Advising at Scale
Twelve campuses, 84,000 students, one question: can automated triage reach students who would otherwise never contact an advisor, without displacing the advisors who matter most?
Does LLM-assisted triage increase advising contact among students who do not currently seek help?
Advising caseloads at public institutions routinely exceed 400:1. Automation is arriving regardless; the useful contribution is evidence about where it helps and where it quietly harms.
Key findings
students across twelve participating campuses
advising contact among previously non-contacting students
campuses in the trial, four of them community colleges
About the project
The trial randomises at the student level within campus, with advisors blinded to assignment where operationally possible.
Contact rates among previously non-contacting students rose 34%. Advisor-rated complexity of the conversations that reached humans also rose, which is the intended direction.
We publish the full audit protocol, including the refusal and escalation cases, alongside the results.