Skip to main content
DoneThat

AI Adoption GuideEducationCredential

Graduation Requirement Gap Alerts

ML identifies students approaching graduation with unmet requirements and triggers proactive advisor alerts before registration closes.

Education processRecruitAdmitEnrollTeachAssessCredentialGraduateAdvance

By Don, DoneThat’s AI coach · updated

Cite the unmet catalog row and remaining terms

A graduation requirement gap alert is usable only when an advisor can open it and see a specific unmet row from the student's posted degree audit, plus how many terms remain before the coded conferral term. Name the catalog requirement identifier and the requirement title exactly as the catalog and the audit display them. State the posted status for that row. State remaining terms from the stored expected conferral term or from an approved academic plan. If those fields cannot be produced from posted records, do not send an alert.

The quality outcome is a citation, not a course pick. The process reads the posted automated degree audit, lists rows that are still unmet, and attaches remaining terms so the advisor can judge whether the next registration cycle is still a realistic chance to attempt the requirement. It does not guess the course the student usually still needs. It does not fill a blank row with a popular substitute. Empty is the correct output when the audit is incomplete, unposted, or internally unresolved.

Student information platforms such as Ellucian, Workday, and Anthology already hold the catalog and program rows the audit uses. A learning management system such as Canvas can show current enrollments and in-progress grades. Treat each as a system of record for its own data. None replaces a posted audit. Current enrollment is not a met requirement until the audit posts the row as met.

Run the check only on a posted, complete audit

Run the gap check against the posted degree audit: the evaluation the registrar treats as official for that student, program, and catalog year. Do not run it against a draft, a what-if, an advisor worksheet, or a batch that has not been posted. Draft surfaces change as transfer credit evaluation finishes, as petitions clear, and as catalog exceptions are applied. Alerting from a draft is a failure mode. Alerts will cite gaps the official record does not contain, or miss gaps that appear only after posting.

Completeness is a gate. Before any alert is queued, the audit run must be posted, every requirement group must have a resolved status, and remaining terms must be computable from a stored conferral term or an approved plan. If transfer work is still unarticulated, a substitution is pending, a final grade is unposted, or no catalog year is assigned, the evaluation is unfinished. The machine output stays empty. Advisors can still open the student record. The system must not pretend the audit is settled.

Do not mix catalog years. Evaluate each active program against its posted catalog. An unmet row on a deactivated program is not a gap for the credential in progress.

If remaining terms cannot be derived, omit the countdown. An alert that cites the unmet row without a clock is weaker. An alert that invents a term count is false urgency.

How the alert should read for one student

A student in Arts and Sciences is coded for Spring conferral, with two terms remaining on the academic plan. The posted audit shows college and major groups as met except one catalog row: WI-300, Upper-division writing intensive, status unmet, with no in-progress attempt mapped to that row. The alert to the assigned advisor names the student, cites WI-300 as the catalog and audit display it, states that the row is unmet, and states that two terms remain before the coded conferral term.

That is the full payload. The alert does not recommend a section, an instructor, or a substitute. It does not enroll the student. It does not place a registration hold. The advisor owns the plan: take the departmental writing course next term, petition a substitution, or move the conferral term.

If WI-300 were in progress on the posted audit, the gap alert would not fire for that row. If the audit were still a draft waiting on transfer composition credit, the alert would stay empty rather than claim WI-300 was unmet. If several rows were unmet, the alert would list each catalog identifier and title. It would not collapse them into a single note that the student is still short on general education. Collapsed language is how invented gaps enter the queue, because the advisor cannot see which row is real.

Remaining terms are a scheduling clock. Two terms remaining means another registration cycle may still exist. Zero terms remaining with an unmet posted row means the advisor needs the citation before the current window closes. After that window, the coded conferral term is not reachable without an exception. The process still does not invent a course to close the row.

Notify the advisor before registration closes

Generate the alert because of the calendar, not because of a confidence score. The advisor needs the cited unmet rows while the student can still register for a course the audit would accept. Queue the notification against the institution's registration calendar for the next open term, for students whose remaining-term count is at or below the threshold the registrar sets.

Send the alert to the advisor of record for the active program. Copy a registrar work queue only when local policy says that office, not the advisor, routes exceptions. Do not send the student a machine-generated warning that they will fail to graduate. Remaining requirements already appear on the posted audit the student can view. Whether to discuss them, and in what language, is the advisor's conversation.

An alert that arrives after the window closes cannot change enrollment. An alert that arrives weeks early from a draft audit trains the office to ignore the queue.

Non-enrollment in Canvas can corroborate that the student is not currently in a course the audit maps to the unmet row. Absence from the learning management system is not proof of a catalog gap. The gap lives on the posted audit.

If the advisor later uses course placement recommendation to choose among eligible sections, that is a later step. Placement is not the gap alert. Combining them is how pressure to auto-register starts: the process that cited WI-300 unmet begins selecting a section. Stop at the citation.

Do not invent a gap, auto-register, or treat the alert as a hold

Three failure modes make this process unusable.

Alerting from a draft audit. A what-if or in-progress evaluation is not posted. Articulations, grades, and petitions will change the rows. Gate on posted and complete, or output empty.

Inventing a gap. If the audit cannot resolve a row, the catalog identifier is missing, or remaining terms cannot be derived, output empty. Do not fill the blank with a high-enrollment course or a historical substitute. Those fillers are not catalog rows. They become advising errors when the student was already waived, substituted, or completed by transfer. Do not invent a missing course.

Treating the alert as a hold. A gap alert is a message to the advisor. It is not a registration restriction, not a diploma block, and not an automatic change to the conferral term. Holds, when the institution uses them for missing requirements, are a registrar policy instrument with their own approval path. The advisor still owns the plan, including the decision that no hold is warranted.

Do not auto-register. Closing a cited row is a human choice among courses, petitions, and term changes. Auto-enrollment from a gap list collides with time conflicts, campus rules, and the student's other remaining rows. The alert's job is to put the unmet catalog row and remaining terms in front of the advisor before registration closes.

When the posted audit later shows the row as met, clear the alert. When the student reaches conferral, digital credential packaging should see a clean posted audit, not a residual gap flag that survived because alerts were treated as permanent holds.

Is this worth automating for you?

Whether this pays back depends on how much time it takes your team today. Most teams estimate that from memory, and the estimate is usually wrong in one direction or the other.

DoneThat reconstructs where the time actually went, with no timers to forget, so you can measure the baseline before committing to a project and check the gain afterward.

Measure the baseline first