Generative artificial intelligence (GenAI) is fast becoming a key tool for boosting productivity and fine-tuning operations. A recent Microsoft survey reveals that 24% of private sector leaders have already integrated GenAI into their workflows, with some even eyeing workforce adjustments in favour of AI systems. Meanwhile, public sector agencies are taking a more measured approach, mindful of the risks involved.
Senior bureaucrats from 22 Australian government bodies shared a range of views on deploying GenAI in policy-making. Some are optimistic, suggesting that the technology can help streamline government functions—as one interviewee remarked, “Why improve the candle when you could use a light bulb?” Others remain cautious, pointing to potential pitfalls such as data hallucinations, biased datasets, and the challenges of handling sensitive information.
There’s a shared understanding among public servants: AI isn’t seen as a replacement for human expertise, but rather as a way to offload routine tasks and free up resources for higher-value activities, like engaging with the community. Additionally, most current GenAI uses are confined to administrative tasks that deliberately avoid sensitive data—a precaution partly shaped by the legacy issues like those experienced during the Robodebt era.
In response, government agencies are rolling out GenAI initiatives incrementally. Senior officials describe these steps as controlled experiments with clear objectives, designed to ensure both value for money and risk mitigation. Some agencies have limited access to freely available models like ChatGPT, instead opting for authorised tools such as Copilot. Nonetheless, the ongoing use of unauthorised GenAI tools by some staff highlights the need for better guidance on safe usage.
While GenAI holds potential for handling tasks at a faster pace, its integration into government work must resonate with Australian values. Public servants find themselves at a crossroads: leverage GenAI to explore innovative policy solutions or proceed cautiously to nurture public trust in new technology. With high expectations on tackling challenges like housing and economic issues, a balanced blend of human insight and machine efficiency could offer the best path forward.
