Verification Load Balancing Orchestrator for Brisbane In-House Counsel: Make Shadow AI Use Visible Before It Becomes a Conduct Problem

You run legal for a Brisbane-headquartered business. Three of your team are quietly using a consumer AI tool to draft first-pass contract clauses. One of them pasted a counterparty’s term sheet into it last week. Nobody told you. The Australian Solicitors’ Conduct Rules apply to every solicitor on your team regardless of where they work, and the conduct obligations don’t soften because the drafting was “just a starting point”. The Verification Load Balancing Orchestrator is built so that AI-assisted work surfaces into a controlled review pipeline rather than disappearing into private browser tabs.

The problem

Shadow AI use inside an in-house team is rarely malicious. It’s a senior associate trying to clear a Friday backlog. It’s a junior counsel who doesn’t want to ask for the third time how to structure a deed of variation. The drafts come back fast, they look fluent, and they go into matters under the name of a solicitor whose duties under the Australian Solicitors’ Conduct Rules — including the duty to act competently, the duty of confidentiality, and the duty of candour where work product is later relied on — apply to every output that leaves the team.

The structural issue is that verification work (a partner or senior reviewer actually reading the AI-assisted draft against source documents) is not load-balanced. It piles up on whoever happens to be the most diligent reviewer, gets skipped under deadline pressure, or never gets assigned at all because nobody flagged the draft as AI-assisted in the first place. The ASCR don’t prohibit AI use. They make the solicitor responsible for the output. A team without a verification pipeline is a team that can’t evidence the responsibility.

What the Verification Load Balancing Orchestrator does

The Verification Load Balancing Orchestrator distributes verification workload across partners and reviewers. It sits between drafting (whatever tool produced the draft) and lodgement or release, and routes each AI-assisted artefact to a reviewer based on three signals: the reviewer’s current verification queue depth, their domain match for the matter, and the risk tier of the artefact itself. The orchestrator does not generate legal content. It does not store client material outside your configured environment. Its job is allocation, queue management, and audit — turning shadow AI use into observed, reviewed AI use.

It pairs with RuleCheck, Exegesis’ open-source citation verifier (github.com/andrefabre/rulecheck), so any draft routed through the orchestrator can be checked deterministically for fabricated authorities before it reaches a human reviewer’s queue.

How it works

  1. Intake. A team member submits any AI-assisted draft (clause, memo, advice note, submission) through the orchestrator. Submission is the only sanctioned path — the policy your GC sets is that AI-assisted work without an orchestrator ticket is not approved work.
  2. Risk tiering. The orchestrator classifies the artefact by matter type, counterparty exposure, and whether it will be filed, sent externally, or used internally only. This determines the verification depth required.
  3. Reviewer routing. The orchestrator assigns the artefact to a reviewer with available queue capacity and domain fit. A partner buried in a transaction does not get routed a fourth contract review at 4pm; a senior reviewer with capacity does.
  4. Deterministic checks first. Before the human reviewer opens the artefact, RuleCheck runs over any cited authorities and flags fabricated, mismatched, or unverifiable citations. The reviewer sees a clean shortlist of what actually needs their judgment.
  5. Audit trail. Each artefact carries a record: who drafted it, what tool assisted, who reviewed it, what was changed, and when it was released. This is the evidence your GC needs to show the team is meeting its ASCR obligations.

Why this matters in Brisbane

Queensland adopted the Australian Solicitors’ Conduct Rules in June 2012, and they remain the applicable conduct framework for solicitors practising in Queensland — including in-house counsel admitted in Queensland. The ASCR are, in the Law Council’s description, “a statement of professional and ethical obligations derived from solicitors’ duties as an officer of the court, the common law and equity, legislation, and the collective judgment of the legal profession”. Those obligations attach to the individual solicitor. They do not soften when the solicitor is employed in-house rather than in private practice, and they do not soften when a draft was produced with the help of a generative AI tool the GC didn’t know about.

For a Brisbane in-house team, the practical question is not whether AI is being used. It is whether the team can show, on any given matter, who reviewed the AI-assisted output and against what. The Verification Load Balancing Orchestrator is built to make that answer trivial to produce.

Sources

Join the waitlist

The Verification Load Balancing Orchestrator is in design with a small cohort of Australian in-house teams. We’re scoping the right deployment model (managed service, self-hosted, or hybrid) based on what GCs actually need to govern AI use across their teams.

Join the waitlist — be the first to know when access opens for Brisbane in-house counsel