Australia’s Fair Work Commission is using governed Agentic AI to automate its IT operations, accelerate case reporting and prepare new public-facing services, demonstrating how government organisations can deploy Artificial Intelligence rapidly while maintaining human oversight. Joshua Pope, Chief Technology Officer at the Fair Work Commission (FWC), tells us more.
Australia’s Fair Work Commission has moved Agentic AI from procurement to production in a matter of weeks, using the technology to automate IT operations, accelerate case reporting and lay the foundations for public-facing AI services.
The deployment comes against a particularly sensitive backdrop for government automation in Australia. The Robodebt scandal, which resulted in hundreds of thousands of unlawful debt notices, a US$1.8 billion settlement and a Royal Commission, demonstrated the potential consequences when automated systems operate without sufficient governance and human oversight.
For Joshua Pope, Chief Technology Officer at the Fair Work Commission (FWC), that history has shaped how the organisation approaches Artificial Intelligence.
“Governance is really hard,” Pope said. “There was a very public bad experience with AI making choices on behalf of people, without a human in the loop. So, for us, the ability to check, validate and verify at every step was non-negotiable.”
Moving beyond business-hours service
The FWC is Australia’s national workplace relations tribunal, handling issues including unfair dismissal, workplace disputes and minimum wages.
Pope believes a significant proportion of people approaching the Commission could receive the information they need through agentic systems rather than having to wait for staff operating during conventional business hours.
Introducing such technology into government, however, brings significant compliance requirements. Security standards, IRAP certification, data sovereignty and rules governing automated decision-making all have to be addressed before systems can move into production.
FWC therefore needed an AI platform that could operate within Australian government security requirements while allowing its relatively small technology team to develop applications quickly.
The Commission selected Workato and took an unusual approach to implementation. Before the contract had even been signed, Pope insisted that his team receive practical platform training.
“I specifically said I don’t want to see a single sales pitch as part of the onboarding week,” he said. “Just show me how to build stuff.”
FWC moved from completed procurement to onboarding within a week, with its technology team attempting to build and deploy its first Workato AI agent, known as a Genie, on the first day.
The Commission has since put an IT Support Genie and License Optimisation Genie into internal production, creating a repeatable framework for deploying additional agents under established governance controls.
Pope also wants applications developed by FWC to provide reusable models for other public sector organisations. “Build once, reuse infinitely,” he said. “That’s the idea.”
AI takes on hearing-room monitoring
One of the clearest examples of the technology’s operational impact can be found across FWC’s hearing-room infrastructure.
The Commission operates around 59 purpose-built hearing rooms across Australia, supporting virtual conferences and physical hearings involving judges, lawyers, paralegals, applicants and respondents.
Around 50 hearings take place each day across Australian states and territories, making the availability of the underlying technology critical. A technical failure can potentially disrupt or delay proceedings.
Previously, FWC’s ICT team had to access four separate portals to check the status of equipment. The process was largely manual, could normally be completed only once each day and created the possibility that problems developing later would remain unnoticed until reported by users.
The IT Support Genie now performs checks several times a day across the systems and produces a consolidated report identifying affected rooms, device problems, polling health and domain status.
“We went from checking once a day across four different portals to running checks multiple times daily through a single agent,” said Kris Deep, Infrastructure Manager, Fair Work Commission. “The Genie catches issues before our users do and that changes the entire dynamic for my team.”
The agent has also been applied to other routine ICT processes.
Provisioning overseas conditional access has fallen from around 30 minutes to five minutes, according to the Commission, while automated onboarding and offboarding can save between 30 minutes and an hour for each employee.
The system can also help diagnose connectivity problems by identifying competing device connections without requiring the ICT team to investigate manually.
Case reports cut from an hour to two minutes
FWC has also begun applying agentic automation closer to its core tribunal operations.
One of its latest agents was developed in a week to address the preparation of investigation reports.
Associates previously spent at least an hour retrieving information from multiple systems, including Microsoft Dynamics, then assembling the material manually. Individual cases could involve more than 30 files that needed to be downloaded and reviewed separately.
The new workflow uses Workato to retrieve information directly from multiple sources and consolidate case notes from applicants and respondents into a master report.
FWC said the process now takes around two minutes, representing a roughly 30-fold improvement over the previous one-hour workflow.
“Our associates used to spend at least an hour pulling case information from multiple systems and compiling it manually,” Deep said. “That now takes two minutes. When you’re dealing with people waiting on outcomes for unfair dismissal or workplace safety, that kind of time saving has a direct impact on how quickly we can serve them.”
For a tribunal handling dispute that can have significant consequences for individuals and employers, the objective is not simply to make back-office processes faster. Reducing administrative workloads could also shorten the time required to progress cases.
Avoiding the productivity trap
Pope argues that AI productivity projects need to go further than simply calculating the number of employee hours saved.
Instead, FWC is attempting to decide in advance how employees should use the capacity released by automation.
“I’m very careful with giving time to people without a specific directive,” Pope said. “We’re not giving time back for people to decide what to do with it – they could fall into the trap of doing the same old things. We’re replacing that time with things we’ve already identified.”
That approach changes the calculation around AI deployment. Rather than treating automation primarily as a headcount or cost reduction exercise, FWC is looking at where human expertise can provide greater value once repetitive administrative work has been removed.
The strategy also reflects the Australian Public Service AI Plan 2025, which set out ambitions to give public servants access to Generative AI tools while putting appropriate governance, assurance and training around their use.
FWC’s internal deployments provide an early indication of how that approach could work operationally.
Public-facing agents next
Having initially used internal systems as a proving ground, the Commission is now preparing to expand Agentic AI into services used directly by the public.
A contact centre agent is already under development.
Pope initially expected frontline employees to be concerned about automation entering this part of the organisation. Instead, he said employees were interested in removing the routine work involved in triaging enquiries so they could concentrate on helping people with more complex problems.
“Those at the frontlines said, no, triaging calls is the last thing they want to be caught up with,” he said. “They wanted to focus on actually helping the people they connected with.”
For Pope, this highlights an important distinction in calculating the return on AI investment.
“Your better ROI comes from removing friction points in people’s day jobs, not from where you think you can see cost savings,” he said.
FWC is also participating in the development of Workato’s Data Genie, with Pope acting as a design partner and providing feedback on government requirements ahead of wider availability.
The Commission’s experience suggests that deploying Agentic AI rapidly in government does not necessarily require bypassing governance or compliance. Instead, its approach has been to establish those requirements early, train teams before implementation and prioritise applications that can be configured and reused.
“Find what you can reuse first,” Pope said. “Don’t get stuck researching where your internal ROI is and commissioning a bespoke solution every time. Research where you can pick something up and just configure it for your use case.”
For government organisations still weighing how to introduce Agentic AI safely, FWC provides an interesting test case. Its most significant achievement may not be that an hour-long process can now be completed in two minutes, but that it has demonstrated how AI applications can move into production quickly while retaining human oversight and governance.
