YDUQS
Generative AI screening of admission cover letters
YDUQS' generative AI accepted or rejected admission cover letters correctly more than 90% of the time
Problem
Applicants' cover letters for simplified enrollment were read by staff. That took time and added screening costs.
Approach
The Brazilian higher-education group started work on Google Cloud's Vertex AI in October 2023 and ran proofs of concept within weeks. It built a solution on the PaLM 2 model that screens cover letters and added it to the website used for simplified enrollment. Staff no longer work on the screening full time and instead give feedback to improve the model.
Results
Takeaway
The group counted a letter as a success when the model accepted or rejected it correctly. The company said other models it compared scored below 80% and took up to 11 seconds.
Categories
- Size
- Large enterprise
- Industry
- Education
- Business function
- Production and quality
- Innovation type
- Process innovation
- AI technology
- Generative AI
- Adoption
- Off-the-shelf tool or platform
- Evidence
- Vendor case study
Sources
Source summaries
Google Cloud's customer story describes how YDUQS used Vertex AI and PaLM 2 to screen admission cover letters. It reports a success rate above 90% and a 4-second average response, and says other models scored below 80% and took up to 11 seconds. The page carries no date; the Internet Archive first saved it on May 22, 2024.
Full details are in the original sources.