Expert Report Provenance Agent for Brisbane Litigation Lawyers: Know Which Sentences of Your Expert Report a Model Wrote
Your expert finalised their report on Friday. It reads well — too well, in places. You know they used a transcription tool for the interview notes, and you suspect a model helped draft the executive summary. On Monday at the ART directions hearing, opposing counsel asks the expert whether any part of the report was generated by AI, and which parts. The expert hesitates. So do you. The Expert Report Provenance Agent is built so neither of you has to.
The problem
Expert evidence in Administrative Review Tribunal matters — and in Queensland Supreme Court proceedings where the expert may be cross-examined on the same report — turns on the independence and authorship of the expert’s opinion. Experts are increasingly using language models to draft, summarise, or polish sections of their reports. When that use is undisclosed, or when no one can say with confidence which paragraphs came from the expert’s own analysis versus a model’s reformulation, two things follow: the report’s weight is open to challenge on attribution grounds, and the instructing solicitor’s candour obligations under the Australian Solicitors’ Conduct Rules become difficult to discharge.
The Administrative Review Tribunal publishes Practice Directions and guidance for professionals and practitioners that govern how expert evidence is to be prepared and presented, including expectations around the basis of the expert’s opinion. A report whose authorship cannot be reconstructed paragraph-by-paragraph is a report whose foundation is contestable.
What the Expert Report Provenance Agent does
The Expert Report Provenance Agent tracks the provenance of every section of an expert report — what was AI-generated, what was human-edited, and what was authored from scratch by the expert. For each paragraph (or finer-grained span) in the final document, it records:
- Source: model-drafted, model-assisted (light edits), human-authored, or human-rewritten
- The point in the workflow where the text entered the document
- Edit history sufficient to show how the expert’s own analysis sits behind the final wording
- A disclosure-ready summary suitable for inclusion in the report or for tendering to the Tribunal on request
The deliverable is a provenance log attached to the matter, plus a one-page authorship statement the expert can sign and the instructing solicitor can rely on when answering questions about how the report was prepared.
How it works
- Ingest the working drafts. The expert (or their assistant) provides the draft versions of the report as they evolve — including any model-assisted drafts and any clean human rewrites.
- Span-level diffing. The agent runs deterministic diffs across the draft sequence, classifying each span of the final report against its earliest source: model output, human edit on model output, or human-authored.
- Provenance ledger. A structured ledger is produced, mapping each paragraph of the final report to its authorship classification and edit chain.
- Authorship statement. A plain-English summary is generated for the expert to review and sign — covering which sections used model assistance, the nature of that assistance, and confirming that the opinions expressed are the expert’s own.
- Archive alongside the matter file. The ledger and statement are exported as markdown for storage with the report, so they can be produced if attribution is challenged at hearing.
Why this matters in Brisbane
Brisbane litigation lawyers running ART matters, or Queensland Supreme Court proceedings that draw on the same expert pool, are operating in a window where Tribunal and court expectations around AI disclosure in expert evidence are tightening but the tooling experts use is loosening. The ART’s published Practice Directions and practitioner guidance set the framework for how expert evidence is to be prepared and the basis on which it is given. The Australian Solicitors’ Conduct Rules require candour to the court and the Tribunal, which extends to the provenance of evidence the practitioner tenders.
An expert who cannot answer “which parts of this did a model write?” is an expert whose evidence is exposed. An instructing solicitor who cannot answer the same question on the expert’s behalf has a candour problem of their own. The provenance ledger closes both gaps before the directions hearing, not after.
Sources
- Administrative Review Tribunal — Practice Directions and other guidance for professionals and practitioners: https://www.art.gov.au/help-and-resources/professionals-and-practitioners/practice-directions-and-other-guidance
- Law Council of Australia — Australian Solicitors’ Conduct Rules: https://lawcouncil.au/policy-agenda/regulation-of-the-profession-and-ethics/australian-solicitors-conduct-rules
- RuleCheck by Exegesis (open-source provenance and verification tooling): https://github.com/andrefabre/rulecheck
Join the waitlist
The Expert Report Provenance Agent is in build. Brisbane litigation teams running ART matters or Supreme Court proceedings with expert evidence can join the waitlist to shape the workflow — particularly how the agent integrates with the way your experts already draft.