evidoria

← Back to browse

Good practice Imported

ChatGPT-Based Virtual Admissions Office — Three Peruvian Public Universities Automate Applicant Support

Peru · Lima · See the Peru profile · See the Lima profile

Evidence: Observational / pre–post Top 89% 27/100 · Ask Evidence Copilot about this practice

Researchers at three Peruvian public universities built the first ChatGPT-based virtual admissions-office assistant, cutting response time to under two minutes (92% accuracy) and lifting Net Promoter Score from 25 to 70 across students, parents and applicants.

over 80 %
Response time reduction versus traditional channels
92 %
Response accuracy
95.2 %
Sensitivity
75 %
Specificity
25 to 70 NPS
Net Promoter Score, before and after
86 %
Student satisfaction
ChatGPT-Based Virtual Admissions Office — Three Peruvian Public Universities Automate Applicant Support

Details

Maturity
Pilot
Promoter
Universidad Nacional Mayor de San Marcos
Period
2024–2025
Keywords
higher education, admissions, conversational AI, public university

Context

Peru's public university admissions offices traditionally rely on phone calls and in-person visits, creating long waits at peak enrolment periods. Researchers from Universidad Nacional Mayor de San Marcos, Universidad Nacional Pedro Ruiz Gallo and Universidad Nacional Intercultural de la Amazonía built what they describe as the first ChatGPT-based conversational system for a public university's admissions office.

Activities

The assistant handles applicant, parent/tutor and external-inquiry traffic; in the reported deployment, students made up 72% of users, parents/tutors 18% and external applicants 10%.

Results

Measured results included response times under two minutes (an over 80% reduction versus traditional channels), 92% response accuracy, 95.2% sensitivity and 75% specificity, and a Net Promoter Score that rose from 25 to 70; satisfaction was 86% among students, 75% among parents/tutors and 78% among external applicants.

Conclusions

The 75% specificity figure signals a non-trivial misclassification rate on some query types, and the study covers a single deployment cycle at one institution's admissions office, with no independent replication reported yet.

Implementation
Indicative cost
Low (< €50k)
Time to results
Short (< 1 year)
Staffing & skills
Universidad Nacional Mayor de San Marcos, Universidad Nacional Pedro Ruiz Gallo, Universidad Nacional Intercultural de la Amazonía

Conditions for success

  • single ChatGPT-based assistant serving three distinct user groups (students, parents/tutors, external applicants)
  • response-quality metrics (accuracy, sensitivity, specificity) tracked and published

Common failure modes

  • 75% specificity indicates a non-trivial rate of misclassified queries on some question types
  • results come from a single deployment cycle at one institution with no independent replication reported yet

Where it fits

Governance type
public university consortium (research-led deployment)
Scale
single admissions office, one deployment cycle
Income level
upper-middle-income country (Peru)

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful