Akademik
By e-pondok team
Student Anomaly Summaries: AI That Flags the Decline Before the Semester Ends
e-pondok's weekly AI summary: stalled memorization, slipping attendance, and rising clinic visits surface early enough for real guidance.
End-of-semester report cards show who failed. By then the problem is old: memorization that stopped three weeks ago, attendance that eroded slowly, or a student visiting the clinic again and again had been signaling since the early weeks. The data was always there. Nobody had time to read it person by person.
The Asisten menu in e-pondok does that reading on demand and presents the result as a short brief of actionable findings.
How it works, plainly
The anomaly summary does not guess. It compares this week against the previous weeks on four fronts:
- Stalled memorization: students with no new setoran for weeks while their target remains open.
- Declining attendance: this week’s pattern worse than the student’s own recent baseline.
- Piled-up gate exits: leave patterns outside the student’s usual rhythm.
- Rising clinic visits: a climbing clinic frequency can signal health trouble.
Every finding names the student, the delta against the prior period, and a severity. Because detection runs on fixed rules, a teacher can verify each finding by opening the student’s history; clicking the student’s name opens their record directly.
From finding to guidance
The summary does not replace guidance; it orders the queue. Without it, mentors review students from memory or from whoever complains loudest. With it, the weekly mentoring meeting starts from a short list: the three to five students whose data declined most this week, with the direction of the problem attached.
The follow-up stays human: ask how they are, check the dorm, talk to the class teacher, invite the parents. What changes is the time-to-detection, from end of semester to the same week.
AI narration, governed numbers
The narrative paragraph is written by AI, but the numbers behind it come from fixed rules. If the AI service is unavailable, the brief still prints as a plain findings list. That keeps the summary honest: nothing invented, no name appears without data behind it.
Per-pondok settings also control the context length and topics the assistant may not touch, so the brief stays aligned with the pondok’s own policy.
Summary
Your pondok’s data already contains the early signs of every student problem. The anomaly summary reads those signs early: stalled memorization, falling attendance, clustered gate exits, and rising clinic visits, presented as a list you can carry straight into the mentoring meeting.
The feature is part of the Pro plan. See it on the Academic AI Assistant page, read more in the academic AI assistant, or start a subscription to use it at your pondok.