For teachers and professionals

This page is for teachers, SENCOs, and the professionals who support pupils — educational psychologists and occupational therapists. It explains what to capture, how the AI drafts, why you approve every word, and the rules that keep every pupil's folder safe. Demonstration · synthetic data only — no real children.

Draft content 5 item(s) marked "to be confirmed" by the programme office.

Who this page is for

Teacher / SENCO is one login. A SENCO sees a wider group of pupils and is where any alert marked *needs a look* goes next. Educational psychologists and occupational therapists review sensitive outputs as the humans in the loop. Professional decisions stay with you; the AI offers examples only.

Support is everyone's job (the Whole School Approach). That is why the Platform has seven views, one per role: teachers capture, the school office manages consent, leadership sees school health.

The model is simple. A pupil is a folder. An evidence item is a page filed into it. A report is a bound stack of chosen pages. Consent decides who may open the folder.

What to capture, and how

An evidence item is an everyday moment, not an assessment. It takes one of three forms: a note, a photo, or a short clip (demo clips up to 20 seconds). Filing one is designed to take about fifteen seconds, one-handed on a phone, with the AI off. A dated professional note is an *observation record*. Offline capture is not in the demo yet.

Notice a moment
A strength, a small win, or something you want to remember. Everyday moments belong here — there are no scores or grades anywhere.
Choose the form
Type a note, take a photo, or record a short clip.
Tag a goal (optional)
Link the item to a framework goal so progress can be followed by area over time — against the pupil's own past, not norms.
Set who may see it
Two flags: *child sees* and *share with family*.
File it
The page joins the pupil's timeline — the record the next teacher inherits.

From evidence to a draft

The journey has six steps: Capture · Insight · Draft · Approve · Share · Oversee. This page follows the four in the middle — where the AI helps, and where you stay in charge. The AI reads only the pupil's own folder, the approved framework and the knowledge bases set up for it. It never invents evidence. Every summary links the records it used, shows where its words came from, and shows how confident it is. You can always write it yourself instead. Missing data is shown as missing, never filled in.

Insight
For a clip, the system derives gentle, non-diagnostic signals — for the teacher's eyes only. These are derived signals, not raw video. **Provisional:** the detection model is still to be specified; the demo shows mock outputs (*spec pending*).
Draft
A small AI drafts a strengths-based report from the evidence — nothing invented. The draft says how many items it is grounded on and cites its sources.
Approve
The teacher edits and approves every word before anything is shared. Reject a suggestion and it is discarded; nothing is saved until you accept. An accepted observation carries the chip *AI-assisted · teacher approved*.
Share
With consent, the report reaches the family — clearly marked AI-assisted, human-approved. Reports are designed for bilingual output (English and Traditional or Simplified Chinese), with checks that the meaning stays the same, and print to PDF.

Why you approve every word

The portfolio comes first, and the Platform never diagnoses. The AI is a helper on top of trustworthy portfolio records — never the product. Every report a family sees is marked *reviewed and approved by the teacher*, so your judgement has to be in every word. In plain terms, here is what the AI will and will not do.

It suggests — it never decides. — Support tiers, plans and interventions are yours to decide. The evidence record informs the 3-Tier Support Model; it never assigns a tier.
It never diagnoses, labels, or scores a child. — No diagnosis, no decision on who qualifies for support, no risk score. Outputs are worded as *possible association* or *example*.
Nothing it writes reaches a student or parent until a teacher approves it. — Every output goes through understand → correct → reject → trace.
Every AI action is logged. — Each output records the model, prompt and version used.
It only reads the folder. — It works only from the pupil's own records; it never makes up evidence or citations.
It speaks respectfully. — Its wording is strength-based and neurodiversity-affirming, following an approved bilingual terminology guide.
It can be switched off. — Every core workflow works with the AI disabled. A switch for each feature hides every AI button and suggestion; the portfolio carries on.

Training and resources

The plan includes in-product help, online learning for teachers, and an AI-security workshop for SENCOs. Completion is recorded, and an acknowledgement is kept in the audit trail. The materials themselves are to be confirmed. The cards below are placeholders — content, format and timing are to be confirmed by the programme. No dates are given on purpose.

In-product help

Placeholder — to be confirmed. Help inside each screen.

Online learning for teachers

Placeholder — to be confirmed. Self-paced modules; completion is recorded.

SENCO AI-security workshop

Placeholder — to be confirmed. Completion is recorded; an acknowledgement is kept in the audit trail.

Guidance for EPs and OTs

Placeholder — to be confirmed (whether it is offered is also to be confirmed). How to review sensitive outputs as the human in the loop.

The audit trail

Every action that matters leaves a trace. The audit trail records every AI call, consent change, share and alert decision, and each time an AI feature is switched on or off. Each AI output records the model, prompt and version used. Entries carry student ids, not names, and can be filtered by event.

An accepted suggestion becomes a teacher-approved observation with the chip *AI-assisted · teacher approved*. A shared report carries a fixed footer, *AI-assisted · reviewed and approved by the teacher*, that cannot be edited. When you handle an alert — resolve, escalate, or not relevant — your decision is recorded. A concern follows a set route to the school or a professional; the AI never acts on it.

Data rules for staff

The law we start from is the Personal Data (Privacy) Ordinance (Cap. 486). All portfolio content is treated as confidential personal data of pupils, teachers and administrators. The rules below are built into the Platform — and they also describe how staff are expected to work.

Collect only what a portfolio needs — There is no field for a diagnosis, a label, a decision on who qualifies, or a risk score. Sensitive SEN details are kept in small pieces, and only the right role at the right school can open them.
Your school only — Each school's records are kept apart, and access is closed by default. You never see another school's records. Any attempt to cross schools or roles is blocked and logged.
Roles are walls — The school office manages roster, staff, classes and consent — metadata only — and has no way into any portfolio. NGO and researcher views show only de-identified totals; small groups are hidden (k-anonymity).
Consent first — Nothing reaches a family until a parent gives consent. Under partial consent, video stays with the school. Withdrawing consent empties the parent view at once. The consent model shown is the delivery team's design — the agreed model is to be confirmed.
Children are assisted users — A pupil reaches the Platform through a supervising adult. The child view is read-only and strengths-only: no inputs, no AI, no numbers that read like grades, no alerts.
Respectful, formative language — No scores, no grades, no comparisons between children. Say *neurodivergent pupils*, never deficit terms.
This demo uses synthetic data — Demonstration · synthetic data only — no real children. Real school data is never used for testing. Data categories, the basis for consent and the wording of notices are confirmed before any pupil data is collected.