Client/Firm Privilege Boundary Orchestrator for Perth Boutique Firms: Stop Privilege Bleeding Between Matters

You run a ten-person firm in West Perth. Same partners, same paralegals, sometimes the same junior drafting a witness statement for one matter in the morning and a submission to the Administrative Review Tribunal for an unrelated client in the afternoon. The team has started using an AI assistant for first drafts, summarisation, and clause extraction. Nobody can tell you with certainty which client’s material has been pasted into which prompt, which session retained context, or whether the model’s working memory from yesterday’s expert brief is now bleeding into today’s. That’s the privilege boundary problem, and it’s the one a boutique firm wears more sharply than a 200-lawyer practice with a dedicated CIO.

The problem

In a small firm everyone touches everything. The same person who drafts for Client A also drafts for Client A’s commercial counterparty next month. When AI tools enter the workflow, the privilege boundary that used to live in physical folders and Chinese walls now has to live inside prompt history, model context windows, and shared workspace logs. Three things go wrong:

Rule 9 of the Australian Solicitors’ Conduct Rules requires solicitors to maintain client confidentiality. Rule 11 governs conflicts between current clients. Neither rule was written with shared model context in mind, but both apply unchanged.

What the Client/Firm Privilege Boundary Orchestrator does

The Client/Firm Privilege Boundary Orchestrator is an Exegesis service shape that manages the privilege boundary between client matters and between client work and firm-internal AI use. It is a control layer, not a chatbot. It sits between your team and whatever AI tools the firm uses, and it does three things: it tags every input with a matter ID and a privilege classification, it enforces session isolation so context from Matter A cannot leak into Matter B, and it produces a per-matter audit log that records which content was processed by which model under which client’s authority.

For a boutique firm, the practical outcome is that a partner can answer the question “has any of Client A’s privileged material ever been put through an AI system in a session that also touched Client B’s matter?” with evidence, not a shrug.

How it works

  1. Matter registration. Each active matter is registered with a privilege classification (client-privileged, firm-internal, public). New matters inherit a default isolation policy.
  2. Session binding. Every AI interaction is bound to exactly one matter ID. The orchestrator refuses prompts that contain material tagged to a different matter unless an explicit cross-matter waiver is recorded.
  3. Context isolation. Model sessions are scoped per matter. Context windows are flushed at session end. Firm-internal sessions (precedent drafting, knowledge management) are kept on a separate channel from client-work sessions.
  4. Expert evidence carve-out. Matters flagged as involving ART or court expert evidence are placed under a stricter policy: no firm-internal precedent pull, no cross-matter context, and a per-prompt log suitable for disclosure if the expert is asked to demonstrate the independence of their opinion.
  5. Audit log. A per-matter, per-user log is generated showing every model interaction, the content classification, and any policy overrides. The log is the artefact that lets you answer a conflict enquiry.

Why this matters in Perth

Perth boutique firms tend to specialise — resources, native title, family, migration, planning — and the pool of opposing parties and instructing experts is small. A migration practice appearing before the Administrative Review Tribunal will see the same expert witnesses, the same departmental representatives, and often the same applicant categories week after week. The ART’s practice directions on expert evidence require that opinions placed before the Tribunal be the expert’s own and that the basis for the opinion be transparent. Where AI assistance has been used in preparing material that accompanies or informs expert evidence, the firm needs to be able to show the boundary held. The same logic applies in the Federal Court under GPN-AI, where practitioner responsibility for AI-assisted material is explicit. A boutique firm doesn’t have a separate information governance team to construct this after the fact — the orchestrator is built to provide it as a default.

Sources

Join the waitlist

Join the waitlist — be the first to know when the Client/Firm Privilege Boundary Orchestrator opens for Perth boutique firms

The orchestrator is in design with a small group of Australian boutique firms. We are scoping per-seat and per-matter pricing models and want input from firms under ten lawyers on what would actually work inside a partnership of your size. Join the waitlist and we’ll be in touch as access opens.