices limits what AI interventions are possible, particularly whether there are sufficient for one-to-one student use. Roll out of AI tools might initially focus on teacher support where only one device per classroom is needed, before moving to programmes requiring student access. For example, use of translation tools might first be used for teaching resources to check scalability and effectiveness, before extending roll out to learner materials, which would require large investments in devices. In situations where few devices are available, adaptations may be needed to ensure successful implementation. For example, in assessment rather than using online tests which require higher device access, students could take paper assessments and OCR be used on a single device to collect data for analysis. 

3. Cultural barriers 

AI offers the potential to help overcome cultural barriers, for example through providing translation into local languages or adapting content to suit local contexts. However, it also has weaknesses founded on the way many models are trained on material from HICs and around inherent bias. Furthermore, teachers may be reluctant to utilise technology due to stereotypes, as well as other concerns such as data protection, job displacement, misuse or transparency. To help overcome these barriers, implementors might consider training models using local languages and teaching materials or culturally appropriate examples. Local experts could help with cultural analysis and training.  

4. Teacher knowledge and training 

Effective tool usage may be impacted by teacher knowledge and training. Upfront training and follow up support are likely to be important, which may potentially be provided through other AI tools, such as chatbots. For example, where data analysis generates recommendations for changes to classroom practice, teachers may need ongoing support and reflection opportunities to implement unfamiliar practices. Users of chatbots might benefit from demonstrations and prompt training.

5. Cost drivers 

As well as internet and device availability, implementors might consider other drivers of cost for implementation. For example, the cost of running some LLMs such as chat-GPT4 and hardware needed to support technologies such as VR and AR may make them inhibitive in a LMIC context. Other cost drivers include printers sometimes needed to automate assessments.  


This section considers other factors related to implementation of AI tools to support FLN teaching in a LMIC context. 

 Selecting appropriate tools 

Firstly, the tool used in implementation is important for accessibility. For example, younger students who are still learning FLN skills need tools which use audio or video, for example speech-based AI chatbots. Teachers can access text-based tools, and these would be recommended as they are cheaper to run, require slower broadband speeds and hence are potentially more scalable.    

Considering the level of data analysis possible 

Educational data to be used in AI analysis can be grouped into 2 categories: data generated from online tasks and data inputted by teachers directly. Using the former may provide higher levels of analysis because the AI tool is able to gather more information on the student, such as time taken per question, especially if the task was adaptive and different levels of questions were given. Analysing teacher inputted data may also provide teachers with new insights, but only if data is sufficient and reliable. Again, device availability and internet access might shape the levels of data analysis which are possible and consequently the potential impact of the AI tool.  

Considering the impact of automating processes 

AI can save time through automating teaching or learning tasks, but developers must be aware that teacher involvement is still important. One area where automation needs to be balanced with teacher involvement is when the thinking involved in doing a task is helpful to their practice. For example, teachers often use lesson planning to help them think through delivery. Here, AI technology might helpfully act as a co-pilot to support teachers in developing plans and reduce time spent, rather than completely automating the process. Furthermore, automation of learning processes such as online learning platforms and automatic marking may limit teachers’ initial exposure to students’ progress and struggles. Ensuring data analysis is helpful to teachers in providing information in an accessible way is important in facilitating good understanding of pupil performance. Finally, where AI is automating decisions in teaching and learning, such as around pupils’ learning pathway on adaptive learning platforms, the involvement of teachers is important to ensuring sensible decisions are being made for individual students.  

Considering how technology shapes practice 

The current strengths of AI technology for education centre around analysis and closed tasks. Consequently, the majority of use cases occur with closed tasks such as content creation and adaptation or in analytical tasks. Although it is important to utilise the strengths of AI well, implementers should be aware of shaping classroom practice around the technology. As best practice in teaching focuses on more open ended, creative tasks, a potential avenue is to explore how AI could support these, for example through analysing collaboration, or potentially in future helping monitor, support and direct small group work in the place of a teacher.   


This section outlines fictional scenarios where AI use cases might be applied and which types of use cases might be most effective in each context. 

Scenario 1: Internet connectivity allows ongoing classroom access. Devices are available for teachers and all students in a class 

Daniel teaches at a large primary school on the outskirts of Nairobi, Kenya. His students speak many different local languages as their families come from different parts of the country. Because of the school’s location, it has regular access to broadband internet and has been part of a programme which provided class sets of basic tablets. Daniel can access experts who help him to troubleshoot technological issues, although he still has occasional issues with power and broadband outages.  

AI application: There is potential for a wide range of AI applications in this context including teacher-facing, student-facing and systems-facing options. Big wins would be personalised learning and the potential to collect lots of student data for analysis. 

An example of an application: Daniel’s students could use their tablets to access AI-powered adaptive learning platforms which support personalised reading instruction. Material could be adapted to the Kenyan context and explanations of the material accessed in different local languages. Daniel could add lessons for students to do or alter their levels where needed. This AI tool would enable Daniel to ensure the teaching is at the right level. It would facilitate data analysis to support future teaching and learning, for example information on students whose progress has stalled and need extra support.  

Scenario 2: Internet connectivity allows ongoing classroom access. Access to devices for teachers only 

Abdul teaches at a medium sized school in a small city in Bangladesh. He owns a reasonable smart phone, which he uses to access the school’s broadband internet. Otherwise, hardware is limited at the school and pupils do not have access to devices. Abdul grew up in the Bangladeshi school system and only received poor quality, limited training before becoming a teacher. Consequently, he relies on a whole class ‘chalk and talk’ methodology for teaching.  

AI application: Effective AI interventions in this context are likely to be more systems-facing or teacher-facing, for example support with training and developing classroom materials and plans.  

An example of an application: Abdul could get ongoing mentorship support to improve his teaching practice through AI from a LLM chatbot. After an initial pedagogy training session (which might be an online video or in), he could ask get answers to questions from the chatbot. Proactive reflective questions from the chatbot might further improve his classroom application of the training.    

Scenario 3: Internet connectivity only for occasional online access. Access to devices for teachers and all students in a class 

Mary teaches at a small school in a rural village on the Solomon Islands. A local NGO has donated old devices for the teachers and a class set for student use. However, the school does not have sufficient funds to support a good quality broadband connection that could be used simultaneously by all the devices. They do have some limited internet, which can be used at off peak hours to download and upload limited data.   

AI application: Effective AI interventions in this scenario would likely be teacher-facing or systems-facing. They might focus on one-off tasks that could be performed when internet is available.  In some cases, for example where learning materials or assessments were downloaded onto student devices in advance, interventions could be student-facing. However, even where devices for students are available, the lack of internet may limit the scope of AI usage because functionality could be compromised. 

An example of an application: AI might be helpful to support Mary with administering assessments and analysing the results. If assessments were downloaded onto devices when internet was available, these could be administered on individual student devices. Accessibility could be increased if tests were in an audio format. When internet was available, scores could be uploaded and analysed, giving Mary insight into performance patterns and recommendations to inform future teaching and learning. 

Scenario 4: Internet connectivity only for occasional online access. Access to devices for teachers only 

In a remote, rural Malawian primary school, Comfort teaches a large mixed class. The school has very few learning resources or books, but she has her own basic smart phone. The school has only solar power for lights and no internet access. However, Comfort goes to a local town once a week where she gets internet access that she can use to upload data she has on her phone and use basic apps.  

 AI Application: Effective AI interventions would likely be teacher-facing or systems-facing. They are likely to involve one-off tasks that could be performed when access to the internet is available, such as generating a plan or lesson content or uploading data for analysis.  

 An example of an application: AI might be helpful to support Comfort with providing better learning materials for her students. As they are learning in English, a language which is unfamiliar to many, Comfort could use AI to generate English audio stories, which are adapted to fit the knowledge and cultural understanding of a rural Malawian context. If she downloaded these onto her phone when she had internet access, she could share them with her class at a later date.   

Article prepared by the team at

Explore the Ideas for How AI Can Help With Education

AI has the potential to transform approaches in education, both inside and outside the classroom. Here, we explore some suggestions of where and how AI technology may be used throughout the education system.

We have established 19 areas of the education system where we envision the use of AI and for each of these areas we have described potential appliations. By outlining these applications we hope to frame dialogue within the community and to establish priorities for investment and development.

Sign Up

Join our mailing list to keep up to date with news and events.

Community          Knowledge          FAQ
          Privacy Policy was set up by Fab Inc. in partnership with Team4Tech. We are grateful to the Bill & Melinda Gates Foundation and the Jacobs Foundation for their support.

Powered by FabData.IO