Powered by
MOMENTUM
MEDIA
While AI can be a useful resource, effective self-representation requires legal knowledge, critical judgement, and an understanding of court procedures that extend well beyond a subscription to an AI platform, writes Madeleine Porter.
Gregory Baker recently broke new ground by securing the first Fair Work Commission win to consider the federal government’s casual conversion laws, which commenced in 2024. The decision found that Baker was entitled to a permanent part-time position at the university, with his minimum hours averaged across the year because he had worked consistently during semesters for three years. While some public commentary has celebrated the case as evidence of the demise of lawyers and the rise of self-represented litigants, I take a more cautious view. This outcome should be seen as the exception, not the norm.
Enthusiastic and emboldened pro se litigants should think carefully before bringing potentially fruitless or unsubstantiated claims to an already overwhelmed court system. A high level of caution should be exercised when engaging with AI platforms for AI-generated legal advice, as a probabilistic algorithm falls short of replicating the expertise, judgement, and compassion that legal professionals provide.
So why is Gregory Baker the exception rather than a new ‘norm’?
As a computing academic at Macquarie University, Baker has expertise in the use and deployment of AI, particularly agentic AI. He prompted and directed the agents with a precision most self-represented litigants simply could not replicate. He directed a team of AI agents that not only identified relevant case law and developed mitigating arguments and rebuttals but also consistently verified citations and sources.
Baker also used a range of leading AI foundation tools, including Claude and ChatGPT, and reportedly relied on a US$200 per month ChatGPT Pro licence. Again, this is not the profile of an average claimant, and such licences may be financially out of reach for many Australian self-represented litigants. Finally, the legal argument itself was narrowly focused, reducing the margin for error. The forum also mattered: the claim was brought in the Fair Work Commission, a tribunal rather than a court, and therefore a comparatively more user-friendly setting.
The reality of AI-generated claims and self-represented litigants
Indeed, a growing number of recent articles and cases tell a different story. They suggest that the recent wave of AI-generated or AI-assisted claims has contributed to baseless applications that, in some instances, border on contempt. Three recent decisions illustrate the risks and pitfalls of relying solely on AI while dismissing or minimising the expertise and judgment of legal professionals.
First, in Ba v Sterling Parts Australia Pty Ltd [2026] FedCFamC2G 1245, Judge Symons of the FCFCOA dismissed an application after inferring that the applicant had inappropriately used AI to produce a key originating process document. The case is a useful example of what can happen when AI-powered assistants “masquerade as legal advisors”, mislead both the litigant and the court by amplifying unmeritorious arguments. Her Honour articulated the danger of relying on AI in place of considered legal judgment: “The sycophantic tendency of large language models to prioritise user approval over truth, can make them especially dangerous and unpredictable unless careful oversight, calibration and verification is applied.”
The next recent case, Ms Wayan Buschman v Good to Great Schools Australia (C2026/5556), was also heard in the Fair Work Commission. Unlike Baker, however, the claimant relied on unverified AI-generated excerpts that proved to be incorrect and outside both the scope of the Fair Work Act and the jurisdiction of the Fair Work Commission. The decision is a timely reminder to read the whole case, not merely the “Cliff Notes” version – or an AI-generated summary.
Most recently, the Fair Work Commission took a firmer stance on the inappropriate and immature use of AI by making the rare decision to order costs against an applicant who relied too heavily on it. In Sadnan Khan v Aldi Pty Ltd [2026] FWC 3144, deputy president Easton found that Khan had constructed his unfair dismissal case around AI-generated arguments that focused on the wrong date, leaving him three days short of the minimum six-month employment period. Despite repeated warnings from the commission that his case was doomed and that costs might follow, Khan continued to submit AI-drafted responses rather than engaging with the correspondence himself. Easton ordered him to pay Aldi’s capped costs of $1,230, describing the losses as entirely avoidable.
The difference wasn’t the tool. It was knowing what to ask it, what to ignore, and when it was telling him what he wanted to hear. The lesson for pro se litigants is that while AI can be a useful resource, effective self-representation requires legal knowledge, critical judgement, and an understanding of court procedures that extend well beyond a subscription to an AI platform. It’s a profession built on years of education, experience and knowledge. It is not something that can be replaced with a free, unspecialised, agentic AI tool. For qualified lawyers, our job is to remind our clients of this and the dangers that can arise when complex cases are placed in the hands of AI-made “lawyers” with no prior legal experience. If anything, the cases referenced in this piece highlight just how important and integral the lawyer’s role is in the era of AI.
Madeleine Porter is the global legal vertical lead at iManage.
