This guide supports New South Wales (NSW) government agencies to safely and responsibly use agentic artificial intelligence (AI). AI agents use models, tools and data to identify tasks, make decisions, take action and learn over time to achieve goals. With agentic AI, the human’s role shifts from completing tasks to managing AI agents that do some tasks for you.

The guidance will help NSW agencies use the technology responsibly, including defining the need for human intervention, undertaking regular reviews of its output, and applying privacy and security controls. It provides guidance to support them to:

• consider their policy position
• identify potential use cases
• set clear ownership and guardrails before launch
• run safe pilots and plan carefully when scaling to production
• understand the unique risks of this emerging technology.

Key points

• Agentic AI is more than chat
• Agents extend generative AI capabilities
• Assign ownership and guardrails upfront
• Standards are still developing.