Research
Ardvarq is a research-driven company. The tools we build are grounded in what we are learning about how higher education actually serves its students, and what would have to change for it to serve them better.
What we learned from the CU Boulder pilot
The students in the deepest conversations supplied context, challenged suggestions, changed constraints, and verified the result. That behavior is promising because it keeps the student—not the system—in charge of the decision.
Easy-A language appeared in 19.2% of typed conversations, and those conversations went 1.8× deeper than the rest. The opening question did not always describe the real advising need underneath it.
None of the 22 distress-related conversations in the study produced an advisor referral. A future version cannot simply promise a human handoff; it must recognize the moment, make the referral, and verify that the path is usable.
Study limits
This was one 44-day product pilot with no comparison group. It measured conversations and product behavior, not academic outcomes. GPA data was self-reported or reconstructed rather than registrar-verified, and a larger review did not support our early claim that lower-GPA students engaged most deeply.