AI marks the reflection. The teacher keeps the final word.
The fear isn't that AI gets it wrong. It's that it quietly takes over. Here's the line we drew, and why every score Skilly proposes is only ever a proposal.
The Skilly Team
When teachers first see an AI propose a level for a pupil's reflection, the reaction is rarely "wow." It's a flinch. Is this thing about to decide things about my pupils that I can't see or change?
It's the right instinct, and we designed around it from day one.
The principle: surface, never decide
Skilly never assigns a final score. It reads each reflection, proposes a level on a transparent five-point rubric, and drafts feedback. Every one of those is a draft the teacher can accept, edit, or override with a single tap. The score the pupil eventually sees is always the teacher's — the AI's proposal is just a starting point that saves the cold-start effort of a blank page.
The same holds for wellbeing. If a reflection contains a signal — low mood, a possible disclosure, peer pressure — Skilly raises a flag. It does not act on it. It doesn't message a parent, it doesn't open a case, it doesn't decide anything. It puts the signal in front of the responsible adult and gets out of the way. The judgement, and the duty of care, stay where they belong.
Why this matters more as the AI gets better
It would be easy to argue that as the models improve, we should let them do more. We think the opposite. The better the proposal, the more important it is that a human still owns the decision — because the cost of automation bias rises exactly when the machine is usually right. The one time it misreads a pupil is the time that matters most, and that's precisely when an over-trusting workflow fails.
So the line stays drawn. Skilly does the reading and the paperwork. The teacher keeps the final word. Always.
See it in your own classroom.
Try the live demo, or book a 15-minute walkthrough with the team.