The same AI tools that businesses are using to draft HR documents are being used by employees to find the gaps in those documents.
That is not a prediction. It is happening now, across Australian workplaces, at a scale that has pushed the Fair Work Commission to the edge of its operational capacity.
This is Part 2 of a two-part series on AI risk in Australian workplaces. Part 1 covered the compliance risks of using AI to generate employment contracts and HR documentation. This blog covers the other side of that risk — what happens when employees use AI, and what that means for your business.
The two risks are directly connected. A business with well-maintained, current documentation is harder to expose on both fronts. A business relying on AI-drafted documents that haven’t been professionally reviewed is exposed on both simultaneously.
The Numbers Every Australian Employer Needs to Know
The Fair Work Commission has recorded a 70% increase in total workload over three years, from approximately 30,000 matters per year before 2023, to 44,039 lodgements in the first ten months of 2025–26 alone.
FWC President Justice Adam Hatcher has pointed to AI as the principal driver. The historical correlation between retrenchment rates and dismissal applications, which held steady for decades, has broken down entirely. In Justice Hatcher’s assessment, delivered to the Victorian Bar Association in February 2026, the surge “is principally being caused by the increasing use of AI tools by potential litigants“, describing it as “the only reasonable inference which can be drawn” from the pattern of lodgements and the language appearing in applications.
While the Commission has not yet published definitive evidence confirming this causal link, the inference is consistent with the case-level patterns the Commission is observing at scale.
The breakdown by claim type is instructive:
- Unfair dismissal claims: up 41% between 2022–23 and 2024–25
- General protections dismissal claims: up 62% over the same period
- Other general protections disputes: up 135%
For employers, these are not abstract statistics. They translate directly into time, management disruption, legal cost, and settlement pressure, regardless of whether the underlying claim has merit.
Risk One: Employees Are Already Checking Your Documentation
What AI Allows an Employee to Do in Minutes
Before a formal claim is lodged, often before an employee has spoken to anyone, AI tools are being used to scrutinise employment entitlements with a speed and precision that was not available to most workers a few years ago.
An employee can upload a copy of their modern award, their payslips, and their timesheet to an AI tool. Within two or three prompts, they have a detailed comparison identifying every potential entitlement shortfall — underpaid allowances, incorrect classifications, missed loadings, pay rates applied against superseded award rates.
Brittany Byrne, Partner and Solicitor at Citation Legal, described the dynamic plainly: an employee identifies a discrepancy, tells a colleague, who tells another colleague, who tells the union, and a problem that might once have gone unnoticed compounds quickly into something far more difficult to manage.
The Citation Group Workforce Pulse 2026 report (registration required to download), based on a survey of more than 500 Australian business owners and managers, found that 42% of businesses have identified payroll errors, despite 87% believing their payroll is accurate. That gap between perceived compliance and actual compliance is precisely where AI-assisted employee scrutiny operates.
The stakes of that gap have increased materially. Wage theft is now a criminal offence in Australia. Underpayment that was once treated as an administrative error now carries potential criminal liability.
The Documentation Gaps AI Finds First
AI does not get tired, intimidated, or uncertain. It cross-references methodically. The gaps it identifies most readily are the same gaps that have always created Fair Work exposure, they are simply found faster now, and by employees who may not previously have had the tools or confidence to act.
The most common vulnerabilities being exposed:
- Employment contracts that have not been reviewed since the role changed. A contract drafted three years ago for a role that has since evolved, whether new responsibilities, changed reporting lines, different hours, may no longer reflect the actual employment arrangement. That misalignment creates risk on multiple fronts.
- Award classifications that no longer match the duties being performed. Classification errors are the most common source of underpayment. In most cases they are not deliberate, a role was recruited at one level and gradually evolved without the classification being reviewed. The Fair Work Ombudsman recovered a record $358 million in back-pay for more than 249,000 workers in 2024–25. In most underpayment cases, the error was a classification that had drifted out of step with the role, not a calculation mistake.
- Policies that reference superseded obligations. A workplace policy drafted two years ago may reference legislative obligations that have since changed. An employee using AI to check their entitlements will find that inconsistency.
- Performance and conduct decisions that were never documented. This is the exposure that surfaces most painfully when a claim is lodged. Decisions that were made but exist only in a manager’s memory, with no written record to support them.
Businesses without a dedicated HR function are disproportionately exposed. They typically lack the in-house HR capacity to run systematic documentation reviews, and employment contracts, policies, and classifications may not have been professionally reviewed in years.
Risk Two: A Surge in Formal Claims, and What Is Behind It
What AI-Generated Claims Look Like
AI tools can produce a ready-to-file Fair Work claim within minutes. The output is polished, formally structured, and written in legal language — it looks like it was drafted by a lawyer. For many employers, particularly those without in-house HR or legal support, receiving a claim that looks authoritative creates immediate pressure to respond at significant cost or settle quickly.
The Fair Work Commission is seeing the consequences directly.
In Application by Pennisi [2026] FWC 352, the applicant lodged 53 pages of AI-generated forms and submissions in support of an application to file a general protections claim six months late. The Commission noted that arguments were repeated multiple times throughout the material, with the reasoning shifting and evolving with each repetition. The Commission found it difficult to identify the relevant considerations buried within the volume of material, and the applicant relied on non-existent case law. The application was rejected.
In Reece Hoverd v M & J D Pty Ltd [2026], the applicant used AI tools to draft submissions for a general protections claim. Deputy President Nicholas Lake found the applicant had consistently relied on provisions of a contract and an award that did not exist as the basis for the argument. Even after being specifically warned by the Commission not to provide false or misleading AI-generated evidence, the applicant continued. The application was dismissed, and the Commission invited the employer to seek a costs order. If pursued, this would be the first costs order of its kind against an applicant for unchecked AI-generated submissions.
These are not isolated decisions. They reflect a pattern the Commission is now managing at scale.
The Cost of Responding to a Baseless Claim
The FWC is not a costs jurisdiction in the traditional sense. The default position under the Fair Work Act is that each party bears their own costs, with limited exceptions. That structure means that even a claim with no reasonable prospects of success costs an employer to respond to — time, management distraction, potential legal fees, and the conciliation process itself.
Employment lawyers are consistently reporting that employers are settling meritless claims at conciliation to avoid the cost of properly defending them. The settlement is not an admission of liability. It is a commercial decision made under pressure, and AI has significantly increased the frequency with which that pressure arrives.
FWC General Manager Murray Furlong confirmed in May 2026 that the increase in AI-assisted applications has coincided with more litigants choosing to represent themselves, many with no prior workplace relations experience, informed primarily by AI-generated content. Self-represented litigants now represent the overwhelming majority of employee-related claims.
What the Commission and Parliament Are Doing About It
The Fair Work Commission published an exposure draft Guidance Note on 24 March 2026 setting out requirements for the use of generative AI in Commission proceedings. Parties using AI to prepare documents must disclose that AI was used, verify that all facts, case references, and legislative citations are accurate, and confirm in writing that this checking has been completed. Legal practitioners and paid agents must include hyperlinks to all case law cited.
On 3 June 2026, the Workplace Relations Legislation Amendment (Building Cooperative Workplaces No. 1) Bill was introduced in response to the FWC’s workload surge, which the Commission has attributed in significant part to AI-assisted lodgements. The Bill gives the Commission new powers to deal with disputes faster, determine certain matters on the papers without a formal hearing, and restrict individuals who repeatedly bring baseless claims from making further applications.
For employers, the 3 June Bill introduces a practical change requiring attention: general protections and unlawful termination disputes may now proceed to conciliation before any jurisdictional objection is determined. Employers should be prepared to engage in the conciliation process earlier than was previously the case, even where they believe the claim should not have been accepted in the first place.
Why Documentation Is the Only Reliable Defence
Whether the risk arrives as an employee checking entitlements before raising a concern, or as a formally lodged FWC application, the employer’s position depends on the same thing: documentation that accurately reflects the role, the entitlements, and the decisions that were made.
A business with current, correctly classified employment contracts and well-maintained HR documentation has less to fear from employee scrutiny, because there is less to find. It is also better placed to respond to a formal claim, because the evidence exists and the process can be demonstrated.
The businesses most exposed are, as Citation Legal’s research confirms, those relying on informal processes, verbal conversations, and undocumented decisions. That is not a description of businesses doing the wrong thing. It is a description of businesses that grew faster than their people systems — a pattern that is particularly common in the 20 to 75 employee range.
There is a direct bridge between the two blogs in this series. If your employment documentation was generated or substantially assisted by AI and has not been reviewed by an HR professional, that documentation is more likely to contain the classification errors, outdated award references, and missing provisions that AI-assisted employee scrutiny finds first. The employer using AI-drafted documents without professional review is exposed on both fronts simultaneously.
What Australian Employers Should Do Now
Audit employment contracts and classifications against current reality. Not against the role as it was recruited, against the duties the person actually performs today, the current award rates, and the NES obligations that apply. If contracts have not been reviewed in the past 12 to 24 months, assume there are gaps.
Check payroll against current award entitlements before an employee does. The Citation Group data shows 42% of businesses have payroll errors they are not aware of. An employee can identify those errors using AI in under five minutes. The question is whether you find them first.
Document management and conduct decisions in writing at the time they occur. Performance conversations, informal direction, conduct concerns, these need a written record contemporaneous with the event. Documentation reconstructed after a claim is lodged is not a defence. A decision that exists only in a manager’s memory is not evidence.
Train managers on what needs to be recorded and how. The surge in claims is being felt most by businesses where managers are making decisions without HR support and without understanding what requires documentation. That is a training and process gap. It is fixable, but it needs to be fixed before a claim arrives, not after.
Respond to AI-generated claims on their merits, not their method. The FWC is clear: the origin of a claim does not change the employer’s obligation to respond properly. An AI-generated claim that arrives looking polished and authoritative may contain significant errors, but the employer’s response must be evidence-led regardless. Documented decisions, accurate records, and a defensible process are the response — not commentary on how the claim was drafted.
Conclusion: The Standard Has Not Changed, the Speed Has.
AI has not changed what good employment practice looks like. It has changed how quickly gaps in that practice are identified and acted on.
The employers best positioned to manage this environment are not necessarily those who understand AI, they are the ones whose documentation, classifications, and processes are current, accurate, and defensible. An employee scrutinising a well-maintained employment contract with an AI tool will find nothing to act on. A business with accurate payroll, correct classifications, and documented decision-making is in a substantially stronger position when a claim arrives, regardless of how it was generated.
That has always been the standard. The only difference is the cost of falling short of it now arrives faster, more formally, and with greater frequency than it used to.
If your employment documentation has not been reviewed recently, or if it was generated with AI assistance and has not been checked against current Australian law, that review is overdue before an employee’s AI tool does it for you.
Learn more about Strategic HR Australia’s HR Compliance & Risk service or start with a structured review with our quick online compliance audit. Alternatively book a consultation to find out how we can help you and your business.
