Relay Relay Relay RELAY
Every student’s confusion, heard.
Every teacher, in the loop, in real time.
RELAY is an AI-supported tool for K-12 schools, built on learning science rather than assumptions. Students get the guardrails research shows improve outcomes. Teachers get real-time visibility of where a class actually is, without any extra marking.
Same underlying model in both. What changed was how it was designed to behave with a student. Bastani et al., PNAS 2025
What makes RELAY different
Three decisions
most tools don’t make.
Purely formative. A student can admit what they don’t understand without it costing them a mark. That honesty is the whole mechanism: it is what makes confusion visible in the first place.
RELAY reports what it saw and stops. It does not recommend an action, contact a student, or put anything on a report. Your teachers keep the professional judgement that only they can exercise.
RELAY draws only on the material your teachers supply, shaped to your curriculum, ethos and cohort. Guardrailed does not mean prescriptive: the constraint is deliberate, the configuration is your school’s.
How RELAY works
Six things it does,
and one it won’t.
Pick any of them. Each shows what a student or a teacher actually puts in, and what comes back.
A student, on their own
I get that the writer uses a metaphor here.
But I don’t really know what “tone” means in this question.
What the teacher sees
Student 04
“I don’t really know what ‘tone’ means in this question.”
Student 11
“I picked the answer but I’m not sure why it’s right.”
The question they would never raise in front of the class.
Students rarely put their hand up to say they’re lost. RELAY gives them somewhere to say it plainly, and passes it on in their own words, so it reaches the person who can actually help.
Wednesday, 9:40pm
Homework, question 4
I got an answer but I don’t know if the method is right.
I think I just copied the example.
No teacher awake. No hand to put up. Nowhere else this goes.
Thursday, 8:15am
Student 07 · last night
“I think I just copied the example.”
Same issue, same question
6 of 28 students
The teacher walks in already knowing what to open with.
The moment nobody sees.
Most confusion happens at a kitchen table at night, hours from the nearest teacher. RELAY catches it there and has it waiting before period one, so the lesson starts where the class actually is.
Tuesday, period 3
The class is working through the set task.
Nothing is marked. Nothing is submitted. Understanding is simply becoming visible as it happens.
Live, mid-lesson
Before the test, not after it.
Confusion surfaces while there is still a lesson left to change. Not in a summative result three weeks later, when the chance to do anything about it has already gone.
What the student is asked
Explain your reasoning in your own words.
No mark is attached. Nothing here counts towards a report. There is no one to perform for.
What comes back
Remove the mark and you remove the performance.
A graded task teaches students to look like they understand. RELAY is purely formative, so honesty costs them nothing, and honesty is the only thing that makes the confusion visible in the first place.
RELAY has surfaced something
Four students are stuck on the same idea.
RELAY reports it. It does not decide what to do about it, and it never contacts a student on your behalf.
Yours to choose
RELAY stops here. That is a deliberate design decision, not a missing feature.
It informs. You decide.
RELAY never recommends an action or replaces a professional judgement. It tells you what it saw and stops. The teacher knows the child, the room and the week, and remains the one who decides.
What you give it
Your syllabus
Your class notes
Your answer sheet
Your rubric
Nothing from the open internet. Nothing from another school.
What it works from
The constraint is deliberate. The configuration is the school’s.
The guardrails are yours to set.
Guardrailed does not mean prescriptive. RELAY only ever draws on what your teachers provide, and the school decides how tightly it is drawn, which is exactly why it isn’t an off-the-shelf subscription.
Before it leaves the school
Every student is a nickname.
No names, no identifiers, nothing that could point back to a child.
What we hold
No personal information. No model training.
Student data is never used to train models, and no personally identifying information leaves your school. If your ICT manager wants the detail before the second meeting, they should have it.
The research underneath
Every design decision
traces to a finding.
RELAY is built from published learning science rather than from assumptions about what AI should do in a classroom. Each choice below links to the work it came from.
No grade, no judgment Black & Wiliam, 1998
Formative assessment produces some of the largest measured gains in education, but only where feedback is separated from grading. Attach a mark and students attend to the mark rather than the feedback. Removing the grade is what makes the feedback do any work.
Black, P. & Wiliam, D. (1998). Assessment and Classroom Learning.
Built to make itself unnecessary Zimmerman, 2002 · Torrington et al., 2024
Self-regulated learning is learnable, not innate. Students move towards it through scaffolding they eventually stop needing. RELAY is designed as that scaffold: guided questioning early, withdrawn as a student demonstrates they can do the work themselves.
Zimmerman, B. (2002). Becoming a Self-Regulated Learner. Torrington et al. (2024).
The teacher stays in the loop McCaslin & Hickey, 2001
Co-regulation describes the adult support that precedes self-regulation. It is a relationship, not a feature. This is why RELAY routes a student’s confusion to their teacher rather than resolving it alone: the point is to prompt the conversation, not to replace it.
McCaslin, M. & Hickey, D. (2001). Self-regulated learning and academic achievement.
Why the relationship is the target Van Bergen et al., 2023 · Burns et al., 2022
Student-teacher relationship quality predicts engagement, behaviour and achievement. It is also uneven: a substantial minority of students describe their relationship with their teacher as difficult or distant. Making a student’s thinking visible gives a teacher something concrete to respond to.
Van Bergen et al. (2023). Burns et al. (2022), Journal of Adolescence.
Guardrails change the outcome Bastani et al., PNAS 2025
In a randomised trial with nearly a thousand secondary students, a carefully designed, guardrailed tool produced a substantial lift in performance. Unrestricted access did not. How a tool is designed to behave with a student decides whether it helps.
Bastani et al. (2025). PNAS
The full argument, including the evidence that did not make the final cut, is set out in our position paper.
Request the position paperSafe with student data
Built for a school
to say yes to.
The questions an ICT manager or a deputy principal will ask, answered before they have to ask them.
Your data doesn’t train our models
Nothing a student or teacher writes in RELAY is used to train a model. Not by us, not by anyone downstream.
No personally identifying information
Students are nicknames. No names, dates of birth or contact details leave your school.
Only your classroom’s material
RELAY works from the syllabus, notes and rubrics your teachers supply, not a generic model answering from the open internet.
Pilot programme
See if RELAY
fits your school.
We’re working with a small number of K–12 schools on our first pilot. Pilot schools work directly with the EQV team at a discounted rate, with pricing agreed before any work begins.
We're looking for schools that:
- Are grappling with how to implement AI responsibly
- Have leadership willing to be involved in the process
- Want outcomes grounded in learning science, not hype
- Are open to collaborative design, not just off-the-shelf tools
Start a conversation
Tell us about your school and we’ll come back to you. No form to fill in beyond a name and an email.
Talk to us about a pilot