AI caseload strains Fair Work tribunal

Australia’s Fair Work Commission is overhauling its procedures after generative AI tools helped fuel a sharp rise in workplace claims, adding pressure to a tribunal already dealing with record demand across dismissal and general protections matters.

The national workplace tribunal has moved to review how applications are filed, screened and managed as self-represented workers increasingly use AI systems to prepare claims, submissions, witness statements and appeal notices. The shift has widened access to the legal process, but it has also increased the volume of weak, incomplete or legally confused filings that still require time-consuming assessment.

Commission President Justice Adam Hatcher has identified AI-assisted litigation as a major factor behind the escalation. Total lodgements, once typically a little above 30,000 a year, rose to about 40,000 in 2023–24 and then to 44,075 in 2024–25. The workload is projected to reach between 50,000 and 55,000 matters in 2025–26, putting the tribunal on course for another record year.

The pressure is clearest in the Commission’s traditional caseload rather than in new powers added through workplace law reforms. From July to December 2025, total lodgements reached 27,192, about 40 per cent above the three-year average. General protections claims involving dismissal rose 62 per cent to 4,464, other general protections disputes jumped 135 per cent to 1,280, and unfair dismissal applications climbed 41 per cent to 9,725.

The rise has placed strain on members and staff, affecting the Commission’s ability to deal quickly with both routine applications and major workplace matters. Its benchmark of issuing 90 per cent of reserved decisions within 12 weeks has slipped to 85 per cent for the financial year to date, marking the first fall below that threshold in many years.

The Commission is preparing changes requiring parties to disclose whether they used generative AI when preparing applications or documents. Applicants who rely on tools such as ChatGPT, Claude, Copilot or Gemini will be expected to confirm that facts, evidence, legislation, case law, legal extracts and quotations have been checked and are relevant. Case law references will also need hyperlinks, and failure to meet the requirements could result in dismissal of an application or a costs order.

The planned disclosure regime reflects growing concern that AI can make poor claims appear legally persuasive. The technology can help an unrepresented worker identify a valid legal pathway, prepare documents within tight deadlines and understand rights under the Fair Work Act. The same tools can also invent facts, misstate law, create false confidence about compensation, or convert an unfair dismissal grievance into a general protections claim when the worker does not meet qualifying-period rules.

Justice Hatcher has cited an internal test in which a chatbot produced a ready-to-file general protections application and witness statement in less than 10 minutes after being given basic facts. The generated material included substantially invented details and suggested a compensation range of A$15,000 to A$40,000 for a matter that did not appear to have reasonable prospects of success.

Appeals are another pressure point. AI-generated notices can list multiple grounds, including procedural fairness claims that require detailed review of the hearing record even when they lack merit. That makes it harder for the tribunal to identify and dispose of cases with little prospect of success, slowing access to justice for applicants with stronger claims.

The Commission’s reforms extend beyond disclosure. New forms for general protections dismissal matters are intended to require clearer articulation of alleged breaches and respondent objections. Some extension-of-time applications with no reasonable prospects may be determined on the papers without requiring employers to respond. Conciliation conferences are also being tightened, with a stronger focus on settlement rather than extended argument about the merits.

The next phase will examine unfair dismissal procedures, where most matters are less complex and may be handled through faster, less formal processes. The Commission is also reviewing how technology can support its own operations, while maintaining its position that generative AI will not make decisions under workplace law. Any decision-making power will remain with a human office holder.

Employers are watching the reforms closely because even unsuccessful claims can impose costs, management time and settlement pressure. Small businesses are especially exposed when AI-assisted applications arrive with legal language that appears sophisticated but may rest on factual gaps or misunderstood statutory tests.