Date context: arXiv v1 preprint posting date.
A preprint examines a multinational tech employer introducing generative AI into HR search. The authors use search logs, 25 survey responses and 10 interviews. Workers selected tools according to task, role, language and tenure. They checked source documents, compared search systems or asked a colleague or HR when an answer needed confirmation. The study highlights the quality and accessibility of the underlying HR material.
Ali, Rodríguez Velázquez, Horta Ribeiro, Liao and Papakyriakopoulos · 2026-06-16
Primary publisher text checked on 2026-11-06. Scope and limits stated below; independent results not reproduced.
What the source shows
A preprint examines a multinational tech employer introducing generative AI into HR search. The authors use search logs, 25 survey responses and 10 interviews. Workers selected tools according to task, role, language and tenure. They checked source documents, compared search systems or asked a colleague or HR when an answer needed confirmation. The study highlights the quality and accessibility of the underlying HR material. Source: Ali, Rodríguez Velázquez, Horta Ribeiro, Liao and Papakyriakopoulos, 2026-06-16
Limits and reading context
This is a single-organisation case with a small voluntary survey and interview sample. Its observations help frame deployment questions. The design cannot establish a general adoption rate or a causal productivity effect. Source: Ali, Rodríguez Velázquez, Horta Ribeiro, Liao and Papakyriakopoulos, 2026-06-16
For HR, AI + JJ
AI + JJ, application: Test the same routine question with several roles and languages. Ask each person to locate the governing document, explain what it supports and show where they would turn if the answer were incomplete. Source: Ali, Rodríguez Velázquez, Horta Ribeiro, Liao and Papakyriakopoulos, 2026-06-16