A board can only challenge what reaches it.

A trust board was told this month that it could not show it had seen something.

On 22 October, Schools Week reported that the Department for Education had issued Athena Academies Trust with a notice to improve. The findings included significant disproportionality in exclusions of pupils with SEND, and trustees' records that showed little evidence that serious concerns reached the board, or that trends and outliers were reviewed. Seven of the trust's nine schools were rated good or met the expected standard. The finding was not that nobody cared. It was about what the board was shown.

A board can only challenge what reaches it.

The same gap, four times in one week

The same Schools Week front page carried a commissioner's review of Slough's SEND services. Deborah Glassbrook found fragmented data systems and a safety valve agreement that rested on inaccurate data. A deficit figure is built from counts of children, plans, hours and places, and those counts begin in schools.

On 22 October, Tes reported on subject access requests. Dixons Academies Trust went from 16 requests in 2023-24 to 133 in 2025-26. A survey of 23 trusts by Leigh Academies Trust found that almost all named generative AI as the main driver of volume and complexity. Asking has become cheap. Answering has not. A request means finding everything held about one person, across systems that were never built to be searched together.

On 23 October, Tes reported that the final KCSIE 2026 differed from the July draft in ways that schools had to find for themselves.

Four stories, one pattern. The information exists somewhere in the school. It is not in a form that can be counted, searched or compared. So the person responsible for the answer assembles it by hand, late, under pressure.

What it costs

For a SENCO or a head, it costs evenings. For a governor, it costs the ability to do the job. A trustee can ask how many of the pupils suspended this term have identified SEND. If nobody has joined the behaviour record to the SEND register, the answer takes a week, and by then the meeting has moved on.

For a child, the cost is quieter. A pattern nobody counts is a pattern nobody acts on.

What changes it

Not a bigger policy and not a longer board pack. A record made once, at the moment the work happens, in a form that can be counted later.

That is the design rule we work to at Aevum Foundry AI. Teacher Manager marks pupil work and drafts feedback that teachers approve, and it runs on the school's own device, so the work stays where the school can see it. I want to be careful about what I claim. We are an early company, we are working with schools to learn what a useful record looks like, and we do not have a finished answer for exclusions or access requests. What we have is a rule: if information is created inside the school's own building, in one place, it is easier to count, easier to search and easier to show a board.

There is a second benefit for anyone responsible for data. A tool that runs on the school's machine leaves fewer copies of a child's work in fewer places. Fewer copies mean a shorter search when a request arrives, and fewer questions about where something went.

A test for governors

Pick one number you would want next term. Pupils suspended with identified SEND, set against their share of the roll, is a good one. Then ask how long it would take to produce. If the answer is a day, the record is working. If the answer is that someone would have to pull it together, that is your finding, before an inspector writes it for you.

If you lead a school and want to talk through what a record that adds up would need to look like, get in touch through the founding school page. I would rather learn from your answer than guess.