Using Generative AI Well: AI Skill Training and School-to-Work Transitions

How to cite

Brenøe, Anne; Thieme, Albert. 2026. "Using Generative AI Well: AI Skill Training and School-to-Work Transitions." AEA RCT Registry. AEARCTR-0017760. https://doi.org/10.1257/rct.17760-1.1

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Anne Brenøe* , Albert Thieme

* Primary investigator

Registry ID
AEARCTR-0017760
DOI
https://doi.org/10.1257/rct.17760-1.1
First registered
Last updated
Status
On going
Trial dates
2026-01-26 – 2028-12-31
Countries
Switzerland ()
JEL codes
I21, J24, O33, C93
Registry record
http://www.socialscienceregistry.org/trials/17760

Abstract

This study evaluates whether a brief training in responsible and effective use of generative AI tools improves students’ apprenticeship application materials, performance in mock interviews, and subsequent transitions after compulsory school in Switzerland. Schools are cluster-randomized to either receive a standard application training only (control) or to additionally receive a 90-minute generative AI workshop delivered before the standard training (treatment). Outcomes are measured via student surveys, standardized volunteer ratings of application materials and mock interviews, and linkages to education registries.

Intervention

Control: standard YES application training.
Treatment: 90-minute generative AI workshop delivered before the standard YES training (plus the standard training).

Experimental design

We implement a field experiment in collaboration with our project partner Young Enterprise Switzerland (YES). Our study population consists of students in the last two grades of lower secondary education as well as students in bridge programs after having completed lower secondary education. Recruitment of classes occurs via YES school registrations plus direct outreach by the research team. We randomize schools into control and treatment with 50% in each group in separate waves (the timing depends on when students fill out the baseline survey).

All participating classes receive a visit by a YES volunteer to cover their standard application training. Treatment classes additionally receive a 90-minute generative AI workshop before the standard application training.

Design details

Randomization method
Randomization done in office by a computer.
Randomization unit
School
Clusters
74 schools
Observations
2200 students in around 148 classes.
Arms
37 schools control, 37 schools treatment
Primary outcomes
1) Quality of application materials (CV and motivation letter), based on standardized volunteer ratings. 2) Mock interview performance, based on standardized volunteer ratings. 3) Upper-secondary placement.

Cited by

No citation data available for this registration yet.

Keywords