Can a school let an AI assistant handle student and parent data?

The hard part is not the data, it is who is asking

A clinic knows who its patient is. A school receives messages from two parents who may not agree, a grandparent, a tutor, and the student themselves - about the same child. An assistant cannot verify any of that from a chat window. So anything specific to one student - attendance, marks, behaviour, whether they were in class - must be refused and routed to staff who can check entitlement, no matter how plausible the asker sounds.

What is masked before anything leaves, and what is not

Structured identifiers are replaced with tokens at intake, before any model is called: national ID numbers, telephone numbers, card numbers. The originals are held encrypted and the replacement is reversible by the school. Names written in free text are NOT detected automatically - so "is Daniel in the Thursday group?" carries a child's name through, which is exactly why that question has to be refused rather than merely masked.

Published facts are safe, and they are most of the volume

Term dates, timetables, fees, what to bring, how to join mid-term, the lateness rule, how a course is assessed. None of that is about any individual, all of it is already decided, and together it is the bulk of what families write in to ask. A school gets most of the benefit from the half of the problem that carries no risk at all.

Three questions to ask before signing

What the assistant refuses to answer about an individual student, verbatim. What is masked before the text leaves, and what is not. Which sub-processors receive it, under what agreement, and whether a hosting jurisdiction can be agreed - it can, per school, on request. A vendor who will not describe the refusals is selling something a school cannot deploy.