Understanding the ‘Ai’ in Aila

An overview of how Aila generates high-quality lessons using AI techniques

In September, we launched Aila, our AI lesson assistant. We know that teachers enjoy lesson planning and like to have ownership over their lessons, but many also feel that they spend too much time on it. We wanted to develop a tool that keeps you in charge of planning the lesson but uses AI to do the heavy lifting—for example, adjusting the literacy level so it’s appropriate for your class or using a local case study to make it more relevant for pupils.

Everyone has different views, perspectives and feelings about AI in education, so it’s really important to us that we are transparent about how Aila works, to help you feel reassured about using it for your lesson planning.

In this blog, we explain how we’ve drawn on our bank of thousands of sequenced teaching resources, subject expertise, our experience in creating these resources and a deep technological understanding of LLMs (more on that later) to build an AI tool to support you in producing high-quality content that you can use in your classroom.

Retrieving Oak’s content to improve quality

The quality of any AI tool's outputs depends on the quality of the inputs you provide to it. Therefore, having a corpus of high-quality UK education-specific content is critical.

We have model curriculum plans for every subject across key stages 1-4 with resources that can be used to plan over 10,000 fully resourced lessons, all created and reviewed by subject-specialist, expert teachers. They are aligned with the national curriculum and available on an Open Government Licence (OGL), making them the perfect base for us to develop a high-quality AI tool for teachers.

Retrieval augmented generation (RAG) is an AI technique that allows us to direct Aila to prioritise where it draws on content from. Unlike other AI tools, we've directed Aila to draw from our curriculum plans and resources, over the general content available on the internet.

We also use a type of RAG called ‘content anchoring’ where Aila can base your lesson plan and resources on a specific Oak lesson from our curriculum to further increase its quality and accuracy.

These techniques significantly reduce the likelihood of hallucinations (when an AI tool provides false information), bias, or lots of Americanisms.

Engineering our best practices into our prompt

Behind the scenes, Aila uses an LLM (Large Language Model). LLMs have been trained on a huge amount of data (usually from the internet) and use this data to predict how to respond to questions or finish phrases based on patterns in language within the model. Aila currently uses OpenAI’s GPT4o, but we can change it in the future when we find a model that performs better.

Providing clear instructions to the LLM is what we refer to as prompt engineering - to maximise the quality, relevance, and specificity of content produced by it.

Our curriculum plans and resources are produced by our curriculum partners and hundreds of teachers, who are experts in their subject areas, such as Raspberry Pi for computing and the Geographic Association for geography. To support this work, we developed clear, detailed guidance for our partners on designing high-quality lessons aligned with our curriculum principles.

We used this guidance when building Aila. We codified it into our prompt, along with examples and non-examples, to support Aila in designing high-quality content. This means that even if you create lessons in Aila on topics not covered by our curriculum plans or resources, they will still be high-quality and pedagogically sound.

Read more about the pedagogy behind Aila.

Building in the open

At Oak, every line of code in our product is open source, enabling anyone to view, use, and build upon it. Furthermore, as a public sector organisation, we’ve just published detailed information about how Aila works as part of the Government’s new Algorithmic Transparency Recording Standard. This document goes into detail about how we built Aila, such as the models and technologies we used, the risk assessments we went through, and how we aren’t using any personal data. This demonstrates our commitment to the highest standards and ensures you can trust our tools.

Explore Aila yourself

Aila is, and always will be, free - so why not take a look and see what it can help you with?

You might also be interested in:

• Aila’s safety guardrails explained
• How to plan a lesson: a helpful guide for teachers
• Webinar recording: How to harness the potential of AI in your school
• Webinar recording: How to plan high-quality, personalised lessons with Aila