An early alert system that opens a case only when two independent sources agree, cites every warning sign to its record, and sends nothing until the coach approves
Every student at risk reached within two days, each warning sign cited to its record.
See one case, screen by screen ↓Early Alert & Student Retention is an early alert system for college advisors and success coaches. Agents read instructor flags, LMS logins, the gradebook, holds and attendance every night, open one retention case per student only when two independent sources agree, and draft the outreach and the fix for each blocker. The student's own coach approves every message, and the office that owns a blocker signs its release.
AI in education: a buyer's guide to agentic AI for colleges, universities and school districts →
Updated
An early alert system is only as good as the signals it puts together, and at most colleges those signals sit in five systems owned by five offices. A student stops logging in to the learning management system, scores 54% on a midterm, and finds a library hold that blocks spring registration. The progress report lives in the advising CRM, the logins in the LMS, the grade in the gradebook, the hold in the student information system, the attendance in the card-scan system. Instructors flag unevenly, coaches carry caseloads near 300, and by the time someone puts the facts together the student has stopped coming.
Two things make it worse. One flag is a weak signal and a strong label: many tools open a case and send an automated message on a single missed lab, so students learn to ignore the messages and faculty stop flagging. And the blocker belongs to another office, so a $380 charge that Student Accounts could move to a payment plan in an hour waits while the coach finds the policy and emails the office. Early Alert & Student Retention puts eight agents in front of that work: they watch, corroborate, cite and draft, and the student's own coach decides.
Estimated split for one student whose warning signs come from three sources, by hand and with the agents, in business days. The 'after' column is about a day and a half, inside the two-business-day first-contact standard; our estimate, kept modest.
Five Watchers read their own sources, a Corroborator opens a case only on two, a Resolver drafts the fix; the coach approves.
Jordan Teague, Brandvale ID 0418236, is a first-year pre-nursing student on North Campus who prefers Spanish. Here is what happens on Wednesday, September 30, 2026, week 6 of the fall term, screen by screen, in the working solution.
One Watcher per source read 54,210 LMS events, 31 instructor flags, 18 new holds and 9,840 attendance scans. The Corroborator opened 38 cases across the college, 5 on Renée Okafor's caseload, and merged 3 duplicate signals. The West Campus attendance feed did not arrive at 2:00 AM, so those strands are drawn as unknown.
The Attendance Watcher never reports a missing feed as present or absent; it reports unknown.
112 single-source signals opened no case, and each says why. Brianna Castillo has one BIO 141 flag and every other source fine. Mateo Ibarra has a MAT 121 flag and a MAT 121 gradebook reading, but both come from the same instructor, so rule R-06 counts them as one source. Every row reads "Nothing sent".
In week 6, three of Jordan's five strands went bad, from three independent sources: no LMS login for 9 days, 54% on the BIO 141 midterm worth 20% of the grade, and a $380 library replacement charge that blocks spring registration. One missed lab is only watched, and the week-5 progress reports say "No concern". Under rule R-01 v3 the strands knot and the case opens.
Renée clicks the citation LMS-LOG-0418236 and the access log opens read-only, with the cited line highlighted: logins on September 17, 19 and 21 in BIO 141-03, the Lab 4 report submitted on the 21st, then nothing.
2026-09-22 → 2026-09-30: no sessions for 9 days (rule R-02: 7 days during term)
The Resolver found Student Accounts policy SA-7, which lets a library replacement charge under $500 move to a payment plan with the registration hold released while the student has an open success case, and drafted the request. It wrote Jordan a text in Spanish, his preferred language from the student information system, with the English beside it. The checks under the draft: no grade, no balance, consent on file, outside quiet hours.
The Resolver never sends a message, releases a hold, or puts a grade or a balance in a text.
Renée reads the three cited records and approves the text and the hold-release request with reason code OA-1, approved as drafted. The two-way texting platform delivers it at 10:14 AM, and Jordan replies at 10:31 AM. Renée books Thursday, October 1, at 2:00 PM.
Sí, el jueves a las 2 está bien. Gracias (Yes, Thursday at 2 works. Thanks)
The coach can ask; only Student Accounts can release. Priya releases the hold under SA-7 and opens payment plan PP-26-0912, three payments of $126.67, the first due November 15, before spring priority registration opens on November 2. Jordan's first contact came 4 hours after his case opened.
Prof. Ortiz sees only his own sections and his own gradebook. When he flags Jordan for low grades, he is told what happened in plain words, without seeing the case. His other flags this term show their outcomes: watched, a coach is in touch, closed.
Recorded. Your flag and your own gradebook count as one source (rule R-06), and a coach is already in touch.
On Monday, September 28, evening sections opened cases at 11.4%, 1.87× the all-student rate of 6.1%, over the 1.25× threshold, so the Equity agent paused rule R-07 for them. The cause: West Campus evening rosters entered 3.4 days late on average. Marcus Bell's proposed fix, R-07 v5, was replayed on weeks 1–5: 56 of 141 evening cases would not have opened, and the ratio falls to 1.13×. Only the dean, Dr. Teresa Lindqvist, can approve it.
Switched to Grace's role, every Approve, Release and Confirm button is gone. The trail for September 30 runs from the Watchers' readings at 2:10 AM through the Corroborator, the Resolver, Renée's approval, the delivery, Jordan's reply and Priya's release, and the evidence pack with the FERPA disclosure log downloads in one click.
The same week-6 morning, seen by the five people who carry it, from the coach's caseload to the census file: what their day looked like, and what it looks like now.
Five Watchers read their own sources, a Corroborator opens a case only on two, a Resolver drafts the fix; the coach approves.
Reads instructor progress reports and early flags every night, one strand reading per student per week, with the report cited.
Reads LMS access logs and submissions against the term calendar, about 6 million events a term, citing the log line.
Reads gradebook entries with each assessment's weight from the syllabus.
Reads student information system holds, the aid document checklist and next-term registration barriers.
Reads card scans and paper rosters for every section.
Opens one case per student when two independent sources go bad in the same week, and merges duplicate signals.
Drafts outreach in the student's preferred language with the English beside it, and finds the fix for each blocker under the owning office's policy.
Every Monday compares case-open rates by student group, pauses a rule over 1.25× the all-student rate, replays proposed fixes on past weeks, and measures persistence against the census.
Approves every message; Student Accounts signs every release. The agents propose; a named person decides.
Ask about any student, strand, rule or office on your caseload, or say what to change. Every answer points to its records, and a change comes back for the right person to approve.
Why did Jordan Teague's case open?
Three independent sources went bad in week 6. LMS: no login for 9 days, last session Sep 21, 11:47 PM. Grades: 54% on the BIO 141 midterm, worth 20%. Holds: a $380 library charge blocking spring registration. Attendance is only watched (one lab) and the flags are clear, so they did not count. Rule R-01 v3.
Which of my cases need me today?
Seven. Six drafts to approve: Jordan Teague, Amara Nwosu, Luis Ferrer Ortega, Kayla Brooks, Minh Anh Tran and Ricardo Salinas. One closure to draft: Abdi Hassan, back on track. Ricardo's first contact is due today at 5:00 PM; his first draft was rejected Monday for length.
Why is rule R-07 paused for evening sections?
Evening sections opened cases at 11.4%, 1.87× the all-student 6.1%, over the 1.25× threshold. West Campus evening rosters are entered 3.4 days late on average, so present students read as absent. Marcus Bell proposed a 72-hour wait; the dean approves it after a replay.
Which holds on my caseload can be released under policy?
Two are open. Jordan Teague's $380 library charge under SA-7 (payment plan, hold released) and Hannah Okoro's $210 parking citations under SA-4. Sofia Marchetti's and Abdi Hassan's holds were already released. Student Accounts signs each release, not you.
13 screens from the working solution, on its sample data. Pick one to see it large.
The clocks, the messaging rules and how far the agents may go are settings, changed only by the administrator, each change saved as a new version and on the record. The signal rules themselves belong to the dean.
| Setting | Default | Choose from |
|---|---|---|
| First contact | 2 business days from the moment a case opens | Stepped by one business day |
| Escalate to the Director of Advising | When first contact is 3 days late | Stepped by one day |
| Hold-release requests answered by Student Accounts | 1 business day | Stepped by one business day |
| Withdrew closures confirmed by the Director of Advising | 3 days | Stepped by one day |
| Equity threshold | 1.25× the all-student case-open rate | Stepped by 0.05×; a rule above it is paused for that group |
| Agent confidence threshold | 0.85 or 0.90, set per agent | Below it, a person reads the record |
| Texts | Consent on file; no grade and no balance | Both locked by policy; quiet hours 8 PM–8 AM on or off |
| Case records kept | 5 years after the student's last term | Disclosure log locked on; legal hold; redact names on export by default |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
Sources: National Student Clearinghouse Research Center, Persistence and Retention reports (69.1% of the fall 2024 cohort still at their starting institution in fall 2025; community college retention 55.0% for fall 2022 starters and 45.7% for part-time starters, June 27, 2024) · NACADA, median advising load of 296 students per advisor · FERPA, 34 CFR 99.31(a)(1) and 99.32 · Title IV return of funds, 34 CFR 668.22. Rules R-01 to R-08, policies SA-4 and SA-7 and reason codes OA and RJ are the working solution's sample rules, not any one college's. "demo" = seen in the working solution, on its sample data · "target" = design goal, measured in the live solution · "typical" = published figures · "estimated" = our split of the days. Brandvale Community College, its campuses, students and the people named on this page are fictional: characters and sample data in the working solution.
Early Alert & Student Retention is an early alert system for college advisors and success coaches. Agents read instructor flags, LMS logins, the gradebook, holds and attendance every night, open one retention case per student only when two independent sources agree, and draft the outreach and the fix for each blocker. The student's own coach approves every message, and the office that owns a blocker signs its release.
An early alert system brings together the warning signs that a student may leave, such as a progress-report flag, days without an LMS login, a low grade, a registration hold or missed classes, so an advisor can reach the student while there is time to help. In this solution each warning sign is a strand cited to its record, and a case opens only when the strands from independent sources agree in the same week.
No. A single-source signal is watched and nothing is sent; on the sample day 112 were watched overnight and every one can be listed with its reason. Under rule R-06 an instructor's flag and the same instructor's gradebook count as one source, and a feed that fails is drawn as unknown rather than read as fine.
It aims to reach every student at risk within the two-business-day first-contact standard, in the student's preferred language, with the blocker already routed to the office that owns it. In the sample data, 79.8% of fall 2025 pilot students reached within two business days persisted to the spring census, against 70.2% of those reached later; the screen labels this a comparison, not proof of cause.
It can, which is why the Equity agent compares case-open rates by student group every Monday. A rule that flags a group at more than 1.25× the all-student rate is paused for that group, its cause is shown, and a proposed fix is replayed on past weeks before the dean decides. No agent can re-enable a paused rule, and no protected characteristic is used as a signal.
The student information system, the learning management system and its gradebook, the attendance system, the advising CRM, a two-way texting platform and the student accounts (bursar) module. Each source shows its last run and state, and a failed feed is shown as unknown, never fine.
People decide. Agents watch, corroborate, cite and draft; the coach approves every message with a reason code, Student Accounts releases every hold, the Director of Advising confirms every "withdrew" closure, and the dean approves every rule change. A rejected draft goes back to the Resolver with its reason, such as RJ-2, tone or length.
The Agentic Solution Engine builds and deploys it from your requirements and documents: your signal rules, first-contact standard, office policies, message rules and role assignments. It goes live once every quality gate has passed, connected to your student information system, learning management system, attendance system, advising CRM and texting platform.
Answers at any hour, cited to the student's own catalog year, and never a promise.
1-min demoAcademic Recovery PlansEvery probation student signs a plan with checked math before registration closes.
1-min demoClinical Placement Compliance TrackingStudents cleared before day one, every document tested against the site's own clause.
1-min demoWe'll run Early Alert & Student Retention on a sample of your own progress reports, LMS activity, gradebooks and holds.