This study examines the applicability of artificial intelligence (AI) in public administration by developing an AI-based employment subsidy recommendation prototype. Policy documents were structured into a dataset using Gemini 2.5 Pro, and the prototype’s perceived usability and accuracy were validated through a pilot test with labor attorneys and SME workers, followed by a survey of 205 SME owners and HR managers. Results showed substantial difficulties with the existing application process, while the AI prototype received high evaluations in trust, perceived usability, administrative efficiency, and expected impact on application rates. These findings provide empirical evidence that LLM-based administrative support can reduce procedural burdens and improve accessibility, demonstrating its potential to support more personalized and efficient public services.
Cho et al. (Sat,) studied this question.