OpenAI runs AI workshop for public servants in Canberra

OpenAI held a Build a Bot workshop for 60 public servants in Canberra as part of Public Sector Innovation Month.

Participants used ChatGPT, ChatGPT Work and Codex to test how artificial intelligence tools could be applied to everyday public sector work. With support from OpenAI technical staff, they identified problems from their own roles and built early prototypes for drafting, research, coordination and service delivery tasks.

The workshop was run with Public Sector Innovation Month and AI CoLab. It moved attendees from introductory discussion to practical testing, with a preparatory enablement session followed by in-person work on potential solutions.

The event reflected growing interest across government agencies in how generative AI could be used in administrative and operational settings. The focus remained on human oversight, with privacy, sensitivity and review built into the exercise.

Brent Thomas, Head of Policy and Global Affairs for Australia at OpenAI, said the aim was to give public servants time to test the technology against real workplace issues.

"We were thrilled to have been invited to contribute to Public Sector Innovation Month. Some of the best ideas emerge when people move beyond talking about AI and start using it to support their work," Thomas said.

He said the workshop focused on practical tasks rather than abstract debate.

"This morning was about giving public servants time and space to build practical tools that can help them do their work better - improving administrative tasks, improving workflows and creating more space for the work that matters most - delivering services and outcomes for the Australian public.

"The opportunity here is very practical. Many public servants are dealing with large volumes of information, complex processes, internal coordination, drafting, research and service delivery challenges. If AI can help reduce some of that friction, that can free up their time and attention towards higher-value, higher-impact work," Thomas said.

Human oversight

During the session, attendees worked alone and in teams to define day-to-day challenges, explore AI-based responses and share the results with the wider room. The exercise was designed to show where the tools might help in routine workflows while keeping final judgement with staff.

Thomas said that principle would remain central as interest in AI grows in government and business.

"Responsible use stays human-led. That means using information appropriately, being thoughtful about privacy and sensitivities, and keeping human judgement at the centre of key decisions," Thomas said.

OpenAI will continue to support participants through two office-hours sessions after the workshop. Up to three individuals or teams may also receive one-to-one coaching and mentoring to further develop the most promising ideas.

Local focus

The Canberra session also points to OpenAI's broader effort to expand its presence in Australia. It is growing a local team focused on working with businesses, government bodies, founders and institutions on practical uses of AI.

Public Sector Innovation Month is an annual series led by volunteer members of the public sector innovation community. The programme brings together public servants to exchange ideas, share examples of new approaches and examine how emerging tools may fit into government work.

The Build a Bot format placed less emphasis on theory and more on direct experimentation. By asking public servants to start with actual bottlenecks in their jobs, the workshop tested whether generative AI could help with repetitive tasks and information-heavy processes without removing human review.

Participants focused on areas where staff handle large volumes of material, coordinate across teams or prepare written work under time pressure. Those use cases have become common starting points for organisations assessing generative AI, particularly where the technology can speed up first drafts or organise information while leaving decisions to trained professionals.

In the public sector, that balance has drawn particular scrutiny because of the sensitivity of government information and the consequences of errors in official work. The Canberra workshop therefore combined practical building exercises with discussion of privacy, appropriate data use and the need for human judgement in key decisions.

Following the event, OpenAI's next step will be to work with a small number of participants to refine the ideas developed in the room, with up to three teams receiving further one-to-one coaching and mentoring.