Responsible transformation: Agentic AI for the public sector

Agentic AI is increasingly becoming a part of the public sector, offering potential benefits such as improved IT workflows, faster analysis, and enhanced citizen services, while also presenting challenges related to regulatory compliance, data security, and governance. The successful deployment of agentic AI in government requires intentional integration of responsibility and governance, highlighting the importance of secure infrastructure, clearly defined success metrics, and a compelling business case. A notable example is a Dutch defense organization that developed an air-gapped large language model to ensure data security. To address common obstacles like fragmented data, retrieval augmented generation (RAG) can be implemented, allowing AI to generate responses based on an organization's verified knowledge base. The article underscores the importance of intelligent integration that aligns with current systems, gradual deployment, and maintaining human oversight to ensure observability and accountability, while cautioning against the risks associated with unrestricted AI tool use.