Flow: signals into Bria (early warning)
Bria, the AI early-warning engine, never asks staff to enter anything. It reads attendance, marks and fee balances straight from the source models on a schedule, and turns them into a risk score and interventions.
Exactly what Bria reads
| Signal | Model | Used for |
|---|---|---|
| Attendance | ep.attendance.line | ep.ai.attendance.forecast — expected vs actual rate, dip detection |
| Marks / grades | ep.mark.entry | falling-performance flag; subject strengths/weaknesses via ep.subject |
| Fee balance | ep.fee.assignment | financial-distress flag |
| Career fit | ep.ai.career.family | matched from a student's strong subjects |
It runs on a schedule
Computation is cron-driven (e.g. _cron_refresh_admission_probability, _cron_compute_class_heatmap), so it never slows down a teacher saving a mark. The score is recomputed nightly and surfaced on the at-risk view and the parent portal's alert rules.
Closed loop — Interventions staff take are recorded against the student, so the loop from flag → action → outcome is measurable, not just a one-off alert.
Was this page helpful?