Voice Preservation & Procedural Fairness Agent for Perth Boutique Firms: Keep Client Voice Intact When AI Helps Draft Submissions

You run a six-lawyer firm in West Perth. An ART matter is due Friday. A junior used an LLM to “tidy up” a witness statement and a statement of facts and contentions. When you read the draft back, it’s grammatically cleaner — and it doesn’t sound like your client anymore. Phrases the client never used. Confidence the client doesn’t hold. A subtle shift in the chronology. You now have two problems: a procedural fairness problem if the Tribunal hears words your client didn’t say, and a disclosure problem under the ART’s expectations around AI use. The Voice Preservation & Procedural Fairness Agent is built for exactly this drafting workflow.

The problem

Boutique firms don’t have the bench depth of a national practice. One paralegal and two solicitors might cover a whole ART matter from intake to hearing, and AI drafting tools are now part of how the work gets done — there isn’t a realistic alternative when the time budget is what it is. The risk is not AI use itself; the risk is that AI drafting silently rewrites the client’s voice, smooths over factual ambiguity the client actually expressed, or introduces material the client never instructed on. When that document goes to the Tribunal:

• The Administrative Review Tribunal expects practitioners to comply with its practice directions and other guidance, including expectations around the use of generative AI in materials filed.
• Australian Solicitors’ Conduct Rule 19 requires solicitors to be candid with the court (and tribunals where applicable) and not mislead it — which includes not presenting AI-generated content as the client’s own evidence.
• ASCR Rule 8 requires solicitors to follow a client’s lawful, proper and competent instructions — which is harder to demonstrate when a draft has been through a model that rewrites freely.

For a boutique firm without a dedicated AI governance function, the disclosure and voice-preservation work has to happen inside the drafting tool, not as a separate audit step nobody has time for.

What the Voice Preservation & Procedural Fairness Agent does

The agent sits between the AI drafting tool and the filed document. Its job is narrow and deterministic:

• Preserve verbatim segments of the client’s own words (from instructions, statements, file notes) and flag any AI-rewritten version of those segments for human review before it lands in a submission.
• Track which paragraphs of a draft are AI-assisted, which are lawyer-authored, and which are direct client voice — producing a per-paragraph provenance record.
• Generate the AI-use disclosure statement the matter will need on filing, so it isn’t an afterthought.
• Flag procedural-fairness risks: new factual assertions introduced by the model that don’t trace back to instructions, evidence, or pleadings.

It is a check on the drafting output, not a replacement for the drafting judgment.

How it works

1. Anchor the client voice. You load the source material — client statements, file notes, instructions, transcripts — into the agent as the “voice anchor” set. These are the segments that must not be silently rewritten.
2. Run your draft through the agent. Upload the working submission. The agent segments it paragraph by paragraph.
3. Provenance pass. Each paragraph is classified: direct client voice, lawyer-authored, AI-assisted, or AI-generated. Rewritten client-voice segments are surfaced for explicit lawyer approval or reversion to verbatim.
4. Procedural-fairness pass. The agent flags factual assertions in the draft that don’t have a traceable source in the anchor set, so you can either remove them or instruct further.
5. Disclosure artefact. The agent produces a draft AI-use disclosure paragraph and a per-paragraph provenance log you can retain on the matter file.

Why this matters in Perth

Perth boutique firms tend to carry mixed practices — migration, NDIS reviews, social services, veterans’ affairs, small commercial — and a meaningful share of that work ends up at the Administrative Review Tribunal. The ART’s published practice directions and guidance materials set the expectations practitioners are working to, and AI-use disclosure is increasingly part of that frame. A six- or eight-lawyer firm filing into the ART can’t run a separate AI-governance workstream the way a tier-one firm can; the controls have to be embedded in the drafting workflow itself, or they won’t happen. That’s the gap this agent is built for.

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
• Law Council of Australia — Australian Solicitors’ Conduct Rules: https://lawcouncil.au/policy-agenda/regulation-of-the-profession-and-ethics/australian-solicitors-conduct-rules
• 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
• AustLII (Australasian Legal Information Institute): https://www.austlii.edu.au/

Join the waitlist

Join the waitlist — be the first to know when access opens for Perth boutique firms

The Voice Preservation & Procedural Fairness Agent is in build. We’re scoping pricing tiers around the way boutique firms actually run AI-assisted drafting — per-matter, per-seat monthly, or firm licence. Join the waitlist and we’ll let you know when access opens; what we hear from you shapes how the tier you sit in actually works.