Abstract

The disconnect between technological design and the needs of frontline civil servants has become a major challenge hindering the development of government artificial intelligence (AI) in public sectors worldwide. Why do policies regarding government AI applications diverge from the practical needs of frontline civil servants? In-depth interviews were conducted with 11 civil servants from various departments at different levels in Shanghai, eastern China, and Shaanxi province, western China, yielding 57 AI application scenarios that civil servants urgently expect, which were systematically compared with the scenarios outlined in central government policy. The findings show that 57.9% of frontline need scenarios are covered by central government policy, with unmet needs being heavily concentrated in two domains: system integration and personal assistance. The gap between policy blueprints and frontline voices stems from the careful assessment of three constraints during the policy formulation process: resource scalability, risk controllability, and barriers to organizational coordination. By elevating scenarios from a policy term to an object of institutional analysis, this study reveals the rational logic underlying the policy choices of government AI application scenarios, offering an analytical framework applicable to public sectors worldwide in planning for AI adoption.