Better Armed But Not Better Advised: AI And The Rising Cost Of Employment Claims

AI is changing the economics of employment litigation.

For employers defending claims under the Fair Work Act, the issue is not just that more self-represented applicants are using generative AI (which they are), or that the volume of claims is rising (which it is). It is that AI is changing how those claims are presented, pursued and defended.

Claims that may once have been short, informal or obviously weak can now arrive as lengthy, structured and apparently sophisticated legal documents. Applicants can generate pleadings, submissions, evidence and interlocutory material in minutes, at little or no cost. They can also run every available argument, however marginal. If an applicant asks the model to find them arguments, it will find many arguments for them.

The result is a new imbalance in litigation. AI may reduce the cost of bringing and pursuing a claim, but it can significantly increase the cost of defending one.

Better armed, but not better advised

A good lawyer does more than build arguments: they exercise judgement. That means knowing which points are worth running, which evidence will hold up, when a procedural step is proportionate and when settlement is the smarter course. This important role of the lawyer is recognised by the overriding purpose that guides the conduct in civil litigation – the just, quick and cheap resolution of claims.

GenAI can help with advocacy. In fact, it can be very good at this. It can identify arguments, impose structure and produce material that looks convincing. But it does not bring the same level of judgement or strategic insight. It does not assess credibility in any meaningful way. It does not understand the forensic risk of over-pleading. It does not weigh the commercial value of taking a minor point of law. And it will not tell an applicant, with the authority of a trusted adviser, that the case is weak or that a commercial settlement should be considered. That is the real distinction.

A self-represented applicant using AI may be better armed, but they are not necessarily better advised. The technology can give them more arguments to run, without the discipline to know which ones to drop.

That is where the cost problem starts. Litigation once imposed its own friction. Parties had to think carefully about which issues were worth pursuing. However, AI removes much of that friction for self-represented applicants. There is little practical incentive to narrow the dispute, while employers must still assess, triage and, where necessary, answer each point.

AI in the hearing room

The issue is not confined to documents filed in proceedings. In a recent matter, we suspected that a self-represented applicant was using a GenAI tool in real time to formulate oral responses during the hearing itself. When the court directed the applicant to stop using AI technology, there was a noticeable change in the structure, sophistication and legal framing of the applicant’s responses (leading us to believe our suspicion was correct). Others have reported similar experiences.

That raises obvious questions about the integrity of proceedings, as well as practical questions for courts, tribunals and employers.

Courts and tribunals are starting to regulate the use of AI in litigation, and the guidance is evolving quickly. The Federal Court has issued Practice Note GPN-AI, which sets out the Court’s expectations for responsible use of GenAI, including that disclosure may be required if the Court directs it. However, the use of AI during oral hearings appears to be a new frontier. This is particularly problematic where hearings are conducted virtually.

The Fair Work Commission has just issued its Guidance Note on the use of GenAI, effective 20 October 2026, which requires applicants to disclose if and how they have used GenAI to prepare their application.

The Fair Work Commission has also shown a willingness to intervene where AI use affects the conduct of a dispute. In a recent stop-bullying matter, the Commission made orders limiting use of generative AI in correspondence between the parties. This was on the basis that the employee’s lengthy, repetitive and accusatory language (thanks to AI and the prompts it was no doubt given) resulted in the employer failing to meaningfully engage with the concerns raised by the employee. The point is not that AI has no place in workplace communications, but rather that AI-generated correspondence can escalate a dispute if it is not tempered by human judgement and a good dose of old-fashioned common sense.

A cost asymmetry

Fair Work Act proceedings operate in a “no-costs” jurisdiction. In most cases, unless there are exceptional circumstances, each party bears their own costs. That position reflects an important premise: workers should not be deterred from bringing genuine claims because they fear an adverse costs order.

But AI is putting pressure on that framework. If an applicant can generate lengthy legal materials at no cost, but employers must pay lawyers and internal stakeholders to triage and respond, the cost burden shifts. A weak claim can still become expensive to defend.

The Federal Circuit and Family Court of Australia recently addressed the tension created by no-costs employment litigation. In Park v Health Insurance Comparison Choosewell Pty Ltd (No 2) [2026] FedCFamC2G 1344, Judge Vasta observed that the no-costs jurisdiction “does not mean that any applicant has a ‘free hit’ at an employer, in effect, trying to get the employer to give them, what is colloquially known as, ‘go-away money’ to settle an unmeritorious claim, simply because it would cost the employer far too much to defend such a claim.” His Honour emphasised that the no-costs framework exists to protect applicants with genuine claims, not to shield those who exploit the cost disparity inherent in the system.

That principle has obvious relevance in an AI-assisted litigation environment, where it is easier than ever to bring claims at little or no direct cost. There is a real risk that the influx of AI-assisted claims will make it harder for the system to focus on genuine claims. Matters that should resolve early may instead be pushed to hearing, putting further pressure on court and tribunal resources.

Just last month, in Sadnan Khan v Aldi Pty Ltd [2026] FWC 3144, a case handled by our firm, the Fair Work Commission took the “very rare step” of issuing a costs order against an employee who “relied heavily on artificial intelligence”, requiring him to pay some of his former employer’s legal costs. Deputy President Easton found that the employee’s decision to pursue his claim, despite clear and specific advice to the contrary, was ill-advised and unreasonable. The decision sends a clear message that the Commission may accept AI as a drafting tool, but blind reliance on AI advice will not be an excuse for pursuing a claim that is plainly flawed.

Looking ahead

The debate about GenAI in the legal system is already well underway, and it will not be resolved quickly. The Federal Court has recently announced a research project into unrepresented litigants and the administration of justice, with the aim of informing how the justice system responds to these issues. In the meantime, there are practical steps employers can take now when managing AI-assisted claims:

• Triage on substance. A well-structured AI-generated submission is not the same thing as a strong claim. Employers should identify the real issues and respond proportionately.
• Do not be drawn into every point. Self-represented applicants may run every available argument. Employers should resist the instinct to answer every weak, irrelevant or unsupported contention in full.
• Prepare for procedural overreach. Be ready for unnecessary subpoenas, amended claims, additional evidence and interlocutory steps. Deal with them firmly, but proportionately.
• Build the costs record from day one. Keep a clear record of unreasonable conduct, unsupported allegations, fabricated or irrelevant citations, rejected offers and unnecessary procedural steps.

In the spirit of full disclosure, AI lent a hand in preparing this blog post. However, the technical judgement, opinions and scepticism are entirely ours.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.