VyCla — A WhatsApp NLP Chatbot Cuts Enrollment Response Times From 11 Hours to 75 Seconds at Universidad de Nariño, Colombia

At Universidad de Nariño in Colombia, the Office of Control, Academic Records and Admissions (OCARA) handled undergraduate enrollment and admissions questions manually, with an average response time the study later measured at 666.04 minutes — over eleven hours — per query, straining a small staff during peak enrollment periods.

The university's team set out to cut response times and improve the student/applicant experience for enrollment and admissions queries using a natural-language-processing WhatsApp chatbot.

The team built VyCla, a natural-language-processing chatbot integrated with WhatsApp, using an agile Scrum process across five sprints and a stack of Node.js, Python (Flask, NLTK) and PostgreSQL. The underlying intent-classification model was trained on 13,996 phrases spanning 21 intents covering common enrollment and admissions questions. An initial diagnostic phase surveyed 4 university staff, 105 enrolled students and 45 prospective applicants to map real information needs.

In the evaluation phase, 122 users interacted with VyCla, generating 462 resolved inquiries; the study reported 99% model accuracy, a mean user-satisfaction score of 4.23 out of 5, a mean recommendation score of 4.07 out of 5, and an average response time of 1.25 minutes, a 99.81% reduction from the manual baseline. The work was published in the peer-reviewed Revista Colombiana de Tecnologías de Avanzada (Vol. 1, No. 47, 2026).

The comparison baseline (666 minutes) reflects OCARA's own prior manual process rather than an external benchmark, and the evaluation sample (122 users) is a fraction of the university's full applicant pool, so the reported figures describe a successful pilot rather than a fully scaled, independently audited deployment; the authors frame further extension to other administrative processes as future work, not something already proven.