The main thing I have learnt this half term is that AI is a tool and should be treated as such. To harness its massive power, you need to be clear on the desired outcome and be able to effectively articulate that in a prompt.

Your ability to effectively write prompts for AI will be the difference between success and failure.

Highlights: local LLMs and reading books

My personal highlights of this half term are:

• Building a local LLM prototype and training it on specific, qualitative pupils’ SEMH (dummy) data. This ability to work with contextual, complex data locally opens up so many possibilities for effective analysis and pupil support.
• Creating bespoke reading books using Storybook inside Google Gemini. I created reading books based on a pupil’s current phonic knowledge and their age-appropriate interests. This resulted in reluctant readers being excited to read.

Staff and AI

My concerns mostly stem from the conversations I’ve had with staff across the board—from teaching assistants right up to CEOs. Nearly everyone believes AI is valuable, yet most staff don’t currently understand how they can use it to make a positive impact right now. My worry is that ‘AI fatigue’ will set in and opportunities will be missed.

I heard from one senior leader who admitted they are currently just using AI as a search engine. They desperately want to do more but simply don’t know where to start. This leads directly to the next major challenge: scaling.

The Infrastructure and Scaling Challenge

I have seen pockets of successful AI use in schools and Trusts. The massive challenge now is scaling that success school/Trust-wide in a way that staff of all IT literacy levels can easily access.

This challenge is compounded by the pressure on IT support teams. They are constantly firefighting daily IT issues, meaning there’s little opportunity to ring-fence the time needed for innovation. And innovation is essential—it involves developing and building valuable AI assets within the secure IT infrastructures that are already in place.

Commercial AI products

Many commercial products are integrating AI to analyse pupil data and proactively flag areas for support. Crucially, much of this currently focuses on quantitative data. While this is a useful starting point, I worry it can misrepresent complex issues as mere ‘trends’ and be ineffective at providing true support at the individual pupil level.

The integration of qualitative data is coming, but because there’s less immediate commercial demand for it, companies, understandably, aren’t prioritising features customers won’t pay for. I believe it’s our responsibility as educators to step up and understand the significant value qualitative data analysis can bring to truly progressing our pupils.

What’s Next for the Break?

Over the upcoming half-term break, I plan to:

1. Develop the local LLM further, specifically by integrating voice commands to streamline single-user workflows.
2. Build an AI Agent to allow colleagues to easily access and replicate the bespoke reading book process in Gemini.

I hope you all have a happy and healthy half term break.