AI-Powered Student Admissions Validation for an Italian University

We built a UiPath robot that works the admissions process like a member of the admissions team: it validates each application end to end, hands the cases it cannot complete to staff, and reports the rest.

How every application is validated

1. Rules from the admissions team. We collected the validation rules from the admissions staff and their knowledge of the process, and turned them into explicit, testable checks.
2. Portal automation. Unattended UI automation logs into the admissions portal, opens each application and reads the submitted form and uploaded documents, for every branch on the portal.
3. API integration. The branch on the API-based admissions system is handled by the same automation through its API.
4. Document understanding. UiPath Document Understanding classifies the uploaded document and extracts its data: ID cards, passports, driving licences and tax ID documents, each in the several formats accepted in Italy.
5. Form-to-document matching. The extracted ID data is compared field by field with what the applicant entered in the form.
6. Data validity. Regex patterns and type- and format-specific rules check that each value is real and well formed, and that the document type and format are acceptable.
7. Photo validation, built in-house. Our own Python algorithms check the photo against the portrait criteria and match it to the face on the ID document.
8. Human in the loop. When the robot cannot extract every field or information is missing, the case goes to admissions staff in UiPath Action Center, and the results are reviewed at the end of the process. The robot never guesses.
9. Exception reporting. Anything wrong or missing is flagged to the university in a regular Excel report, so staff contact only the applicants who need it.

Built to run every intake

• Orchestrated from UiPath Automation Cloud, with the robots on virtual machines inside the university's own network.
• Designed and built to UiPath best practice.
• We administered the Automation Cloud tenant and the robots' Windows infrastructure.
• Support, upgrades and maintenance of the robot and its infrastructure from 2022 through 2025, with admission-process corner cases handled as they appeared.

Technology

The original project documentation is lost, so the figures below are approximate: working figures and recollections from the delivery team, not audited measurements.

• Around 20,000 admissions validated a year across two intakes, from 2022 to 2025: more than 60,000 applications over the engagement.
• Around 90-95% of applications handled correctly by the robot. The rest, and every case with missing information or fields it could not extract, went to people.
• The manual checking moved to the robot. Staff now handle only wrong or incomplete cases; the delivery team describes the savings as enormous, though they were never formally measured.
• One validation standard for every applicant, every intake, across all branches and both channels.
• 3+ years in continuous use, supported, upgraded and maintained by us throughout.

The risk most teams miss

An admissions robot fails at the worst possible moment: the first week of an intake, when volume spikes and a portal change, a new document format or a Windows update on the robot's machine stops it cold. Our support covered the robot and the infrastructure it ran on, and corner cases were handled as they appeared, so each intake started on a working robot.