Conference Agenda

This paper investigates the institutional readiness of public administrations to engage with artificial intelligence technologies, focusing on the challenges frontline bureaucrats face as AI tools increasingly enter administrative work. While governments worldwide are incorporating AI into policy agendas and investing in digital capacity-building initiatives, far less empirical attention has been paid to how these technologies are encountered and interpreted within bureaucratic practice. The Stanford AI Index 2025 reports that artificial intelligence has become an expanding focus of government activity: since 2016, 204 AI-related laws have been enacted globally, while mentions of artificial intelligence in legislative proceedings have increased more than ninefold. At the same time, AI tools—particularly large language models—are increasingly used in routine administrative tasks such as drafting documents, retrieving information, and supporting policy analysis. As a result, the practical governance of AI is shaped not only by national regulation but also by how frontline public officials interpret and use these technologies in practice.

This paper therefore asks: How prepared are frontline public administrations to engage with emerging AI tools, and what institutional challenges shape how public officials interpret and use these technologies in governance practice?

These questions are particularly important in emerging economies, where governments face strong incentives to adopt AI-enabled governance while confronting constraints related to institutional capacity, technical expertise, and digital infrastructure. Research on algorithmic systems in policymaking shows that computational models can enhance analytical capacity and policy coordination, but their usefulness depends heavily on how they are interpreted and embedded within organizational contexts (Kolkman, 2020). Similarly, empirical studies of AI adoption in government show that AI technologies are often used to improve public service delivery and administrative management, while raising governance challenges related to accountability, transparency, and decision-making processes (van Noordt & Misuraca, 2022). These findings suggest that effective AI governance depends on the institutional readiness of public administrations and the ability of public officials to critically engage with these systems.

The analysis is grounded in three complementary theoretical perspectives. First, theories of state capacity emphasize that effective governance depends on the competence and institutional capability of bureaucratic systems. Second, the concept of street-level bureaucracy highlights that frontline administrative officials exercise discretion in interpreting policy directives and translating them into administrative decisions (Lipsky, 1980). Third, research on algorithmic decision-support systems suggests that the benefits and risks of AI depend on how human decision-makers interpret, trust, and incorporate algorithmic outputs within institutional settings (Kolkman, 2020). Together, these perspectives suggest that AI governance is not only a regulatory challenge but also an administrative one, shaped by the institutional environments in which public officials encounter and use these technologies.

To investigate these dynamics, this paper conducts an empirical study of administrative officers in Bangladesh using a mixed-methods research design combining survey data, qualitative field research, and institutional analysis. A nationwide survey of approximately 455 administrative officers across central ministries, district administrations, and upazila-level offices examines digital competencies, familiarity with AI tools, and perceptions regarding their potential role in administrative decision-making and public service delivery. The survey is complemented by focus group discussions, ethnographic observations, and key informant interviews with policymakers and training officials, enabling the study to examine how routines, administrative hierarchies, and institutional environments shape how AI tools are encountered within public service practice.

Preliminary findings highlight several institutional challenges shaping AI engagement within the public sector. Public officials increasingly encounter AI-enabled tools within administrative workflows, yet many report limited structured guidance on appropriate use in public sector contexts. Organizational factors—including hierarchical decision-making structures, risk-averse bureaucratic cultures, and limited opportunities for experimentation—shape how such technologies are interpreted and integrated into routine administrative work. In addition, uncertainty regarding data governance, accountability, and responsibility for AI-assisted decisions creates hesitation about the role these tools should play in policy analysis and service delivery.

These challenges do not reflect deficiencies among individual public officials. Rather, they reveal a broader institutional gap between the rapid diffusion of AI technologies and the slower development of governance frameworks, training systems, and administrative guidance within public sector institutions. By examining how frontline bureaucracies encounter AI technologies in practice, this paper contributes to research on digital governance by shifting attention from national AI strategies toward the institutional realities of administrative systems responsible for implementing them. The findings suggest that effective AI governance depends not only on regulatory frameworks but also on the institutional readiness of public administrations—including the knowledge, organizational support, and governance structures that enable public officials to engage with AI systems critically and responsibly in public decision-making.