Expert Evidence AI Attribution Agent for Perth Litigation Lawyers: Make AI Use in Expert Reports Auditable Before Tender
Your expert is a forensic accountant in West Perth. She used a model to cluster a spreadsheet of transactions, then wrote the opinion herself. The report is annexed to your affidavit and tendered next week in the Administrative Review Tribunal. Opposing counsel asks, on the stand, which parts of the analysis the model produced and which the expert did. She isn’t sure. Neither are you. The Expert Evidence AI Attribution Agent exists so that question has a verifiable answer before the report leaves your office.
The problem
Expert reports increasingly pass through AI tools — for data analysis, transcription, summarisation of source material, or drafting assistance — without a structured record of where the model touched the work. When that record is missing, three things tend to happen. First, the expert can’t accurately disclose AI use when asked, because the use was incidental and undocumented. Second, the practitioner instructing the expert can’t satisfy themselves that the opinion is genuinely the expert’s. Third, the tribunal or court has no basis to weigh the reliability of the model-assisted components against the expert’s independent reasoning.
The Administrative Review Tribunal’s practice directions on expert evidence set expectations for the form and independence of expert opinions tendered in proceedings. Where AI tools have shaped any part of the underlying analysis, transparent attribution is the only way to keep the report’s evidentiary weight intact. The Australian Solicitors’ Conduct Rules (Rule 19) require candour to the tribunal; that duty extends to how AI-assisted material is presented through a solicitor’s tender.
What the Expert Evidence AI Attribution Agent does
The Expert Evidence AI Attribution Agent attaches verifiable attribution metadata to expert evidence outputs. For each artefact the expert produces — a working paper, a chart, a section of narrative, a calculation — the agent records which components were model-generated, which were model-assisted, and which are the expert’s own unaided work. The metadata is bound to the report as a structured appendix that can be inspected by the instructing solicitor, opposing counsel, and the tribunal.
The deliverable is not a substitute for the expert’s independent judgement. It is an audit trail for it.
How it works
- The expert (or their assistant) registers the report draft and any associated working files with the agent before tender.
- The agent extracts the structure of the document — sections, tables, exhibits — and prompts the expert to classify each block: own work, AI-assisted with expert review, or AI-generated for narrative summarisation only.
- Where AI tools were used, the agent records the tool, the date of use, the prompt category, and the expert’s verification step.
- The agent produces an Attribution Appendix in PDF and a machine-readable metadata sidecar, both suitable for annexure to the expert’s affidavit or report.
- The instructing solicitor receives a reviewer summary highlighting any sections flagged “AI-generated” without a recorded verification step, so those can be addressed before the report is finalised.
Why this matters in Perth
Western Australian litigation lawyers who appear in federally constituted tribunals — including the Administrative Review Tribunal sitting in Perth — work to the ART’s published practice directions on expert evidence alongside the rules of the relevant superior court. Where the matter touches a Federal Court proceeding, the General Practice Note on AI (GPN-AI) places direct responsibility on practitioners for the accuracy of AI-assisted material in documents filed. Expert reports filed in support of an application are part of that filed material.
The practical risk is concentrated at the point of tender. An expert who cannot answer attribution questions on the stand puts the weight of their opinion at risk. A solicitor who tenders that report without an attribution record put their own duty of candour in issue. The agent gives both a defensible record produced before the report leaves the expert’s desk — not reconstructed under cross-examination.
Sources
- Administrative Review Tribunal — Practice Directions and Other Guidance: https://www.art.gov.au/help-and-resources/professionals-and-practitioners/practice-directions-and-other-guidance
- Federal Court of Australia — Use of Generative Artificial Intelligence Practice Note (GPN-AI): https://www.fedcourt.gov.au/law-and-practice/practice-documents/practice-notes/gpn-ai
- Law Council of Australia — Australian Solicitors’ Conduct Rules: https://lawcouncil.au/policy-agenda/regulation-of-the-profession-and-ethics/australian-solicitors-conduct-rules
- AustLII (Australasian Legal Information Institute): https://www.austlii.edu.au/
Exegesis capability references:
Join the waitlist
We’re scoping pricing (per-report, per-expert, or firm licence) based on how Perth litigation teams actually want to deploy attribution into their expert workflow. Tell us how your reports flow and what you need the appendix to look like at tender — it will shape the tier you sit in.