The Litigant Brought a Team of Agents to a Tribunal

A worker brought AI-generated legal arguments into Australia's Fair Work Commission, aimed them at the wrong legal question, ignored repeated warnings, and left owing his former employer $1,230. Another worker told ABC News he used a team of AI agents like a software build system, checked the citations and logic, and won a narrow employment case against Macquarie University.

Episode 59 is about the gap between those two stories. Not AI versus no AI. Supervised AI versus oracle AI. Legal work versus legal-looking text.

The Fair Work Commission's new generative-AI guidance, published August 24 and taking effect October 20, does not ban AI in filings. It makes the human reappear: who used the system, how they used it, who checked the facts, who checked the law, and whose words are in the document.

The tribunal found the load

The Commission says its total workload increased by more than 70 percent in three years, a rise it links principally to increasing use of generative AI by potential litigants. Its commissioned research, prepared by Pivot, surveyed 408 applicants and 211 respondents. Approximately 40 percent of surveyed applicants reported using generative AI to prepare or manage their case; among those AI users, approximately 77 percent used ChatGPT and about 60 percent used the free tier.

Those numbers cut both ways. AI lowers the cost of entry for people shut out by money, language, fear, or procedure. It also lowers the cost of filing work that may be wrong, overconfident, irrelevant, or impossible to test until someone else untangles it.

The access tool creates an access burden.

Khan is the warning label

Sadnan Khan filed an unfair dismissal application against ALDI. The problem was simple and fatal: he had not served the required six-month minimum employment period. Deputy President Michael Easton wrote that Khan's AI-generated materials focused on the wrong date — the date the dismissal took effect, rather than the date he was notified.

Before the hearing, the Commission explained why the case was unlikely to succeed, encouraged Khan to discontinue, and warned that costs could follow. Khan's AI-generated reply again addressed the wrong end date. ALDI warned it would seek costs. Khan continued anyway.

At the hearing, he conceded he had not completed the minimum employment period and discontinued. Easton then ordered Khan to pay part of ALDI's legal costs: $1,230. The decision says Khan relied heavily on artificial intelligence to start and continue the claim, and acted unreasonably by continuing after it should have been clear, on his own version of events, that he could not win.

Easton's closing sentence is brutal because it is also procedural: Khan's AI arguments were “just plain wrong.”

ABC News later quoted Khan saying, “The main thing AI suffers is they do things not the Aussie [court] way.” He also said he planned to use Claude, ChatGPT, and other tools for a future appeal. The pattern survived the cost order.

Baker is the other path

Gregory Baker's case points in the other direction.

Baker, a computer science lecturer at Macquarie University, challenged his casual employment status in the Fair Work Commission. The official decision confirms the outcome: Commissioner Crawford ordered Macquarie University to treat Baker as a 0.1 full-time-equivalent part-time employee, averaged over 12 months.

The official decision does not say Baker used AI. That part comes from ABC's interview with him.

Baker told ABC he used a team of AI agents and treated the claim like software development: “I created a repository for my source code and my programs, which were my [tribunal] filings and came up with a build process that it checked.” Then he described the checks: “Are all the citations correct? Is everything logically coherent?”

He said asking ChatGPT as a kind of oracle with no context produced “a terrible job.”

That is the hinge. A chatbot can produce fluent legal language in both cases. The difference is whether a human has built a verification process around it.

The new rule is really an ownership rule

From October 20, the Fair Work Commission's guidance will require parties who use generative AI to prepare a Commission document, beyond spelling, grammar, or formatting, to state that use and explain how.

Parties must check that facts and evidence exist, that legal authorities support the positions claimed, and that extracts and quotes come from the sources named. For witness statements and declarations, the witness has to confirm the document reflects their own knowledge, words, and truth.

The plainest line is the most important one: the checking must be done by a person. You cannot check an AI filing by asking the same AI, or another AI, whether it is right.

The guidance says noncompliance may mean documents get less weight or are disregarded. It may lead to costs. It may lead to dismissal.

The old friction did invisible quality control

This is where one Australian tribunal becomes a broader story about legal institutions under machine-assisted load.

The old intake system used friction. Bad claims sometimes died before lodgment because the person could not write the application, navigate the procedure, afford help, or got advice that the case was weak. That filter was ugly. It excluded people with real claims. It also did quality-control work the institution did not have to name.

Generative AI breaks that filter. It can turn confusion into a formatted submission, a weak claim into a formal letter, and a wrong deadline theory into something with headings, citations, and a tone that sounds like law.

The Commission's research says document quality has become a degraded signal. That means someone else has to do the checking.

Sometimes that someone is the tribunal. Sometimes it is the employer on the other side. Sometimes it is a front-line advice worker trying to pull the applicant back from a claim the chatbot made sound winnable. The Pivot report quotes one respondent describing the new dynamic as “AI fighting AI.”

Key points

The Fair Work Commission's guidance begins October 20 and requires disclosure when generative AI is used to prepare a Commission document beyond spelling, grammar, or formatting.
Parties must check that facts, evidence, legal authorities, extracts, and quotes actually support the positions claimed.
Witness statements and declarations must reflect the witness's own knowledge, words, and truthfulness.
Khan's $1,230 costs order is an August 19 case proof, not evidence that the October 20 guidance already applied.
Baker's AI-agent workflow is sourced to ABC's interview; the official Fair Work Commission decision is used only for the legal outcome.
The Commission's research draws a useful distinction between GenAI-assisted users who verify outputs and GenAI-dependent users who treat the system as a quasi-authoritative advisor.
The access-to-justice promise is real, but access to justice cannot mean access to a legal-sounding machine that confidently walks a person into a wall.

Listen to Episode 59

Episode 59, "The Litigant Brought a Team of Agents to a Tribunal", is live now.

Download the episode or subscribe to the show feed.

Sources

The show sent source questions to the Fair Work Commission and Professor Genevieve Grant at Monash University on August 30, then sent follow-ups on August 31. No reply, bounce, human-route request, listener tip, or episode-relevant source response had arrived by the final pre-publication sweep on September 1.

Send source tips, corrections, or field notes to [email protected]. If you work in a court, tribunal, legal-aid service, union, employer response team, or community-law clinic and you're seeing AI-generated filings change the work, use the subject line AI filings. Anonymous notes and source-protection requests are welcome.