NOVA ECE's AI layer is not a chat box but an assistant doing specific jobs within specific limits. Its most visible function is turning accumulated observation records into something a teacher can read.
A transparent three-step structure
Every insight is presented the same way. Observation: which records it rests on. Interpretation: what those records might show. Suggestion: what could be tried next. The separation lets a teacher distinguish interpretation from raw data — and reject an interpretation they disagree with.
Each insight carries a confidence level. An interpretation built from few records is marked low confidence, and that mark is not hidden. Every output also carries the statement that it does not replace the teacher's assessment.
Consent gate and privacy
- Without a parent's consent for AI processing, the AI layer never touches that child's data.
- Every access is written to the audit log: who, when, for what purpose.
- Insights are not stored permanently; what is recorded is the observation, not the interpretation.
- Diagnosis, labels and child-to-child comparison do not exist in the system.
Intelligence about missing data
The layer's most deliberated behaviour: with no data, it produces no interpretation. “No observation recorded for this child in 21 days” is a record alert, not a developmental judgement. The system never conflates the two — few records does not mean a developmental problem.
The INSIGHTS dashboard
On the administrator side, patterns become visible at institution level: where observation is thinning, which periods see records drop, how family engagement is running. The dashboard does no per-child ranking; it helps an institution see where to direct support.
The measure of this layer is not how much it says but when it stays silent. In early childhood, good AI is AI that says what it does not know.
Let’s begin the journey to every child’s light.
Discover NOVA ECE at your school. Try all three roles in the live demo — nothing to install.