Would you like AI with that? Lessons from Rogers v McDonald's Australia Ltd (AI-use) [2026] FCA 1264

In Rogers v McDonald's Australia Ltd (AI-use) [2026] FCA 1264 (Rogers), Lee J directed the parties to investigate and confer on the use of AI technology in preparing a class action for mediation and trial. The direction appears to be the first of its kind under section 37P(2) of the Federal Court of Australia Act 1976 (Cth).

The Court did not require the parties to use AI. Rather, it required them to investigate whether AI could assist with particular tasks and, in doing so, identified four requirements for its use: traceability, testability, confidentiality and human supervision.

Why AI became relevant in Rogers

Rogers is a class action brought by the Shop, Distributive and Allied Employees Association against McDonald's Australia Ltd and numerous franchisee respondents. The proceeding alleges that managers were routinely required to perform unpaid "pre-shift" and "after-shift" work.

The proceeding has been referred to mediation. Preparing for that mediation will require the respondents to provide detailed information about group members, including their employment circumstances, classifications, shift records and remuneration (Primary Information).

The difficulty is that the Primary Information is dispersed across business records maintained in different repositories and formats. The practical task is to collect, organise and analyse that information in a form that can be used by the parties and the Court.

It was in that context that the Court considered the potential use of AI.

What did the Court direct the parties to do?

Among other things, the orders require the parties to confer with the assistance of a suitably qualified technology expert on the suitability of using AI technology in preparing the matter for mediation and trial, specifically for use in:

• collating and analysing quantification data for mediation;

• preparing chronologies and summaries of primary documents, including by reference to specified issues; and

• managing and interpreting documents and other data relevant to the proceeding more generally.

The orders were made under section 37P(2) of the Federal Court of Australia Act, which gives the Court a broad power to make directions about the practice and procedure to be followed in a proceeding.

This appears to be the first time section 37P(2) has been used to direct parties to investigate the adoption of AI technology in litigation preparation.

Importantly, Lee J did not prescribe a particular technology or require the parties ultimately to use AI. The immediate task is to investigate its suitability and confer about its potential use.

What did the Court say about AI?

Lee J's broader observations about the role of AI in litigation are significant.

His Honour recognised that AI allows the legal profession to reconsider processes developed in a world constrained by human capacity to collect, collate and synthesise information. The "lodestar", his Honour observed, should be to use AI to allow lawyers to spend more time exercising professional judgment, rather than performing mechanical tasks of collection and synthesis that technology can perform reliably and at a fraction of the cost.

That does not mean entrusting adjudicative tasks to AI. The objective is to use technology for work that does not require human judgment, leaving lawyers and the Court to concentrate their time on the work that does.

His Honour identified four requirements for any technological process adopted in the proceeding:

1. Traceability: it must preserve traceability to the primary records.

2. Testability: its outputs must be capable of being tested and corrected.

3. Confidentiality: it must protect privileged and confidential material where necessary.

4. Human supervision: it must remain subject to appropriate human supervision.

Lee J did not prescribe any particular technology or provider, but noted that existing commercial AI tools are likely to provide a satisfactory solution. What matters is that the use of AI "be examined seriously, cooperatively and with an eye to the overarching purpose, rather than dismissed because the conventional alternative is known".

Why does Rogers matter?

Much of the recent response by Australian courts to generative AI has focused on managing the risks associated with its use. Court practice notes and guidance have addressed issues including the accuracy of AI-generated material, confidentiality, disclosure of AI use and the continuing responsibility of practitioners for material put before the Court.

Rogers is notable because rather than responding to the misuse of AI, the Court required the parties to investigate whether AI could assist with particular aspects of preparing the proceeding for mediation and trial. It treats AI as a potential case-management tool, rather than focusing only on the risks associated with its use.

The decision does not establish a general obligation on litigants or practitioners to use AI, nor does it suggest that AI will be appropriate in every proceeding. Rogers does, however, show that in an appropriate case a court may require parties to investigate whether technology can assist with particular litigation tasks.

That is particularly relevant to document and data-intensive proceedings. Class actions are an obvious example. They can involve large volumes of documentary material, information held across different systems and substantial exercises in identifying, organising and quantifying group member claims. The Primary Information in Rogers illustrates the point.

The direction also shows that these issues may sometimes need to be addressed cooperatively. In Rogers, the parties are required to confer with the assistance of a suitably qualified technology expert. Questions about how large bodies of documents or data will be processed may therefore become part of broader case-management discussions between parties in appropriate proceedings.

What should practitioners consider?

Lee J's four requirements provide a useful framework for considering AI-assisted litigation beyond the particular circumstances of Rogers:

Traceability. Where AI is used to prepare a chronology, summary or other analysis, practitioners should consider how the output can be traced to the underlying source material. A chronology entry, for example, should be capable of being linked to the document or record from which it was derived. The level of documentation required will depend on the task and the purpose for which the output will be used. The concept of traceability and testing of AI-generated outputs also extends to verification of outputs that may appear comprehensive but in fact omit relevant material or fail to surface key aspects of the underlying records. An often underappreciated limitation of AI is that it does not guarantee exhaustive coverage of source material. An AI tool may analyse a document set and return an output that appears complete and well-reasoned, but which silently omits relevant information or considerations. This is not the same as hallucination; the output is not fabricated, but it may be incomplete in ways that are not apparent from the output itself. Where each element of an output can be traced to its source, it becomes easier to identify gaps in coverage and to verify that the analysis is complete, rather than simply plausible.

Testability. AI-generated work should be capable of being checked against the underlying material. For large-scale exercises, this may involve sampling or other quality assurance processes. More extensive verification may be appropriate where an output will underpin evidence, quantification or submissions to the Court.

Confidentiality. Before client material is provided to an AI system, practitioners need to understand how the provider will handle that information, including where it will be processed and stored, whether it will be retained or used for other purposes, and who may have access to it. Those matters may also be relevant to confidentiality and legal professional privilege.

Human supervision. AI can assist with collection, organisation and synthesis, but responsibility for the legal work remains with the practitioner. The level of review required will depend on the task and how the output is to be used.

The last of these requirements goes to an important theme in Lee J's reasons. Used appropriately, AI does not replace professional judgment. Its value lies in reducing the amount of time lawyers spend on mechanical work so that more time can be directed to analysis, strategy and judgment.

Where to from here?

Rogers does not establish that AI should be used in every complex proceeding. But it does suggest that, where litigation involves large volumes of documents or data, parties may increasingly need to consider whether technology can perform some of that work more efficiently.

Whether AI is suitable will depend on the task and Rogers makes clear that any efficiency gained must be accompanied by traceability, testability, confidentiality and appropriate human supervision. But the development signals a broader shift: in data-intensive litigation, the question is no longer whether to use AI but how that technology can be used responsibly.